BioNTech SE - ADR Aktie

Welcome to BioNTech's Second Quarter 2026 Earnings Call. I will hand the call over to Doug Maffei, Vice President, Strategy and Investor Relations. Please go ahead.

Thank you, operator. Good morning and good afternoon. Thank you for joining BioNTech's Second Quarter 2026 Earnings Call. As a reminder, the slides we will be using during this call and the corresponding press release can be found in the Investors section of our website.

On the next slide, you will see our forward-looking statements disclaimer. Additional information about these statements and other risks are described in our filings with the U.S. Securities and Exchange Commission, or SEC. Forward-looking statements on this call are subject to significant risks and uncertainties and speak only as of the date of this conference call.

We undertake no obligation to update or revise any of these statements. On Slide 3, you can see the agenda for today's call. I'm joined by the following members of BioNTech's management team: Ugur Sahin, Chief Executive Officer and Co-Founder; Özlem Türeci, Chief Medical Officer and Co-Founder; and Ramon Zapata, Chief Financial Officer.

Also available for the Q&A portion of the call is Annemarie Hanekamp, Chief Commercial Officer. Related to yesterday's Chief Executive Officer announcement, we will also be joined today by Helmut Jeggle, Chairman of BioNTech's Supervisory Board. With this, I will hand the call over to Helmut.

Thank you, Doug, and good morning, everyone. Before we begin with a business update from the Management Board, I would like to provide further color on the appointment of BioNTech's next Chief Executive Officer. As announced yesterday, Guido Oelkers will take office as CEO as of February 1 at the latest. From the outset, the Supervisory Board's CEO search was guided by 3 clear priorities: proven strategic leadership, the ability to scale a global biopharmaceutical business and a strong track record of building and growing innovation-driven, science-based organizations.

Guido is an excellent fit on all 3 dimensions. Most recently, as CEO of Sobi, he more than quadrupled the company's revenues over 9 years by strengthening its global capabilities and maximizing the value of his late-stage pipeline. He brings deep expertise in launching and commercializing innovative products with a focus on the U.S. market as well as extensive leadership experience across Europe and Asia Pacific.

The Supervisory Board believes that Guido is the right leader for BioNTech's next phase. His expertise in scaling innovative organizations in a focused and capital-efficient manner, combined with his deep knowledge of the markets most relevant to BioNTech will position the company well to deliver on its key objectives, evolve into a multiproduct biopharmaceutical company and continue its remarkable success story.

With that, I would like to hand over to Ugur and will be available for questions during the Q&A at the end of the call.

Thank you, Helmut, and a warm welcome to everyone joining us today. I believe it is important to note that this transition reflects the natural evolution of BioNTech from a pioneering research organization into a global biopharmaceutical company with multiple commercial products. The next phase of this evolution requires the corresponding leadership skills, and I am confident we have found this in Guido.

During our exchanges, I have come to know Guido as a leader who combines a deep understanding of the pharmaceutical industry and strategic acumen with genuine respect for the culture and people of our organization. He understands what we have built and importantly, he understands what it will take to scale it. To ensure continuity and a seamless transition, I will remain actively engaged in supporting the preparations for Guido's onboarding.

As for BioNTech's next phase, our mission remains constant to translate science into survival. To achieve this, BioNTech has successfully built a diversified toolkit of modalities, including next-generation immunomodulators, ADCs and mRNA cancer immunotherapies. Our multiproduct portfolio has progressed further and growing share of its now in late-stage clinical development and pivotal trials. The second quarter was a period of significant progress for BioNTech towards this.

First, we are accelerating the late-stage development of our oncology assets. We shared encouraging global data in first-line NSCLC from the Phase II portion of our Phase II/III trial at ASCO from our potential next-generation I/O backbone Pumitamig. Second, our combination therapy strategy is gaining momentum. We have expanded our novel, novel combination programs and presented data from ongoing combination trials with our ADCs with more to come soon.

Third, we continue our shift from platform-centric to a tumor-centric clinical development approach around cancers with greatest unmet medical need. Notably, in our GU tumor area, we dosed the first patient in our Phase III trial, evaluating our B7-H3 ADC elfetabart drozuntecan, formerly known as BNT324 in metastatic castration-resistant prostate

cancer.

BioNTech is well positioned for the next phase. With a growing pipeline of potentially registrational trials, strong partnerships and financial strength, we are on track to become a diversified, multiproduct company by 2030. We are targeting more than 17 late-stage and pivotal trial readouts through 2030 and beyond, spanning multiple tumor types and different lines of treatment. Our progress to date sets us up for an impactful second half of 2026. We enter the remainder of this year with momentum and diligent execution as we continue to progress towards our long-term vision.

With this, I will hand over to Esther for an update on our oncology execution.

Thank you, Ugur. I'm glad to be speaking with everyone today. Our ambition is to address the full continuum of cancer, utilizing the approaches that Ugur outlined. We have defined a tumor-focused strategy to address significant unmet medical need where our novel combinations can extend survival outcomes for patients and maximize the potential of our pipeline.

As such, we are advancing multiple assets from our multimodal oncology pipeline into late-stage development. During the first half of 2026, we made progress across our pipeline, and I'll cover some of these updates today. I'll begin with lung cancer, which is one of the cancers of highest unmet medical need and the tumor area where we have the broadest and most diverse coverage. We are aiming to tackle unmet medical needs at every stage of the lung cancer patient journey.

Our lung cancer strategy covers various disease stages, settings and enlists various modalities, next-generation immunomodulators, ADCs and mRNA cancer immunotherapy. For certain settings, we have multiple opportunities with the aim to change the standard of care and move forward with our combination strategy.

At the core of this tumor-based oncology strategy is pumitamig, our investigational bispecific immunomodulator targeting PD-L1 and VEGFA, now in development with our partner, BMS. In lung cancer, we are now running 4 registrational programs for Pumitamig, ROSETTA LUNG-01 in first-line extensive stage small cell lung cancer and ROSETTA LUNG-02 in first-line non-small cell lung cancer, where global Phase II data were presented at ASCO.

ROSETTA LUNG-202, our pivotal trial in first-line PD-L1 high non-small cell lung cancer is now enrolling and ROSETTA LUNG-201, our pivotal trial in unresectable Stage III non-small cell lung cancer is also underway. Moreover, we are generating novel-novel combination data to inform the first wave of combination trials with registrational intent.

Zooming in on pumetamig, here, we again have pioneered by delivering the first global Phase II data for PD-L1 VEGF bispecific in first-line non-small cell lung cancer. At ASCO in June, we presented Phase II data from ROSETTA LUNG-02, our global randomized Phase II/III trial evaluating pumetamig in combination with chemotherapy in patients with previously untreated advanced non-small cell lung cancer.

In 40 evaluable patients with both squamous and non-squamous histology, pumetamig, plus chemotherapy demonstrated robust clinical activity with unconfirmed and confirmed overall response rates for combined doses of 72.5% and 62.5%, respectively. Two features of these data deserve particular emphasis. First, the encouraging activity observed across PD-L1 expression levels is noteworthy.

Second, the particularly strong response rate in PD-L1 low disease across histologies. This includes patients with PD-L1 TPS less than 1% who represented approximately 58% of patients in this cohort, a subgroup typically with poor response to anti-PD-1, PD-L1 treatment. In that population, the confirmed objective response rate was 47.6%. In patients with TPS between 1% and 49%, it was 77.8% and all 6 patients with TPS 50% or above responded.

The safety profile was manageable in both histologies with no new safety signals. These data support the ongoing global Phase III program for pumetamig in lung cancer. The robust clinical responses across PD-L1 strata aligned with our expectations. It speaks to the potential of pumetamig to confer benefit in the all-comer patient population, including the PD-L1 low expression levels where unmet medical need is high.

The ROSETTA LUNG-02 trial is currently recruiting in its Phase III portion, and we look forward to presenting additional Phase II data from this trial as the data mature. The central question in the PD-1, PD-L1 VEGF class has been whether the clinical activity observed in trials conducted in China would be consistent with the data in global populations.

We have been able to address that question with our own asset across 3 key tumor types in Phase II trials. Firstly, in first-line small cell lung cancer. The China Phase I/II demonstrated a disease control rate of 94% and a confirmed overall response rate of 82%. With the global Phase II trial, we showed a disease control rate of 100% and a confirmed objective response rate of 76%. Then there is the first-line non-small cell lung cancer indication.

We observed in the China monotherapy Phase I/II, a confirmed objective response rate of approximately 47% in PD-L1 positive patients. In the global Phase II pumetamig plus chemotherapy, which includes the PD-L1 unselected population, we saw a confirmed objective response rate of approximately 63%. In TNBC, the disease control rate was 92% in both the trials conducted in China and globally. In the China Phase I/II in first-line TNBC, we reported a confirmed objective response rate of approximately 74% and in the global Phase II cohort, which included a heterogeneous population of first- and second-line patients, we reported a confirmed objective response rate of approximately 62% as expected given the treatment line population.

Across these 3 tumor types of high unmet need, we are observing a meaningful consistency between the data generated in China and globally. While cross-trial comparisons must be interpreted with caution, we are encouraged by this cross-regional consistency. This gives us increased confidence in the global potential of pumitamig. We are advancing multiple pivotal Phase III programs to confirm these signals. As covered here on our tumor map slide, we are deploying multiple modalities to tackle lung cancer. Gotistobart is a critical component of [indiscernible].

As a reminder, gotistobart is our selective Treg depleting antibody targeting CTLA-4 developed in collaboration with our partner, OncoC4. We are advancing gotistobart through the pivotal second stage of ESERVE-003, our global Phase III in patients with metastatic squamous non-small cell lung cancer who progressed following platinum-based chemotherapy and PD-L1 inhibitor treatment.

This is a setting with very few effective options and poor prognosis. gotistobart differentiated mechanism of selectively depleting regulatory T cells in the tumor microenvironment is designed to reengage the immune system even after prior checkpoint inhibitor exposure.

Earlier this year, at ELCC, we presented updated data from the nonpivotal Stage 1 of PRESERVE-003, our global Phase III trial. The data are very encouraging. The 12-month PFS rate of 25% for gotistobart versus 0 for docetaxel is a signal of durable disease control. Gotistobart reduced the risk of death in this IO pretreated patient population by 54% compared to docetaxel with a hazard ratio of 0.46.

The median OS in the gotistobart arm has not yet been reached compared to approximately 10 months with docetaxel. At 12 months, 63% of patients treated with gotistobart were alive versus 30% in the docetaxel arm. The safety profile was consistent with the previously established profile for gotistobart with no new signals of concern.

We expect to present longer follow-up data at the World Lung Conference next month. Based on current event accrual projections, we expect to conduct the first interim analysis from the pivotal stage of the trial towards the end of this year. At ASCO this year, we presented overall survival data from the Phase II study evaluating gotistobart in combination with pembrolizumab in ovarian cancer patients who had received prior platinum-based chemotherapy.

The data showed a compelling and differentiated signal with a median overall survival of 18.9 months. Together, these data reinforce the potential of gotistobart to provide an extended overall survival benefit and serve as a potential chemo-free treatment option for lung patients. I will now turn to [indiscernible] Elfe-D, our or BNT324, our B7-H3 targeted ADC developed in collaboration with Duality Bio. B7-H3 is overexpressed across multiple tumor types, including prostate cancer, non-small cell lung cancer, small cell lung cancer and others.

The target biology, combined with the pharmacology of a topo1 inhibitor ADC with drug-to-antibody ratio of 6 positions Elfe-D as a potentially versatile oncology asset across a wide range of solid tumors. More than 1,000 patients have now been treated with Elfe-D across more than 10 tumor types, including 400 patients treated with Elfe-D in combination with pumitamig. The growing body of clinical evidence demonstrates antitumor activity across multiple indications with a favorable safety profile.

This quarter, we dosed our first patient in the Phase III clinical trial for Elfe-D, evaluating it against docetaxel in patients with [indiscernible] metastatic castration-resistant prostate cancer. Prostate cancer is our first Phase III indication for Elfe-D, and it represents one of the strongest B7-H3 expression profiles of any tumor type.

The trial targets a patient population with substantial unmet need following progression on second-generation androgen receptor pathway inhibitors. In parallel, Elfe-D is being evaluated in combination with pumitamig across multiple Phase I/II programs. These results will help inform the optimal clinical design for upcoming registrational combination trials. We expect to present some of these data at a medical conference later this year.

Moving now to our portfolio of innovative mRNA cancer immunotherapies, which aim to activate and educate the immune system with precision. Our individualized neoantigen-specific immunotherapy autogene cevumeran developed in collaboration with Genentech is advancing in 2 ongoing randomized Phase II trials. In adjuvant ctDNA Stage 2 high-risk or Stage III colorectal cancer, we have a Phase II trial evaluating autogene Cevumeran monotherapy against the standard of watchful waiting.

Enrollment is now complete and in June, an interim analysis based on the centrally assessed primary endpoint of disease-free survival was reviewed by the independent Data Safety Monitoring Board with a recommendation to continue the trial without modification. Thus, the study will continue as standard protocol, and we will remain masked to the data until the final analysis.

The data readout from the final analysis of this trial is event-driven and expected in 2027. In adjuvant pancreatic cancer, recruitment for the Phase II IMCODE003 trial is well underway. Data from a Phase I investigator-initiated trial, including a 6-year update presented at AACR this year, continue to demonstrate durable immune responses against the encoded neoantigens for up to 6 years, evidence that supports our therapeutic rationale in the adjuvant and minimal residual disease setting.

On our FixVac platform, BNT113, our off-the-shelf HPV16 targeting immunotherapy is advancing in the [indiscernible] MERIT Phase II/III trial in combination with pembrolizumab as a first-line treatment for patients with PD-L1 positive HPV16 positive head and neck squamous cell cancer. A Phase III interim analysis is expected for PFS this year. For BNT116, our mRNA immune therapy targeting multiple non-small cell lung cancer-associated antigens, we expect to present data at WCLC 2026 from Cohort 6 in combination with cemiplimab and chemotherapy.

These programs reflect our conviction that mRNA cancer immunotherapy, particularly in combination with checkpoint inhibition can deliver meaningful benefit in defined patient populations. In closing, today's review underscores the significant progress across our portfolio. We have reached multiple key milestones and continue to execute on plan in 2026 and beyond.

Within our late-stage programs, we anticipate 3 further readouts this year. gotistobart in squamous non-small cell lung cancer, our FixVac immunotherapy BNT113 in head and neck cancer and TPAM in breast cancer. For gotistobart, we expect the first interim analysis in late 2026 based on the projected event accrual rates. This initial review by the independent data monitoring committee is intended as an early checkpoint before the next preplanned interim analysis.

For BNT113, based on current event accrual projections, we expect a Phase III interim analysis for progression-free survival later this year. Overall survival, which is the trial's other co-primary endpoint is not expected to be mature at this interim. TPEM is currently being advanced in 2 pivotal clinical trials, one in second-line endometrial cancer and one in HER2 low hormone receptor positive metastatic breast cancer.

The candidate has generated encouraging data to date in both indications. With the primary analysis in breast cancer expected in the fourth quarter of 2026, we will determine the optimal regulatory pathway based on aggregated data across both indications. With this data-driven approach, we aim to pursue a value optimization strategy for TPEM in an evolving treatment landscape while prioritizing opportunities where we can deliver significant benefit for patients.

Following our midyear review of upcoming late-stage milestones, we have updated the expected timing for the Phase III pumitamig trial in triple-negative breast cancer in China and for the Phase II [indiscernible] trial in second-line castration-resistant prostate cancer, both of which are now expected in 2027.

With regards to our earlier stage novel-novel readouts, we have already published data on some of those combinations and expect more soon. I want to highlight one data set that is strategically significant for our ambitions in lung. the upcoming readout for the Phase I/II trial evaluating pumitamig3 in combination with Elfe-D, our B7-H3 ADC, across advanced non-small cell lung cancer and small cell lung cancer.

This will be the first clinical data for our PD-L1 VEGFA bispecific antibody in combination with an antibody drug conjugate in lung cancer. The combination brings together 2 mechanistically distinct and potentially synergistic approaches, the immune reactivation enabled by pumitamig with the targeted cytotoxic payload of Elfe-D. We execute these earlier combination studies to provide the signal-seeking evidence we need to inform and potentially derisk our next steps.

This data generation will guide the entry of our novel, novel combination strategy into the pivotal stage, and it is the foundation of the next chapter towards BioNTech's growing leadership in oncology.

With that, I will now turn the presentation over to our CFO, Ramon Zapata, for the financial update.

Thank you, Ozlem, and a warm welcome to everyone joining us. I will cover 3 topics today. Firstly, our second quarter and first half 2026 financials. Secondly, our full year 2026 financial guidance; and lastly, the execution of our share repurchase program announced in May this year as part of our capital allocation strategy.

Note that all figures will be in euros unless otherwise stated. Starting with the second quarter financial performance. Revenues for the second quarter of 2026 were EUR 106 million compared to EUR 261 million in the prior year quarter. This decline mainly reflects lower demand for our COVID-19 vaccine in the U.S. In addition, the prior year quarter was positively impacted by a onetime revenue effect. This related to a compensation payment from Pfizer opting out from our shingles vaccine development program.

Moving to R&D. Adjusted R&D expenses decreased to EUR 477 million from EUR 509 million in the prior year quarter. This change mainly reflects the execution of our disciplined prioritization across the portfolio. Lower spending on our focused programs, together with favorable cost-sharing effects from our collaboration partners supported an efficient cost structure. At the same time, we continue to invest in our prioritized immuno-oncology and ADC programs, including pumitamig and gotistobart.

Moving to SG&A. SG&A expenses on an adjusted and IFRS basis were EUR 198 million compared to EUR 137 million in the prior year. This increase was mainly driven by a global initiative on scaling our processes and ERP infrastructure to strengthen efficient operational execution and our ongoing prelaunch activities for late-stage programs. Our cost base in 2026 also reflects the inclusion of CureVAC operations post merger. At the same time, we continue to realize meaningful savings through pipeline prioritization measures and enhanced cost discipline across the organization.

When comparing IFRS and adjusted results overall, the key adjustments are as follows: within R&D expenses, the difference is driven by impairment charges related to intangible assets outside our focused products. Within other operating results, the difference relates to actions we are taking following our manufacturing footprint consolidation, the decision we announced in May. The costs are mainly employee-related expenses and impairment charges. These charges reflect our progression from announcing the acquisition to actively executing [indiscernible]

While these costs weigh on our near-term results, they are a deliberate investment in reshaping our future prospects. We are positioning the company for enhanced operational efficiency and expect sustainable savings going forward. Importantly, we are acting from a position of strength, allowing us to make these adaptations proactively since we maintain a strong financial position of EUR 16.6 billion in cash, cash equivalents and security investments at the end of the second quarter compared to EUR 16 billion as of June 30, 2025. This empowers sustained investments across the pipeline, our preparations for commercialization and in our long-term goal to become a global multiproduct biopharmaceutical company.

Before we go into our full year guidance, let's first start from the quarterly view to our year-to-date financials, comparing the performance of the first half of 2026 with the prior year period. The drivers are mostly in line with the factors I have described for the quarter. In the first half of 2026, revenues were EUR 224 million. While we expect the seasonal phasing of the COVID-19 vaccine business throughout the year, as mentioned, this decline mainly reflects lower demand in the U.S.

Secondly, lower adjusted R&D expenses of EUR 1,004 million in comparison to the prior year period reflect our focused R&D investments approach in our prioritized programs and positive cost sharing effects with our collaboration partners. Thirdly, higher adjusted SG&A expenses of EUR 349 million reflect the ongoing prelaunch activities and commercial buildup for our first oncology launches as well as costs newly incorporated in 2026 like our ERP infrastructure initiative that I already mentioned.

As we look to the second half of the year, we are taking a disciplined view on our full year outlook. We anticipate changes in some of the factors driving our business and are revising our previously disclosed full year 2026 financial guidance. We now expect revenues in the range of EUR 1.6 billion to EUR 1.9 billion, adjusted R&D expenses in the range of EUR 2 billion to EUR 2.3 billion and adjusted SG&A expenses remaining unchanged in the range of EUR 700 million to EUR 800 million. I will now detail the factors driving our revised revenue guidance. While COVID-19 has been endemic for some time, we continue to monitor the evolving vaccine market and expect softer-than-anticipated global COVID-19 vaccine.

In addition, the European Medicines Agency recommendation issued in May allows the use of the previous year's vaccine formula as an alternative to the newly recommended XFG variant adapted vaccines. This year, for the first time, Germany will utilize previously manufactured on-stock vaccine doses for the upcoming vaccination season. As a result, we expect significantly reduced sales in Germany recently.

Also, the timing of milestone-related revenues resulting from an out-licensed R&D program, which are no longer expected in 2026. In terms of revenue phasing through the rest of the year, we continue to expect the majority of our 2026 revenues to be realized in the second half of the year, specifically in the third quarter when we expect to recognize the EUR 613 million BMS collaboration payments.

Moving to operating expenses. We have revised our adjusted R&D expenses to the range of EUR 2 billion to EUR 2.3 billion. This reflects our focus on optimizing our R&D resources and continued cost discipline as we prioritize the development of our late-stage clinical pipeline. We expect these cost savings based on prioritization and optimization to continue into future years.

Our assumption for adjusted SG&A expenses remain unchanged in the range of EUR 700 million to EUR 800 million as we continue to gradually build out our commercial capabilities. Despite these changes, our strong balance sheet and disciplined cost management position us well to continue investing in strategic [indiscernible]

Turning to my final slide and giving you an update on the execution of our capital allocation framework, which we presented during our first quarter earnings call. Our approach remains clear and disciplined centered on 3 priorities: first, focused R&D investments; second, disciplined capital deployment as well as third, optimize operational efficiency and sustainable value creation.

We are executing against these priorities with consistency and intent. As shown on the slide, with respect to the second pillar, we started the execution on our up to EUR 1 billion share repurchase program and repurchased an amount of EUR 162 million so far. This execution reflects our conviction in the intrinsic value of BioNTech.ile importantly, we retain full optionality to advance our pipeline, execute on partnerships and pursue corporate development opportunities.

Taken together, our strong financial position and the 3 pillars of our capital allocation strategy continue to serve as a clearly defined long-term objective to become a global multiproduct company, addressing the high unmet medical needs of cancer patients worldwide. On a final note, regarding the announcement of new company and potential contributions by BioNTech, the discussions are ongoing. And as with all potential deals, BioNTech's guiding principle in the negotiations is to maximize value for patients and our shareholders. With that, I will hand back to the operator to open the call for questions.

[Operator Instructions] We will now take our first question from the line of Cory Kasimov from Evercore ISI.

I wanted to ask on Elfe-D, the B7-H3 ADC candidate. And what data that you've seen that prompted the decision to choose metastatic CRPC as the first indication for Phase III development? And kind of what gives you the confidence that you have a competitive ADC construct here?

Thank you, Cory. Ozlem, will you please answer the question from Cory?

Cory, I can take the question. So what is very encouraging for our B7-H3 is the combination of durable disease control plus the so far excellent safety profile that we are seeing for the B7-H3. To remind you, various ADCs come either with challenges related to hematosuppression, stomatitis or ILDs. And with our B7-H3 ADC, we see a very tolerable safety profile, allowing us not only to get disease control -- temporary disease control with this compound, but enable long-term application. We have a range of patients who have been dosed now for more than a year without significant significant ILD events [indiscernible].

And.

If I may add to that, Cory, we have presented this data at ASCO 25 and ESCOGU26, also follow-up data from our heavily pretreated population, which we have tested in the Phase I/II study, which is still ongoing.

We will now take our next question from the line of Tazeen Ahmad from Bank of America.

Can you just give us a sense about the data updates that are expected for the remainder of the year? Can you just remind us how your updated guidance for what data to expect has changed since earlier in the year? Specifically, are we still expecting -- to [ China TNBC] Phase III data for triple-negative breast cancer this year? And then should there be any expectation that this data would be presented in a press release versus just released at a medical meeting?

So with regard to changed milestones and specifically also the China interim analysis test, even for first-line TNBC, we are continuously monitoring events and the data readouts are event driven. And in our China TNBC study, we observed that the events take longer. So the readout is pushed out to 2027 which is in principle is also actually a good sign for us.

Then another study, which will come a bit later with regard to its readout is our pumitamig study in first-line CRCC in China, which is a Phase II study. Here, we have decided to look in more chemo combinations than originally planned. So they are added on top of this. And another readout, which will -- is now projected for 2027 is the gotistobart Phase II in second-line positive prostate cancer. And the reason is that we want to see more maturity.

Major milestones and readouts this year are gotistobart interim analysis for Part 2, meaning the pivotal part of our non-small cell lung cancer study in squamous non-small cell lung cancer, second line in combination with docetaxel. So we are excited about that.

AS a control,

Yes, as a control. We are excited about that. Then the BNT113 trial where we expect the Phase III interim analysis for progression-free survival later this year and the [indiscernible] study in HER2 low hormone receptor positive metastatic breast cancer.

We will now take the next question from the line of David Dai from UBS.

Just on pom plus chemo in frontline non-small cell lung cancer, we've seen some encouraging Phase II data so far, especially showing translation from China to global. I'm just curious what additional preclinical evidence or, let's say, efficacy, safety or biomarker features that give you confidence that [indiscernible] can be a differentiated VEGF PD-1 versus other VEGPD-1 approaches?

Maybe I take the question. So with regard to the bispecific class, they have one thing in common that they both due to the bispecific activity enable either improved binding to PD-1 or PD-L1. We have one feature, which we believe is differentiating with our molecule binding to PD-L1 in the tumor microenvironment. This gives us in principle the opportunity to have a double tumor microenvironment directed compound. Whether this translates at the end of the day to a differentiated efficacy, we have to see. There are, of course, no head-to-head price here. We believe that the true differentiation will come with the overall portfolio. in which we combine not only with chemotherapy, but with a differentiated set of ADCs and other compounds.

We will now take our next question from the line of Daina Graybosch from Leerink Partners.

I wonder if you could help us understand your T-Pam comments more. Give us the details, is there a specific outcome or threshold in the HER2 low breast cancer you're looking to exceed and how various outcomes from that study could impact your strategy forward with regulators and launching -- potentially launching T-Pam.

So our T-Pam program, as you know, is broader. We are developing T-Pam in endometrial cancer, second line and in breast cancer with our Phase III trial in hormone receptor positive HER2 low. And we have what is also important to note ongoing signal-seeking studies of T-Pam with pumitamig in different breast cancer patient segments, which is an important part of the strategy because ultimately, our ADCs are part of our portfolio because of the potential to further elevate pumita and allow us to leapfrog. With regard to our breast cancer study, the benchmarks if you compare with published benchmarks, median PFS in the range of 9 to 13 months and 18-month median OS of around 85% according to other studies in this indication and approved treatment.

We will now take the next question from the line of Geoff Meacham from Citigroup.

Just had a bigger picture one on capital deployment. You guys have a substantial cash position. But you're also streamlining the pipeline with the cost savings initiatives. So on the latter, I guess, can you talk a little bit about what your sort of North Star is in this? Is it deprioritizing overlapping indications? Are you eliminating some earlier stuff based on competitive landscape? I just want to get a sense for the strategy there.

Thank you for question. so I think Because of the cash position that we have and the strength of our balance sheet, we are able to take on the number of pivotal and Phase III trials that we are running now with pumitamig and the strategy with our partners is to add on these efforts as much as possible to really widen the net of all the indications where we can use Pometamig and no novel combinations.

So I would say that our capital allocation priorities remain unchanged. We continue to fully fund our priority pipeline and the commercial capabilities needed to support these upcoming launches. Second, we maintain the flexibility to pursue attractive external opportunities that have the potential to strengthen our portfolio or our capabilities.

And third, we continue to return capital to shareholders through the authorized share buyback program that we announced last quarter. So on your comment on portfolio optimization, we continuously review our pipeline to ensure that the resources are focused on the areas with the greatest strategic and value creation potential. And that means continuing to invest behind our core programs while reducing or stopping investments in noncore assets where we feel it's appropriate.

And maybe I can add here another aspect. So this year is the year of combination trials where we evaluate pumitamig in combination with our larger ADC portfolio. The results of the studies, of course, will provide further prioritization of the best combinations, thereby reducing maybe the investment in some of the ADCs in certain indications. So the overlap at the moment is by purpose to identify the winners. In 2027, we expect that we will have identified the winners and engage into several Phase III clinical trials, including assets from our partner, BMS.

We will now take the next question from the line of Akash Tiwari from Jefferies.

So you have 3 interim Phase II readouts expected in the second half of this year. You have your HER2, your CTLA-4 and then your head and neck cancer vaccine. Can you talk about your confidence on a positive interim analysis for each of these programs? Is there a particular program where maybe the team's internal view is particularly bullish?

It's a difficult question. So we have to see the data. And -- but we are positive. We have, of course, positive expectations for each of the trials. You know that if the data for gotistobart that we have seen in the first part of the Phase III clinical trial is recapitulated, this would become a game-changing result in this indication. Everyone knows that docetaxel remained unbeaten for decades now. And this would be the first time that if the data are recapitulated, we would have a significant benefit with a mono compound as compared to the standard of care.

We will now take the next question from the line of Terence Flynn from Morgan Stanley.

I appreciate the update on your iNeST CRC data coming next year. But I was wondering if you could help us think about potential read-through from the Moderna Merck INT adjuvant melanoma Phase III data that we might get this year. So what would you be looking for in that data to give you confidence in your own iNeST program?

So in terms of biology and indication, there we don't see any read-through opportunities. Melanoma versus colorectal cancer, these are very different biologies and indications and responsiveness to immunotherapy and in particular, antigen-specific T cell antigens. We remain committed to the way we are conducting together with our partner, Genentech, our program focusing on adjuvant settings, focusing also on cancers where checkpoint inhibition immunotherapy has a lower probability of success and does not serve the medical need. And neoantigen vaccines are not created equal. So it's difficult to read from one platform to the other...

We will now take the next question from the line of Jessica Fye from JPMorgan.

Ramon, I was hoping you could help us out with the EUR 400 million reduction to guidance at the midpoint. Can you just quantify how much of the change was driven by the milestone pushout versus the [indiscernible] decision to use existing inventory? And how much is just softer COVID demand? And then what specifically was the partner milestone that was pushed out?

Thank you for the question. Most of the adjustment of the guidance is a reflection of the weaker COVID-19 vaccination rates and the current regulatory and public health environment. It's really incorporating the latest inputs from our partners and the teams that are operating in our key markets. I would like to remind you that the COVID revenue is back-end weighted into the late Q3 and Q4. So this will only reflect on the event. And also that we are expecting approximately EUR 613 million revenues from the BMS collaboration that is also giving us a good uplift for the second half of the year.

So based on the information currently available, we believe that this revised range is appropriate. Now in relation to the comment of the license milestone that we were expecting, yes, we were expecting that as well as the second half of the year, but it's not the key driver of the revenue adjustment. So it's mainly lower demand across every market. And then specifically for Germany, it impact us a little bit different because Germany is a direct market versus the other countries that are Pfizer managed markets. So then we get the impact on our revenues a little bit more acute than versus the other markets. So I would say, if you would think about percentage, 80% is COVID-related completely and then the rest is the loss of the out-license milestones that we were expecting.

We will now take the next question from the line of Yaron Werber from TD Cowen.

I wanted to ask about ROSETTA Lung-02.

Clinicaltrials.gov is showing data in 2029, but you changed the endpoint now to PFS and not OS. Is there any chance that we can get this data potentially earlier? And any sense when you might finish enrollment?

We can't, at the moment, not speculate on when to expect the data. As compared to waiting for OS, it will be obviously earlier that PFS reads out, but we don't have any guidance for the final readout yet.

And maybe can you just remind us when you changed the endpoint, did the interim analysis change in any way? Anything you can share would be great.

The changed endpoint and the interim analysis for PFS will become now an opportunity to file if it's positive. We can't still -- as said, it is expected to read out earlier than the interim OS. But we can't, at the moment, say when we are going to expect because this is, again, an event-driven trial and it's the first time in this indication that we evaluate this large population.

We will now take the next question from the line of Evan Seigerman from BMO Capital Markets.

This is [indiscernible] on for Evan.

Just one from us. So as we think about the broader opportunity for Pumitamig, can you discuss how the asset might be differentiated in MSS CRC given the limited efficacy of checkpoint inhibitors in the indication and immune collection type of tumors? And then also, what gives you confidence that Pumi's mechanism could overcome these past challenges?

Yes. to provide -- because we do not yet have data in CRC, the strongest evidence that Pumitamig3 or the bispecific class is differentiated comes from the observations of objective response rate, but also durable disease control in patient populations who are PD-L1 negative. For example, in TNBC, we have documented data showing more or less the same rates of objective response in the patients in PD-L1 positive, PD-L1 low positive and PD-L1 high positive patient population.

So CRC is an indication in which checkpoint blockade was not successful. We have now this combination, and we have to see from our interim analysis, which is coming in 2027, we are going to test or we are testing at the moment different chemotherapy combinations, whether this translates to better data as compared to traditional benchmarks with chemotherapy alone.

And so -- but the broader pumitamig opportunity, again, is based on the one side, improving response rate and durable disease control and OS, in indications where checkpoint blockade PD-1 is approved, opening up indications in which PD-1 treatments are not approved. And as a third component, combining pumitamig as a potential next standard of care with a new generation of ADCs that combine -- that allow disease control even in advanced disease with a good safety profile.

We will now take the next question from the line of Mohit Bansal from Wells Fargo.

I have a question regarding squam versus non-squam. So there will be a lot of data coming this year from competitor as well. So the first trial that is reading out is in for squam lung cancer for VEGF PD-1. My question is how much read-through there could be for the non-squam program if squam were to be successful? And what specifically you would be looking at the competitor data to gain confidence in your own programs or think about the future trials?

With regard to the read-through, in principle, these histologies are different diseases, right? So we would be very cautious to read from data in squamous to non-squamous or vice versa. So we really need to produce the clinical data for both histologies and in our ROSETTA Lung-02 trial, for example, we have, therefore, also separated both histologies in subtrials.

But on the other side, the data, our own data, but also the data coming from [ OS ] indicate that in both indications, PFS is improved. And we have seen now in the recent update that the improved PFS appears also to translate into OS thickness in other indications. So we are cautiously optimistic that we will see in both indications, PFS benefit and OS benefit.

Thank you. This was our final question. This concludes today's conference call. Thank you for participating. You may now disconnect.

Welcome to BioNTech's First Quarter 2026 Earnings Call. I will now hand the call over to Doug Maffei, Vice President, Strategy and Investor Relations. Please go ahead.

Thank you, operator. Good morning and good afternoon. Thank you for joining BioNTech's First Quarter 2026 Earnings Call. As a reminder, the slides we will use during this call and the corresponding press release can be found in the Investors section of our website.

On the next slide, you will see our forward-looking statements disclaimer. Additional information about these statements and other risks are described in our filings with the U.S. Securities and Exchange Commission, or SEC. Forward-looking statements on this call are subject to significant risks and uncertainties and speak only as of the date of this conference call. We undertake no obligation to update or revise any of these statements.

On Slide 3, you can find the agenda for today's call. I'm joined by the following members of BioNTech's management team: Ugur Sahin, Chief Executive Officer and Co-Founder; Ozlem Tureci, Chief Medical Officer and Co-Founder; and Ramon Zapata, Chief Financial Officer. Also available for the Q&A portion of today's call is Annemarie Hanekamp, our Chief Commercial Officer.

With this, I'll hand the call over to Ugur.

Thank you, Doug, and everyone, welcome to everyone joining us today. As BioNTech has grown, our vision has remained confident translating signs into survivor. Cancer is a complex systems disease with heterogeneity across patients and variability within individual tumor. The future of cancer treatment will therefore center around rationally designed therapeutic combination, telling potent and precise mechanisms of action that create biological synergies.

To address this, BioNTech has built a diversified toolkit of modalities comprising immunomodulators, ADCs and mRNA cancer immunotherapy. We believe that combination approaches will be key to elevate patient outcomes meaningfully across solid tumors. To execute against that vision in 2026, we have 3 priorities: first, accelerate the late-stage development of our oncology assets. Key late-stage data readouts are anticipated from our first day of oncology programs this year.

Second, big momentum in combination therapy. In 2026, we are expanding our novel-novel combination strategy centered around Pumitamig as potential next-generation IO backbone. This includes the first expected readouts of combination tiles with our ADCs as well as with other next-generation therapies including the recently announced partnership with [indiscernible] to combine Pumitamig with quick standing.

Third, shift from a platform-centric to tumor center clinical development approach around high incident cancers such as lung cancer, breast cancer and other tumor types. The foundation of this matrix approach is our combination strategy, which allows us to address several lines of treatment with different asset combinations. In March, we announced plans to pursue next-generation mRNA innovations in a new independent company, founded and led by SM and I, with BioNTech and our new company focusing on their respective strategic priorities, we aim to maximize value for patients and shareholders. Selling at [indiscernible] is ongoing, and we expect to share details of an agreement later this year.

BioNTech is well positioned for this next phase. With a growing late-stage pipeline, strong partnerships and financial strength, we are on track to become a diversified multiproduct oncology company by 2030. We are targeting more than 17 late-stage and pivotal [indiscernible] to 2030, spanning multiple tumor types and different lines of treatment. We end up the remainder of this year's momentum solid execution and which set of catalyst opportunities ahead.

With this, I will hand over to Ozlem for an update on our oncology execution.

Thank you, Ugur, I'm glad to be speaking with everyone today. BioNTech's clinical development strategy seeks to address the full continuum of cancer. From resected high-risk tumors in the early setting to advanced and metastatic disease as well as treatment resistant and refractory cancers. We have defined a set of key tumor focus areas with high incidence and high unmet need, including lung cancer, breast cancer and others.

Across these 2 more focus areas, our goal is to leverage novel combinations to maximize the potential of our pipeline and to elevate solid tumor treatment outcomes. As such, we are advancing multiple assets from our multimodal oncology pipeline into late-stage development. During the first quarter of 2026, we continue to make progress here, and I look forward to speaking to some of these updates today. I'll begin with lung cancer, which is our most advanced example of our disease area focused approach.

Our lung cancer strategy is built as a matrix across firstly, disease stages and settings from resectable tumors to unresectable Stage III disease for metastatic first-line and later lines of therapy. Second, clinical and molecular subgroups including patients with and without actionable alterations and with different PD-L1 expression levels. And third, treatment backbones and combinations with pumitemic at the core. This quarter, we continued to add to the body of evidence for our lung cancer approach, including the presentation of new data at the European noncancer Congress.

Starting with Bumitamic, our PD-L1bGFA bispecific antibody and the IO backbone of our combination-based development strategy. In March, we presented Phase Ib/IIa data at the LCC. This trial evaluated Pometamic as a monotherapy in patients with previously untreated advanced non-small cell lung cancer, enrolling both squamous and non-squamous histologies. The results are encouraging in an overall patient population with PD-L1 expression of at least 1%. We observed a confirmed objective response rate of 46%, a median progression-free survival of 13.6 months and the median overall survival of 27 months. The disease control rate was 96%. Two features of these data deserve particular emphasis. First, the activity observed across PD-L1 subgroups is noteworthy and second, the particularly strong response rate of 71% in PD-L1 high squamous disease. The tolerability profile was manageable with a low rate of treatment discontinuation. These data support the ongoing global Phase III program for Pumitamig in lung cancer. The RecetaLANg-02 trial is currently recruiting in its Phase III portion comparing to metamec chemotherapy to pembrolizumab plus chemotherapy in first-line non-small cell lung cancer. Phase II data from this trial are expected to be presented at ASCO 2026.

As Ugur mentioned in his opening remarks, another component of our lung strategy is our recently announced collaboration with [ Bringelly ]. The study combines DLL3 targeting T cell engager or [indiscernible] with Pumitamig. The clinical trial aims to develop a novel treatment regimen that delivers more sustained tumor control in extensive stage small cell lung cancer, one of the most aggressive and underserved forms of cancer.

Small cell lung cancer progresses rapidly metastasizes early and almost always recurs within a year after initial treatment. While the addition of immune checkpoint inhibitors to chemotherapy has led to improved survival outcomes for patients with extensive stage disease. Most patients progress within months after treatment and the prognosis remains poor. The collaboration combines 2 complementary immunoerapeutic mechanisms to explore a potential new path to enhance and sustain antitumor immunity.

[indiscernible] T cells to cure DLL3 expressing tumor cells while pumitamic aims to restore TC's ability to recognize and destroy tumor cells while cutting off the blood and oxygen supply that feeds the tumor with the intention of preventing it from growing and proliferating. As you can see on our lung cancer slide, we are deploying multiple modalities, next-generation immune modulators, ADCs and mRNA immunotherapy.

Gotistobart is a critical component of that picture. Gotistobart is our selective Treg modulator targeting CTLA-4 developed in collaboration with our partner, OncoSpand it is designed to precisely address the patient population that sits beyond Pumitamig's current focus, namely patients with metastatic squamous non-small cell lung cancer whose disease has progressed following platinum-based chemotherapy and PD-1 PD-L1 inhibitor treatment. This is a setting with very few effective options and poor prognosis.

Gotistobart's differentiated mechanism, selective regulatory T cell depletion in the tumor microenvironment is designed to reengage the immune system even after prior checkpoint inhibitor exposure. In January, Gotistobart received orphan drug designation from the FDA for squamous non-small cell lung cancer, building on its existing fast track designation.

In March at the ECC, we presented updated data from the nonpivotal dose confirmation stage of Preserve-003, our global Phase III trial. The data are very encouraging. The 12-month PFS rate of 25% for Gotistobart versus zero for docetaxel is a signal of durable disease control. Gotistobart reduced the risk of death in this IO pretreated patient population by 54% compared to docetaxel with a hazard ratio of 0.46. The median OS in the Gotistobart arm has not yet been reached compared to approximately 10 months with docetaxel. At 12 months, 63% of patients treated with Gotistobart were alive versus 30% in the docetaxel arm. The safety profile was consistent with the previously established profile for Gotistobart with no new signals of concern. These encouraging data are derived from a small patient population and require further validation.

Based on current event accrual projections, we expect interim data from the pivotal stage of Preserve-003 later this year. This program reinforces the breadth and depth of what we are building in Lung cancer.

I'll now turn to gynecologic cancers, another of our tumor focus areas and 1 where we have a late-stage asset, trastuzumab Permian Tian, our HER2-targeted ADC developed in collaboration with our partner, [indiscernible] Updated TPAM data were presented at the Society of Gynecological Oncology Annual Meeting in April, and patients refer to expressing previously treated advanced or metastatic endometrial cancer, including patients who had received prior immunotherapy, TPAM demonstrated a confirmed objective response rate of 49%, with a median duration of response of 9.9 months and a disease control rate of 79%.

Responses were observed across all HER2 expression levels, IHC1+, 2 plus and also [indiscernible]. The safety profile was manageable and consistent with what has been previously reported for ADCs and HER2 targeted agents in this setting. The confirmatory FRNC-01 Phase III trial continues to enroll. In addition to our studies of PAM in endometrial cancer, the ADC is also being evaluated in the Phase III clinical trial in HR-positive, HER2 low metastatic breast cancer the Dynasty BREAST 02 trial. The Phase III interim analysis for this trial is expected later this year based on current event accrual projections.

Moving now to our portfolio of innovative mRNA cancer immunotherapies, which aim to activate and educate the immune system with precision. Our personalized approach includes autogenevumiran, which is partnered with Roche Genentech. In 2025 and early this year, we published data from multiple trials that support our focus on the adjuvant setting where tumor burden and heterogeneity is lowest. The biology and our clinical experience point to greatest relevance in earlier disease settings where lower tumor burden allows for immune system to consolidate control. Updated long-term follow-up data from the PDAC Phase I trial were presented at the AACR Annual Meeting this year among the 8 patients who mounted an immune response to the immunotherapy, 7 remained live for up to 6 years after surgery and demonstrated persistent cytotoxic cancer killing lymphocytes.

In contrast of the 8 patients who did not exhibit an immune response, only 2 were still live was a median overall survival of 3.4 years. In adjuvant, CTDNA positive, Stage 2 high risk or Stage 3 colorectal cancer. We have a Phase II trial evaluating autogenerumiran monotherapy against watchful waiting. The final analysis with disease-free survival as primary endpoint is event-driven and according to projections to be expected in 2027. For fixed rack in first-line HPV16-positive PD-L1 high HNSCC. We have a Phase II/III trial in combination with pembrolizumab. Recruitment is ongoing in the Phase III interim analysis is expected in 2026. In Q1, we generated additional data and evidence to support lung and gynecological cancer 2 of our tumor disease focus areas, in particular. Looking ahead for catalyst calendar for the remainder of the year remains rich. In our late-stage programs, we anticipate 5 more readouts in parallel, early data from our pometamicplusADC combination trials will begin to inform the design of our first pivotal combination trials, a milestone that marks the next chapter of our novel strategy.

We are in the midst of a sustained evidence led data generation phase [indiscernible] is designed to advance our pipeline, derisk our programs and bring us closer to our goal of delivering meaningful new treatment options for patients with cancer.

With that, I will now turn the presentation over to our CFO, Ramon Zapata, for the financial update.

Thank you, Ozlem, and a warm welcome to everyone joining us. Today, I will cover 3 topics. To begin, our first quarter 2026 financial results. Second, our reaffirmed full year 2026 financial guidance; and third, an update on our capital allocation strategy, where I will speak to our planned share buyback program and our manufacturing footprint consolidation initiative.

Note that all figures will be in euros unless otherwise stated. Our first quarter performance is in line with our expectations and reflects the seasonal demand pattern we expect across quarters for COVID-19 vaccines. Revenues for the first quarter of 2026 were $118 million. This compares to $183 million in the same period last year. The decrease was primarily driven by lower demand from our COVID-19 but since as expected.

R&D expenses were $557 million compared to $526 million in the prior year period. The increase was driven by higher spending on our immuno-oncology and ADC programs in particular, to Pumitamig and Gotistobart as well as R&D costs from BioNTech China, previously named by [indiscernible] and [indiscernible] which were acquired in 2025. These increases were partly offset by lower expenses from our COVID-19 vaccine collaboration with Pfizer.

On an adjusted basis, R&D expenses were $527 million, excluding an impairment charge for an intangible asset. SG&A expenses were $151 million compared to $121 million in the prior year period. The increase was mainly driven by our ongoing commercial buildup and the post-acquisition inclusion of operations from BioNTech China and [indiscernible] Adjusted SG&A expenses were identical to the results under IFRS accounting standards. We ended the first quarter with $16.8 billion in cash. Cash equivalents and security investments.

Our strong financial position continues to support sustained investment across our pipeline, late-stage oncology programs and our preparations for commercialization. Turning to the next slide. We are reaffirming our previously disclosed full year 2026 financial guidance. All guidance is provided on an adjusted non-IFRS basis. We expect total revenues for 2026 in the range of $2 billion to $2.3 billion. As stated at the beginning of the year, we anticipate lower COVID-19 vaccine revenues compared to 2025, driven by declines in both the United States and European markets. The U.S. market continues to be competitive and dynamic.

In Europe, we expect lower revenues as we defend our market share and begin managing the transition away from multiyear contracts. In Germany, specifically, we recognize direct sales of our COVID-19 vaccines as revenue. Hence, the anticipated declines in our sales of COVID-19 vaccines in the country will have a direct impact to our top line, whereas revenues outside of Germany only affect our top line as part of the 50% gross profit split with our partner, Pfizer.

Revenues from our collaboration with EMS from the Badami preparedness contract with the German government and from our services businesses are expected to remain stable. On revenue cadence, we anticipate COVID-19 vaccine revenue pacing to be similar to last year's, with the last 4 months of the year, driving the majority of the full year revenue figure. The BMS collaboration payment of $613 million is expected to be recognized in the third quarter of 2026.

We expect adjusted R&D expenses in the range of $2.2 billion to $2.5 billion. Investments will be concentrated on our priority late-stage programs. We will continue applying disciplined portfolio prioritization across all development stages. We expect adjusted SG&A expenses in the range of $700 million to $800 million reflecting our continued commercial build-out in oncology.

Turning to capital allocation. Let me highlight 3 key components of our approach to create long-term shareholder value. The first component is focused R&D investments to maximize the potential of our pipeline. We actively manage our portfolio, focusing our resources and programs that have the greatest potential to elevate patient outcomes. This means increasing investment into our late-stage priority products, namely Pumitamig, our ADC assets mRNA immunotherapies and the respective combinations while reducing spend outside of those areas. The second component see us mobilizing our strong balance sheet as a statement of confidence in the business.

We plan to initiate a share repurchase program of American depositary shares of up to USD 1 billion over the coming 12 months. Let me walk you through some principles that guided this decision. One is opportunistic flexibility. This program gives us the ability to deploy capital during times when our share price may be disconnected from intrinsic company value. Another principle is that our pipeline remains the primary driver of value. The buyback is supportive of the share price. But it is not determinative. The real value creation story at BioNTech remains the clinical execution of our oncology pipeline. Also, disciplined capital management. This program implements our R&D investment. We retain full optionality to advance our pipeline, execute partnerships and corporate development opportunities.

Our balance sheet with $16.8 billion in cash, cash equivalents and security investments gives capacity to do all of this simultaneously. In short, the share repurchase program reflects confidence in our science. Capital management discipline and a commitment to delivering long-term value for our shareholders. The third key component of our capital allocation strategy relates to the optimization of operational efficiency and commitment to sustainable value creation at sites. This will affect just over 1,800 positions. For each of these manufacturing sites, we are exploring divestment options through the end of 2026. This includes a partial or total sale. We expect cost savings to ramp up over time.

Once the measures are fully implemented, we expect approximately $500 million in recurring annual savings in alignment with our capital allocation approach, these savings are intended to further support the advancement of our oncology pipeline towards commercialization. This is a decision we have taken after careful assessment. Our commercial and R&D drug supply will be covered by our broader manufacturing network. Supply of our COVID-19 vaccine will be fully handled by our partner Pfizer via their established manufacturing capacity beginning at the end of 2026. These plans underline our commitment to continuously steel our capacities in support of our strategy to become a multiproduct company by 2030. As we look across these 3 icons on the slide, we are energized by the progress we have made to date and the path ahead. We are making progress towards our strategy -- we are progressing key programs into pivotal stage, leveraging our partnership with BMS and our fortified balance sheet to fund our pipeline.

From 2026 through 2029, we will drive execution at scale and speed, advancing combination therapy studies, accelerating pivotal trial execution, building tumor indication specific portfolios and executing our first oncology launches. By 2030, our goal is to be a diversified multiproduct global biopharmaceutical company, addressing the high unmet medical needs of cancer patients worldwide. BioNTech's robust financial position empowers us to pursue that goal. We remain fully committed to translating our science into survival for patients.

With that, I will hand back to the operator to open the call for questions. Thank you.

[Operator Instructions]

Our first question comes from the line of Dain a Graybosch from Leerink Partners.

We're excited to see the initial data from Rosetta Long 02 at ASCO. And I wonder although I have a question more about the statistical design of that study. We've all noticed and I think you shared in the last earnings call that you changed the primary end point from a dual PFS OS to a single PFS primary. And I wonder if you could talk more about why you made that change, including any conversations you had with BMS and with FDA.

First question from Daina about [indiscernible] which is coming at ASCO and a question about the rationale behind the endpoint change, which we announced, I believe, about 2 months ago.

Yes, I can take that, Doug. Daina, thank you for the question. We have made this change because PFS is a well-accepted endpoint in non-small cell lung cancer and is very large, and we expect the largest and earliest benefit a signal in this endpoint and wanted to make sure that we allocate the full alpha on this endpoint and have a statistically robust read out.

This does not mean that we need let overall neglect overall survival. Overall survival is in fact, a key secondary endpoint and will also be assessed. And as you know, this is a well-trodden regulatory path in particular, for non-small cell lung cancer, which has also been extensively used by Kate Roder.

We'll now move on to our next question. Our next question comes from the line of Jessica Fye from JPMorgan.

Thinking ahead to [indiscernible] the HR-positive HER2 trial for TPAM, on what metric or endpoint do you expect the data to best underscore differentiation from HER2.

Okay. Question from Jeff at JPM on essentially how we see differentiation of TPAM versus in HER2.

So you asked for the endpoint metrics. The primary endpoint is objective response rate in connection with duration of response. And we have provided the data from the largest recurrent endometrial cancer population at SGO, which you might have seen where we demonstrate the objective response rate and duration of response together with a manageable safety profile. The differentiation is that we have now a data set, which shows that our ADC has also a clinically meaningful benefit in the lower HER2 population and the 1-plus and 2-plus population, which is a differentiator.

I was asking for Dynasty Breast 02, the HER2-low trial, where we have benchmark data from in HER2?

So this is -- yes, just to uses a pile of Tamesimotherapy. There's not a direct comparison of HER2. Of course, there are data where you can benchmark the results of this part which no. We have to see the readout and ensure, first of all, that there's a positive study and then whether we can make a cost comparisons to other [indiscernible]

And I'll add this is [indiscernible] Chief Commercial Officer for Biotech. We've always signaled that TPAM is an important asset for BioNTech also predominantly as a strategic asset not just for building out our commercial engine, which will be the first time for buying tech in the oncology space, but also as a combination partner.

So towers point, we will wait for the data readout the physicians we spoke to always signal that they like to have more than one option. So we do see a meaningful place for [indiscernible] in the West center space as well. But again, a strategic asset that we predominantly also focus on in combination.

We'll now move on to our next question. And our next question comes from the line of Tazeen Ahmad from Bank of America.

For the upcoming data that you're expecting to show at ASCO for Punis chemo in the frontline non-small cell setting. How can we best frame expectations? What would be good data there?

Okay. Thank you. So -- that question was on our upcoming data that we are presenting at ASCO PMI frontline non-small cell lung cancer, what are our expectations in terms of that data set.

Ozlem, would you like to take that?

Yes, I can take that. So this -- the data we present at ASCO is stronger Phase II heart of this trial. And what we will show is efficacy profile and the safety profile of 2 different doses of Tomita in the combination with chemo in this patient population. And that data might have to inform about what to expect then from the ongoing pivotal Phase III part of the trial.

Our next question comes from the line of Akash Tewari from Jefferies.

This is Manoj on for Akash. Just 1 from my end. Given the recent disclosures around the PFS interim from the HARMONY 3 global trial, do you think any changes in FX size assumptions or design changes needed to be considered for the Ocotlan trials? -- and also the Optoplan of trial showed interim overall arable hazard ratio around 0.6. So do you still think the chemo combos are the optimal approach -- market entry approach in the city?

Okay. Thank you, Manoj. So I caught that first question is on Harmony if that checks our perspective on the space? And second question, it was a little tricky to hear that all your within about best option for chemo combinations.

Could you just clarify the question?

Yes. So [indiscernible] option from the fact TMDslike hazard ratio of 0.6 overall as a ratio point -- so just wondering like whether chemo combos are still the option or like going for the ADC combos will be the most optimal option to enter the market first?

Okay. Great. So also, should we pass over to you for the HARMONY-03 data? And then Ugur, if you could offer some context on the second question, please.

Yes, sure. So the recent disclosure of Harmony free data is about interim analysis of PFS, which is which was a late edit early look into PFS. But we don't know much about the metrics behind that. So we cannot comment extensively. However, Summit management has signaled that quote, they have deliberately used in minimal alpha to set the bar high, which is a very valid approach at that in this case. However, it also means that statistically this interim analysis is uninformative on the hazard ratio. So we have to wait for the next analysis, which will be later this year.

Yes. And so in short, we are -- this does not change anything for our overall strategy. We would like to remind everyone that our overall strategy has several days of development. The first day of development is Kometani such Simon. But we have already started more than a year ago, with first combinations, ADC combination. And at the moment, we have more than 10 clinical case ongoing to assess the combination of Puritane our ADC to our 4 to 5, 3 to 6. And we were in part on the studies end of -- in the second half of 2026. And this study is, of course, provide our differentiation that actually what comes next as the second phase, which will be a combination of comatoselected ADCs in different type of indications.

We'll now move on to our next question. Our next question comes from the line of David Day from UBS.

I just wanted to come back to the LOI, where you changed the primary endpoint from dual PFS OS to PFS primary endpoint. How do you think this will help with regulatory pathway? Does that mean that you're able to actually get approved just on PFS with accelerated approval and then full approval on the OS, just also thinking litter around how should we think about Redapt using PFS as a primary endpoint.

Okay. Great. Thank you, David. So Alan, maybe if I pass to you. So it was a question -- a follow-on question on Resetting on the endpoint changing from dual to primary on the rationale for that, specifically what it helps us to do with the development.

Yes. So first of all, this position was discussed with our partner, BMS and also with regulators -- and the point is, in fact, that PFS is the earliest potential readout. We know that this type of next-gen -- is that PFS is the earliest and also the largest endpoint to cover the mechanism of action of this next-gen IOs. And with having PFS as only primary endpoint, we can put the entire alpha on this PFS and ensure that it has the highest speed out power -- so this is the rationale behind that.

Again, still overall survival is a key secondary endpoint. And having it as a secondary end point allows us to get a clean path to approval with even a delay or soft OS.

Our next question comes from the line of Asad Haider from Goldman Sachs.

The updates on the trial progress. Maybe just shifting gears quickly for Ramon on capital allocation. just given the substantial cash balance, it would be helpful to hear your updated thoughts on deployment and what the considerations were that went into the $1 billion share program you repurchase program you announced this morning. And then just on the revenue guidance reaffirmation despite the seasonally lower Govin 1Q that you're calling out, just talk us through how you're thinking about the revenue progression through the rest of the year.

Okay. Great. Thank you so -- thank you, Asad. I appreciate the questions. So first, talking about our capital allocation. I think our capital allocation strategy remains the same. We acknowledge that we are in an investment phase that we are building biotech into a commercial stage multi proto-oncogy company by 2030. And the good thing is that the strength of our balance sheet allows us to invest in the pipeline, continue to build our commercial capabilities and preserve flexibility for target opportunities in the M&A or the BD space.

And additionally, now it's also allows us to return capital to shareholders. So the repurchasing program is not at the expense of our innovation efforts on pipeline, but it's more to be seen as an element of our overall capital allocation strategy. And then if I move to the revenue guidance and the dynamics of the COVID vaccine revenues. So I would say that our current guidance already assume assumes lower pro basin revenues versus last year. And as you rightly point out, so the regulatory and the recommendation landscape remains dynamic. And as you can expect, we are monitoring these developments very closely.

Now based on the information available today and including the expected seasonal profile of comminate revenues, we are reaffirming our 2020 revenue guidance.

Our next question comes from the line of Terence Flynn from Morgan Stanley.

Just two for me. I was wondering if there's any update on the CEO search? And if you can provide a time line for when that might be finalized? And then also with respect to your TPAM FDA discussions, similar type question, just any update there and expected time line for visibility.

Thank you for the question. On the succession process. So this is being led by the Supervisory Board, so I cannot comment on specific timing or process details. What I can tell you is that both Ugur and Ozlem, together with the full management board and the overall organization, we remain committed to delivering our 2026 priorities.

Our operating focus and strategy has not changed. And we will over the market as appropriate when we have more information on that.

Okay. Great. Thank you, Ramon. So now on TPAM, maybe if we pass to Osman, first of all, and then Marie, you could add some color as possible.

And this time it's about endometrial that study, right? Sorry for missing that for the other questions. So this cancer Phase II cohort is fully enrolled, and we have presented data, the confirmatory Phase III trial, the FERC continues to enroll, and we are in discussion with the FDA. We haven't changed our plans to submit Yes. And I would add to that what I stated before, TTM continues to be an important asset for us to lay our groundwork for commercial stage biopharmaco company and we continue to see the started launch as a very strategic opportunity to build our commercial infrastructure and prepare for potential future launches where, as you know, especially in the United States, time to peak for oncology assets go around time lines of potentially 9 months.

So we don't have time to learn on the fly sort of saying, especially if we look at the potential for mid where we also partner with Bristol-Myers Squibb on the commercialization. This together would set us up nicely for success even though currently, we're not experienced in oncology launches as of yet.

Our next question comes from the line of Evan Seigerman from BMO Capital Markets.

We're looking forward to the data at ASCO. I want to follow up on Terence's question. As you think about the management change. Can you talk to the profile of a new executive team that you might want to bring in? Is it still R&D focused? Are we going to shift more towards commercial as you transform the company.

Thank you, Evan. Again, sorry if I am not going to be able to give too many specifics and details because the management board is not running this process. It's our Supervisory Board. Now having said that, our Chairman Helene has shared some characteristics last quarter when we disclosed the change in the management board. But I think it's -- so what we are looking is for skills and capabilities in late-stage development as well as commercialization -- production and commercialization and scale of pharmaceutical products. So I think I would be close to what Heather would be commenting on that.

Our next question comes from the line of Cory Kasimov from Evercore.

Do have 1 question. Let me ask if a competitor bisect data shows like an OS , how does that change the bar for percent of Lent, ike what is strong and just a West trend here beenough? Or does that just the antibias second, I'll clearly [indiscernible]

Sorry, the audio was not so clear on that. Would you mind clarifying? Were you talking about Pfizer's data or a different data set?

I would say ask historic data in the PD-1 BGS space here that does show like a clear OS benefit? How does that change or raise the bar for your studies?

Okay. Yes, understood. We get back now. So I'll pass over to [indiscernible]

Can you it's about how many fixed did I a. On is positive is a how this would change our view.

Okay. So we are also excited to see the data. Remind you that this is a China study, which means that the comparator is titles past chemo, not embolus chemo. So it would not have a direct full for our [indiscernible] study.

We'll now move on to our next question. Our next question comes from the line of Mohit Bansal from Wells Fargo.

Given the Harmony 3 versus how many 6, and we don't know the data in alpha spend on FEI, but there has been some -- there are some questions around the translatability of China data to the global data. So I'm not asking you to comment on HR, but would love to understand when you are seeing your own China data versus global data, what gives you confidence that you would be able to replicate what you saw in China into a global trial.

So generally speaking, there are data sets. For example, Punit, small cell lung cancer, Puma TNBC data IVO second-line EGFR-mutated non-small cell lung cancer data, which are reproductions of previous China data on a global level. So we continue to be very positive about the regional report usability of this data.

Having said that, with regard to the molecules to the molecule class, there seems to be reusability of data. However, there could be still setting specific frictions on data reproducibility in populations or indications in which there are major differences between global and regional population. For example, small cell lung cancer or non-small cell lung cancer, we are in China of smoking rates -- smokers rates are different to global. And that means we have to continue to monitor and follow the data and we'll see from the data, which comes out whether such setting specific frictions on reproducibility will show up.

Our next question comes from the line of Yaron Wage from TD Securities.

This is Jen on for Yaron. So to make the tire catalyst rich here with 5 more late-stage pipeline data readouts across Godin, BMC, et cetera. side a company premedicine ASCO, how should we think about the order and timing for the rest of these CLA stage readouts? And then secondly, on Kanuma. Beyond our 3 lead indications, obviously, 1 other Phase III trials during this year. How are you doing Bristol evaluating where Communic has the most potential.

Okay. Thank you for that. So I thought that, that is essentially around timing and cadence of our late-stage data readouts, I would imagine that in the coming year because that's what we've disclosed and then also how discussions are going with BMS in terms of which indications to prioritize. So Ozlem, should I pass it to you for this?

Yes, I can start with the second one from a scientific and clinical and medical point of view. I can say that BMS and we are very aligned in the understanding of the potential of Punita and that it is a very broad opportunity, and we are deciding on the sequence and on the specific indications together based on data and all the other dimensions, which are relevant for making strategic decisions for anetumab portfolio.

With regards to data readouts, we will have a couple of data readouts on Pumitamig over the last -- over the next -- this year, 12 to 18 months. One of these readouts, for example, at ASCO, the Rosetta LAN GO-2 trial. Later this year, hightop readout from Phase I/II studies of combinations with our ADCs with Punita and additional readouts will follow in the next year. Yes. And I would just add on the BMS and medic strategy is that we have a very deep and strong governance ongoing with BMS at a different level.

So from a scientific, from a clinical perspective. And also we're looking, of course, at where can we address unmet medical needs the most. And as you know, the oncology space is in constant evolution providing more options for patients and making sure that by the time our designs or trials read out, we're still relevant in what the current standard of practice clinical practices, and that is something where we can leverage both DMS and biotech capabilities as we're coming together to make those decisions. And sometimes, that also includes changing some of our initial thinking to maximize the opportunity for Pumitamig for both BioNTech and [indiscernible]

Our next question comes from the line of Jeff Meacham from Citigroup.

This is Jae on for Jeff. Maybe just following up on earlier questions on TPAM. Are there any outstanding data maturation requirements for TPAM that could push the time line beyond the current 2026 submission? And then earlier on the comments on TPAM having efficacy in low HER2 as well, is the strategy to pursue a broad pan-HER2 label?

Okay. Great. So we caught whether TPAM has any outstanding data request that could impact regulatory pathway. And then clarification on HER2 low and what our approach might be there. So also would you like to take the data question.

Yes. I can take both and the after we can also get for commercial input here. So no, we don't have outstanding data questions around TAM what we are currently monitoring is the enrollment of confirmatory trial to ensure a harmonized timing of BLA submission and the time lines for data to come out of this confirmatory trial.

With regard to the populations, we are interested in a broad labor. We -- that's our goal, given that we have a large data set for all HER2 IHC levels, including the low expression on -- and I would add from a commercialization perspective, that -- I mentioned this before and talking to our customers or treating physicians that secondary option is always welcome. I think Part from our commercialization strategic launch and making sure that physicians start to get familiar with TFM itself as we're also moving forward with combination strategy it's going to be important for us to understand where we can leverage the strategic launch for TPAM specifically and then move through in commercialization.

Our next question comes from the line of Harry Gillis from Berenberg.

I have a follow-up on Catalyst timings. I was wondering based on the latest event accrual projections you have. Can you be any more specific on the timing of the Stage 2 portion of the gate Stobart reading. And then so on the [indiscernible] had a net trial, when we might expect those within 2026? And following on from that, for each of these if they were to be positive, should we just expect a press release at the time, stating that? Or would we expect any specific data? And then given got a [indiscernible] interim and I believe the fix back as well if these were to pass the interim readout, would we just nothing and then maybe get an update at the next quarterly results. So just exactly when we might expect those and how we should expect an uptick.

Thank you, Harry, for those questions. So we first question on Stage 2 GOT data. and then fix back head and neck and whether each would be likely to have interim data readouts or not. So Ugur, I'll pass over to you for this one.

Yes, yes. I think from the timing, nothing changed, we had guided to the second half of 2026 for [indiscernible] studies. We are on track on -- with regard to the enrollment of the -- in the study, and we are also on track with regard to the event count study, yes. This will be entering read out with -- in both studies, this is challenging hazard payback ratio.

So it is a first income readout. If the internet out is positive, of course, we will document that if the study continues to go we will also inform the market that the [indiscernible] was performed and the study will continue to go on.

Thank you. This concludes today's conference call. Thank you for participating. You may now disconnect.

Welcome to BioNTech's Fourth Quarter Full Year 2025 Earnings Call. I will now hand the call over to Doug Maffei, Vice President, Strategy, Investor Relations. Please go ahead.

Thank you, operator. Welcome to BioNTech's Fourth Quarter and Full Year 2025 Earnings Call. As a reminder, the slides we will be using during this call and the corresponding press release can be found in the Investors section of our website.

On the next slide, you will see our forward-looking statements disclaimer. Additional information about these statements and other risks are described in our filings with the U.S. Securities and Exchange Commission. Forward-looking statements on this call are subject to significant risks and uncertainties and speak only as of the date of this conference call. We undertake no obligation to update or revise any of these statements.

On Slide 3, you will see the agenda for today's call. I'm joined today by the following members of BioNTech's management team, Ugur Sahin, Chief Executive Officer and Co-Founder; Ozlem Tureci, Chief Medical Officer and Co-Founder; and Ramon Zapata, Chief Financial Officer. With this, I'll hand the call over to Ugur.

Thank you, Doug and a warm welcome to everyone as you join us today. As BioNTech has grown, our vision has remained constant, namely translating science into survival. Our focus is on oncology. Cancer remains a contract systems often varying between patients and further within individual tumors. We believe that the future lies innovational restarter project combinations that [indiscernible] posted precise mechanisms of action to achieve biological synergies.

We have [ coveted ] a diversified clinical pipeline, spanning next-generation immunomodulators antibody drug connect and immunotherapies that enable effective, personalized precision medicine and novel-novel combinations across solid tumors. In 2025 and early 2026, we have made strong focus towards realizing our ambitions. The year was marked by important achievements in 4 key areas.

We have maintained our leadership in the core vaccine market and launched our [ Varian ] adapted vaccine in partnership with Pfizer. Our vaccine is now distributed in over 180 countries with more than 50% market share in major markets. Second, we have advanced our oncology programs with the [ registrational Nucleus ] advancing in lungs and breast supported by both clinical evidence base with more than 4,000 patients enrolled across Phase II and Phase III study.

On the other side, we anticipate multiple late-stage events even read out in 2026. In parallel, we now have more than 10 novel combination price with Pumitamig in progress. We executed key strategic deals, most importantly with BMS to strengthen the execution of and help derisk our Pumitamig program. The acquired by sales thus gaining full rights to our cornerstone asset Pumitamig and completed the acquisition of CureVac, strengthening our position in the [ RNAi ].

And last not least, we exceeded our already increased 2025 revenue guidance and ended the year in a strong financial position with more than EUR 17 billion in cash equivalent and securities. We maintained disciplined resource allocation with active portfolio management, focusing on late stage programs that provide a clear potential to drive value appreciation.

For 2026, we are focused on 3 key priorities. The first is to accelerate the late-stage development of our first wave of oncology assets, and we anticipate key late-stage data readout this year. Second is building momentum in our combination-based approach. Most of the data readouts from our novel, novel Pumitamig combination cars are expected this year and that inform our Pumitamig ADC pivotal trials.

Third is to continue our evolution from our [indiscernible] approach to our tumor-centriclinical development program centered around high incident cancers, including lung cancer, breast cancer and other tumors. The foundation of this matrix approach is leveraging our diverse clinical assets for combination processes which will allow us to address several lines of treatment with different combinations.

Our current late-stage pipeline illustrates the growth and robust approach we are taking to advance our ambition to become a multiproduct company. Today, we have a growing set of late-stage and pivotal programs across high incidence tumors with a clear registrational [ part ] and stage expansion options where we believe we can make meaningful difference for patients. We expect a sustained cadence of event-driven late-stage readout across different tumor tests from 2026 to 2030.

Our clinical program provides multiple approval opportunities, and we are building launch readiness now, deepening indication-specific expertise and advancing commercial and market access capabilities, the tumor task that we anticipate launch. Earlier today, we announced [ 10 ] to pursue next-generation mRNA innovations in a new independent company as BioNTech advances towards becoming a multiproduct company by 2030.

The new company will be founded and led by [ SM and me ] and we are both excited at the prospect of this new chapter on our personal journey towards our vision to translate our funds into meaningful advances for patients. In order to do this, our new company will be built with distinct resources, operations and funding options. BioNTech tends to contribute related rights and MR technologies to the new company. In exchange BioNTech will hold a minority stake in the company. This will enable and support private development of these innovative technologies.

The binding agreement is expected to be signed by the end of the first half of this year. NNI will transition to lead our new company by the end of 2026 than our current biotech service agreements and [indiscernible] BioNTech calendars and significant shareholders will remain close to the company. BioNTech will continue to sharpen its strategic focus on the development and commercialization of its own late-stage pipeline spanning innovative immunomodulators, ADCs and mRNA candidates, combination approaches are a core target of BioNTech strategy to maximize the value of our next-generation new oncology own candidate, Pumitamig.

BioNTech's potential stake in the new company will provide both organizations with opportunities to collaborate on combination approaches involving their candidates with the potential to create new complementary or synergistic treatment strategies.

Second, our new companies each have unique capabilities were [indiscernible] leading expertise and the focus on their respective strategic priorities to maximize value for patients and shareholders. Over the past 18 years, we have built BioNTech from a start up into a global biopharmaceutical company with a strong and diversified pipeline. During the COVID-19 pandemic, we expanded beyond oncology to develop the first approved mRNA vaccine, helping to protect people [indiscernible].

None of this would have been possible without the extraordinary dedication of our team, the cost of our shareholders and Supervisory Board and the commitment of the partners who have supported us along the way. Today, BioNTech is well positioned to advance its mission and become a commercial multi-product company. I look forward to updating you on our progress throughout this year. Thank you all. With this, I will hand over to Ozlem for an update on our oncology execution.

Thank you, Ugur. I'm glad to be speaking with everyone today. 2025 was a year where we laid important foundational elements to enable us to execute our strategy in 2026 and beyond. We are executing a synergy-driven development strategy across 3 modalities in oncology. At the core of our approach is the rationale the combination across these modalities can help prevent and address resistance and create conditions for more durable treatment responses ideally translating into better outcomes for cancer patients from early to late stage.

Last year progressed the development of assets across these modalities as one of ERP or in combination with current standard of team. We also gained a better understanding of how to prioritize sequence and stage [indiscernible] our development plans based on evidence, feasibility and potential impact. During 2026, we expect to meaningfully advance our novel, novel combination strategy with multi data set expected.

The [ IO 10 ] tumor backbone of our combination-based development strategy is our PD-L1 VEGF bispecific antibody [indiscernible]. We and our partner, [ BMS], are pursuing a 3-wave plan to develop Pumitamig broadly, deeply and in a differentiated manner across indications, disease areas and treatment lines. Wave 1 is anchored in 3 foundational first-line programs, [ SCLC NSCLC and TNBC], each with a global Phase III trial design for registration and supported by studies that derisks those [indiscernible] setting.

Through these trials, we aim to establish Pumitamig foundational first-line indication in combination with standard of care chemotherapy for global registrational Phase III trial.

Speed to the initial label is the key value inflection point, and it creates the platform for stage evidence let expansion thereafter. In parallel, Wave 2 expands into additional indications. Alongside trials with registrational intent, we are running an expanding set of signal-seeking studies across tumor types to quantify effect size and guide evidence selection of the next registrational opportunities.

Wave 3 comprises novel, novel combination beginning with our in-house ADC. You can already see this happening, and we are also starting to combine Pumitamig with our other next-generation immunomodulators. Wave 3 is designed to build river differentiation and options and where the biology supports it to increase steps and durability of response.

The first 2 ways seek to establish and expand Pumitamig in combination with current standard of care. These trials labor foundation for our novel-novel combination. In 2025, we announced many of these indications and made significant development programs. For non-small cell lung cancer, small cell lung cancer and triple-negative breast cancer, we completed our global Phase II program, selected trade 3 doses and initiated each global Phase III trial.

With our partner, BMS, we were able to accelerate the expansion into other tumor types and settings. In January, we announced our intention to have 8 global Phase III trials running by the end of this year and depending on data from some of the signal-seeking Phase I trial listed here, we may further expand our Phase III program.

Two of the most recently announced Phase III program, further expand our focus in non-small cell lung cancer, an area of high unmet need. Lung cancer has a significant incidence and the majority of patients are diagnosed with late-stage GPs leading to poor long-term survival despite the treatment advances with checkpoint it. Our [ Rosetta LENS-02 ] trial, which is evaluating Pumitamig in combination with chemotherapy as a first-line treatment for patients with metastatic non-small cell lung cancer whose tumors do not have any actionable genomic alterations is well underway.

We are expanding our registrational program with 2 new non-small cell lung cancer trials, the first [ Roselan201 ] is evaluating it as a treatment for patients with stage II resectable non-small cell lung cancer, who have not progressed after cutting on these concurrent chemoradiation therapy. The second [indiscernible] is evaluating puritan as a monotherapy in first plant treatment for patients with PD-L1 high metastatic non-small cell lung cancer. [indiscernible] and expect these 2 trials will initiate this year.

Progress and insights from the first 2 waves, both signed and however first [indiscernible] -- this way seek to elevate Pumitamig's reach and maximize its clinical impact for novel, novel asset compilation. This is where we believe we can have the most meaningful clinical impact and are expecting to make significant progress in 2026. We are well positioned to advance Pumitamig in combination with our in-house ADCs, supported by extensive monotherapy evidence.

Across our first 4 ADC programs, we have generated single-agent clinical data in more than 2,800 patients to date, providing physician grade insight into activity, durability and safety and guiding indication prioritization and combination design. Our primary objective is to combine the ADCs with Pumitamig with registrational pathway. In parallel where monotherapy activity is compelling and clinically meaningful.

We will advance on ADC as a stand-alone opportunity, and there are intact couple of signals which we are encouraged about. For instance, we have evaluated activity and safety of [ N24 ] our [ B7-H3 ] ADC in a broad early-stage development program consistent with the expression profile of B7-H3 across a variety of [indiscernible]. The [ 324 ] has demonstrated [indiscernible] activity and favorable safety profile across a broad range of tumor categorized by low single-digit rates of grade 3 treatment-related adverse events and low rates of any great [ ILD ] moment.

One area of particular interest is metastatic castration-resistant prostate cancer, we observe strong activity in heavily pretreated patients. [indiscernible] the goal of moving to earlier lines of treatment, we have designed a Phase III trial in the first line of this indication. We express a recruitment to begin in the coming weeks. We believe BNT324 is well positioned to address the need for an easily administered, well-tolerated treatment option with the potential for more durable responses.

The progress and wealth of insights we have generated on our ADCs as monotherapy and Pumitamig in combination with chemotherapy has informed our evaluation of Pumitamig plus ADC combination in a number of Phase I/II trials in certain tumors. We apply a light effect of [indiscernible] not signal alone, including effect, [indiscernible] tolerability [ hydro ] addressable population competitive context, operational feasibility and [ CMC ] readiness to nominate the first pivotal combos.

Moving now to our portfolio of innovative mRNA cancer immunotherapies, which aim to activate and educate the immune system more precision. Our personalized approach includes outages of [indiscernible], which is partnered with Roche [ Genentech]. In 2025 and early this year, we published data from multiple traits that support our focus on the adjuvant setting where tumor burden and heterogeneity flows. The biology and our clinical experience point to greater relevance in earlier disease setting, where [indiscernible] tumor burden allows the immune system to consolidate control.

Recently, we and our partner, Roche, the sponsor trial, decided to discontinue the trial in high-risk muscle invasive [indiscernible] carcinoma. The reason for this decision is a rapidly emerging treatment landscape and shifting standard of care. Our other randomized Phase II clinical trials evaluating [indiscernible] in [ actual ] pancreatic ductal adenocarcinoma and actual corrected [indiscernible] continue as [indiscernible], we and our partner Roche Genentech remain committed to the development and advancement of autogenous to address the high unmet medical need in this indication.

In [indiscernible], [ ctDNA ] positive Stage 2 high risk of Stage I colorectal cancer, we have a Phase II trial evaluating [ audigensumeral monotherapy ] against watchful waiting. The final analysis, which primary endpoint is event driven and according to updated projections to be expected in 2027.

For fixed back in first line, HCV positive PD-L1 high head neck cancer. We have a Phase II/III trial in combination with [ pembro ] tumor. Recruitment is ongoing and the Phase III interim analysis is expected in 2026. 2026 will be a year packed with potentially value-creating reach offs and cattle.

In summary, I'd like to highlight a few of our late-stage potential registrational trials. For TPAM, we present Phase II data in endometrial cancer and Phase III interim analysis in HR-positive HER2 low breast cancer later this year. For [indiscernible], we expect a Phase III interim analysis in the second line and beyond games nonsmall-cell lung cancer state. For Pumitamig, we expect a Phase III interim edits from our China trial in first line TNB.

In total, we anticipate 6 readouts from late-stage trials. Looking across our pipeline, we believe the potential to relist survival grids for patients is immense. With that, I will now turn the presentation over to our CFO, Ramon Zapata, for the financial update.

Thank you, Ozlem and a warm welcome to everyone who is joining us today. Today, I will be covering 3 main topics: First, our full year and fourth quarter 2025 financial results. Second, adjustments we will be making to our reporting and guidance going forward and third, our full year 2026 guidance.

Financially, 2025 was a strong year for BioNTech. We exceeded our revenue guidance, which we had raised during the year. We were also in line with our already reduced R&D and SG&A expenses guidance for the year. These results were informed by our active portfolio management and strategy where we are focusing our resources on programs that have the biggest potential to elevate patient outcomes and deliver value for our shareholders.

Also important is our [indiscernible] innovative partnership model which contributed meaningful revenue and cost sharing across multiple products. Our total revenues in 2025 were EUR 2.9 billion, a slight increase from the prior year despite the year-over-year decrease in COVID-19 vaccine revenues. This decline was offset in part by the recognition of EUR 613 million in revenue derived from the noncontingent upfront and anniversary payments from our [ BMS ] collaboration.

R&D expenses were approximately EUR 2.1 billion which is a slight decrease from prior year despite the acceleration of our late-stage oncology programs. This was enabled by cost savings resulting from our active portfolio management as well as positive effects resulting from our Pumitamig cost sharing with BMS. We continue to drive value creation through active portfolio management, shifting towards later stage risk programs that have the potential to really deliver a new era of growth for BioNTech. We ended 2025 with EUR 17.2 billion in cash, cash equivalents and security investments. Our strong financial position and dynamic R&D cost discipline will empower continued investments in our late-stage priority progress and preparations for commercialization of our diversified oncology portfolio.

Starting today, we will be supplementing our IFRS reporting with certain adjusted non-IFRS measures, as you can see on the slide. These adjustments are intended to provide complementary information and context to understand the company's underlying business performance and will be reflected in our guidance mix. These non-IFRS measures will exclude expenses and income from legal proceedings, impairments and reversals, employee-related expenses from restructuring and income from bargain purchase and income and expenses from divestiture-related IPOs.

In 2025, these factors impacted our cost of R&D and mainly our other operating results under IFRS. When excluded, we ended 2025 with an adjusted non-IFRS net loss of EUR 117 million. On the fourth quarter figures, revenues were lower than in the same period previous year, driven by reduced demand for our COVID-19 vaccines. Our R&D expenses were also lower in the last quarter of 2025 compared to Q4 2024. Again, this was mainly driven by cost savings resulting from active portfolio management and positive effects resulting from our cost sharing with BMS.

Turning to the next slide. Let me highlight our financial outlook for 2026. All guidance we provide will be on an adjusted basis. We expect total revenues for 2026 in the range of EUR 2 billion to EUR 2.3 billion. Compared to 2025, we expect the same amount and quarterly timing of revenue from our BMS collaboration but expect lower COVID-19 vaccine revenues.

On other revenues, we expect similar revenues in 2026 from the [indiscernible] prepares contract with the German government and from our services business. However, we do not expect any onetime positive revenue effects such as the payments from prices popped out of our [ Singles ] program that occurred last year. On COVID-19 vaccine revenues, we anticipate lower commodity revenues compared to 2025, driven by declines in both the European and United States markets.

The United States continues to be a competitive and dynamic market. where we expect lower revenues this year as a result of this. In Europe, we expect lower revenues as we defend our market share and begin managing the transition of multiyear contracts. Germany, specifically we recognize direct sales of our COVID-19 vaccine as even. Hence, the anticipated declines in our sales of COVID-19 vaccines in the country will have a direct impact to our top line. Whereas revenues outside of Germany only affect our top line as part of the 50% gross profit split with our [indiscernible] with Pfizer.

In terms of revenue guidance, we anticipate profit COVID-19 vaccine revenues facing similar to last year with the last 4 months of the year driving the full year revenue figure. As in 2025, the EUR 613 million [ VMS ] payment recognition is expected in the third quarter of 2026. [indiscernible] remains a strong brand and a leading global COVID-19 vaccine cultures. Given the lean structure of the business under the collaboration with [ Pfizer ], we have in common, a cash-negative franchise with favorable economics, which we expect to continue as markets adjust to the endemic environment.

Turning to operating expenses. In 2026, we expect adjusted R&D expenses to be in the range of EUR 2.2 billion to EUR 2.5 billion and adjusted SG&A expenses to be in the range of EUR 700 million to EUR 800 million. We expect to increase investment into our priority late-stage programs in 2026 compared to the prior year, namely Pumitamig, our ADC pipeline, mRNA immunotherapies and respective combinations. Consistent with our portfolio prioritization strategy, we also expect to lower [indiscernible] outside of our priority areas [indiscernible]. We will continue to follow the data generated by [ 5G ].

As part of this prioritization efforts, we follow a rigorous go/no-go decision-making process across all development stages. This allows us to focus on the programs, which we believe represent strong best opportunities, reserve cash and have strategic flexibility to assess inorganic opportunities as they come through. Our SG&A spend will be driven by our commercial build-out for oncology and preparations for our first oncology launch.

2025 was a year of great progress during which we advanced important components to empower the execution of our strategy. We advanced our pipeline while derisking our R&D investments and efforts. We progressed key programs into pivotal stage established our partnership with [ BMS], all while maintaining a strong balance sheet. During 2026, we will continue to focus on driving our execution at scale and speed by accelerating pivotal trials, advancing combination of therapies and continuing to build indication-specific oncology portfolios.

We are energized as we look towards a phase of sustained clinical data output from 2026 to 2029. By 2030, we envisioned BioNTech as a diversified multiproduct company focused on achieving long-term sustainable growth and generating value for patients and shareholders. Lastly, before opening the call for the Q&A, on behalf of the Management Board, I would like to thank Ugur and Ozlem for what they have built here at BioNTech.

Your vision talent, dedication and [indiscernible] pursuit of excellence has had a lasting impact on the world and all of us. We are excited to see and support what comes next. BioNTech is in an optimal position to execute this next phase of growth. You have truly inspired us all to be both and to continue to push the boundaries of what we believe is possible. With that, we would like to open the floor for questions.

[Operator Instructions]. We will now take the first question from the line of Daina Graybosch from Leerink Partners.

Well, thank you for the question. Congratulations to Ugur on your new pursuit. I'm excited to see where you take it. But certainly, it feels like a transition today. And I think I'll ask my question there. So can you help us better understand how you'll split the mRNA therapeutics what remains in the parent BioNTech? And what kind of innovation will you take to pursue in the new company.

Okay. Thank you, Daina. So Ugur, I think that's one for you in terms of what you -- what could potentially go to the new company from mRNA technologies.

Daina, great to hear you. So first of all, there is no -- nothing that is going to change from the BioNTech perspective. As I'm saying what is visible today. As the state is BioNTech, clinical, clinical and everything what we have communicated so far. I don't want to speak too much about a new upcoming company because this is not disclosed and it is still under discussion.

But you can imagine, Daina is -- and you are very close to that, that the [indiscernible] in the M&A space is rated advancing and we are seeing a lot of innovation happening, particularly in combination with AI. And we together felt really the need to address that M&I focusing on this sector on this type of endeavor to ensure that we can use basic technologies and basic IT that comes from BioNTech to build something completely new and completely [indiscernible] means really next generation and we call it the next-generation everything that goes beyond the current generation. So that's the idea.

If you allow me to add-on, on the topic, thank you, Daina, for the question. So I think it's just to reconfirm that there is no split of BioNTech [indiscernible]. Our strategy and pipeline remains unchanged. If the income [indiscernible], we retain our mRNA oncology programs. And what is being discussed, as Ugur was just mentioning, with this new company related to certain rights and mRNA technologies to advance next-gen innovation while BioNTech continues to focus on executing its late-stage pipeline and, of course, preparing all of this for commercialization.

I think it's also worth mentioning that we will continue to innovate. In BioNTech that we have our innovation engines in Germany, in China and in the U.S., we will continue to deepen our efforts and pipeline in our immunomodulators our ADCs and our mRNA technologies and progress.

We will now take the next question from the line of Tazeen Ahmad from Bank of America.

Another one maybe about how you're thinking about management of the company to the search for the new CEO? Are you looking at internal candidates or do you think that you would want somebody external? What is the profile that we should be thinking about for who you think should be leading the company into its next phase?

And then one question about TPAM. How are you preparing for that launch in endometrial cancer? And is that going to serve as sort of an infrastructure build for other launches? Or is this just going to be tailored for this particular launch?

Okay. Thank you, Tazeen. So just to confirm, we've got one question on the search for replacement CEO and CMO and the criteria and then on TPAM prep for EC and whether the infrastructure is just for that launch or for future launches as well.

Thank you, Tazeen, for the question. So your first part, Ugur and Ozlem will remain in the role through the transition period. And the Supervisory Board has already initiated executive search to identify the next leadership on their successors. [ Focus ] is our leaders with strong experience in late-stage development and commercial execution which reflects BioNTech's next phase of growth. At the same time, now with [indiscernible] and all of our commercial teams, we are already preparing the organization for these potential launches, including endometrial cancer and other programs as we are reading the commercial, medical and market access capabilities needed to support all of this pipeline coming through.

We will now take the next question from the line of Asad Haider from Goldman Sachs.

Congratulations on the move and best of luck. I guess just one question, high level, just on the timing of the departure. Just seems like it's a very critical time for the company for a transition given all the repositioning in recent months and the momentum in the late-stage pipeline. So I guess the question I have to ask us why now ahead of very important readouts and the need for very precise execution during this important time. Thank you.

Thanks, Asad. So that was a question around timing. So it's a critical time for the company, which we recognize getting ready for the launches of certain products and Pumitamig why make this decision now.

I take over the question and then Ramon, you can add. I think from timing wise, we are talking now end of 2026 and not today, yes. So we have a clear plan for milestones and data readouts in 2026. And we believe it's really a perfect timing for transition -- transitioning because the company, at the time point end of 2026, we'll already have a number of weeks important us, but also the number of Phase III studies that we plan end of 2026, 15-plus Phase III clinical [indiscernible] and this is really about industrialization and we need to get people on board who connect that -- connect this with the scale that is needed at that moment.

So I would add to the answer. I think the plan aligns with BioNTech's continued efforts to sharpen our strategic focus on our growing line stage pipeline. And as you know, this is spanning innovative and [indiscernible], as I was mentioning, ADCs and candidates. So now you have 2 companies focusing on these priorities and 2 tailored investment cases. And I don't expect to maximize value for patients and shareholders alike. And in terms of our collaboration and contribution to the new costs, we also retain the possibility to participate in new cost of side to its minority stake. So I hope this provides clarity.

We will now take the next question from the line of Cory Kasimov from Evercore ISI.

First, just a quick clarification question. I just want to be clear, does BioNTech contribute any capital to this new company? Or is it just planning to be a minority investor. And then on the pipeline front regarding [ got], if you were to replicate the results you saw in Part 1 and Part 2 of the study, how do you think about the market opportunity in second line plus squamous non-small cell lung cancer?

Okay. Great. Thank you, Cory. So just to clarify. First question was whether we plan to whether biotech plans to contribute any capital to the new company or whether it's a minority stake. Second question on [indiscernible]. If we were to revisulate the Part 1 results in part 2, what do we anticipate the market opportunity would be.

Thank you for the questions. So let me answer the first one. The short answer would be no. Based on what is contemplated today, BioNTech's contribution to the [ NewCo ] related to certain rights and RNA technologies, not cash. The new company will have the ability to pursue funding from other resources, while BioNTech remains focused on advancing our late-stage pipeline and keep preparing for commercialization and further innovation in our key priorities. [indiscernible] over?

Yes. I think I said how difficult second [indiscernible] lung cancer is, there is more disruptive innovation space for almost [ 20 ] years now. And if the data are dedicated with [indiscernible] in the range of 0.5. This will be a disruption. It will be game changing for patients. And as you know, this is a very sizable patient population in non-small cell lung cancer. So we will come up this market. Market projections once we really see that a result.

We will now take from the line of Geoff Meacham from Citigroup.

This is [ Gary ] on for Geoff. Maybe a question on the management transition. I know during the call, you guys mentioned the prioritization of R&D efforts. And I guess given the upcoming transition, how should we feel about the current late-stage pipeline prioritization and mid-stage pipeline prioritization versus I guess, stability of it looking ahead.

And then maybe another question on [indiscernible]. If the interim data were positive, could that open an avenue for accelerated regulatory filing?

Okay. Thank you for the question. So we have one on the Management Board transition and then the second question on portfolio prioritization efforts and current late stage versus mid stage. So you -- let me take the first part of the question, and then I'll allow [indiscernible] to take the second one. I think in terms of priorities, in particular in strategic priorities, this transition does not change any of these at all. BioNTech remains focused on advancing again the late-stage pipeline and our mRNA oncology programs, where we continue to defend commit and prepare for commercialization.

Reorganization, our governance structures, our scientific leadership that Ugur and Ozlem have been building over the past years provides the stability that we need to bring this next -- to bring all the pipeline for the next stages of either innovation or development or commercialization and we -- and as another will continue to lead the company through the transition period when supervisory board conductors for these successors. I believe that the company is well positioned to continue to move these programs to the different stages at speed and with the right focus and really making this as available as soon as possible to our patients. I can take the second question. So -- Ozlem, you want to talk?

Yes. Yes, I can also take it. So depending, obviously, on the data we will see the later number interim analysis of [ Vergote ] study. And if we can replicate the data we have shown in the initial part of the study, there is absolutely a potential regulatory path forward for an accelerated [indiscernible].

In this second part of the session with regard to the pipeline, so we have really built an extremely rich pipeline. And the pipeline is not only individual drugs, but we believe is our established, expand and elevate strategy. We are building a combination approach that could allow us now and by transitioning from Phase II into Stage 3, really address multiple indication spaces with our current pipeline. There have a number of test assets, including again, next-generation IO molecules, including again, [ ADCs ] that we have in our pain but never shared data [indiscernible]. But you will hear also in the early-stage clinical -- for the early-stage clinical assets in the end of this year, beginning next year late after.

We will now take the next question from the line of Terence Flynn from Morgan Stanley.

This is Chris on for Terence. We have a 2-part question for auto trial in colorectal cancer. Just kind of wondering what level of details are you planning to give for the update in early 2026? And then for the DFS primary endpoint, how do you define the [indiscernible] success?

Thank you, Chris. So that was [indiscernible] in CRC. So what level of details in early 2026? And what is the bar for success?

So our final analysis will be later. We have just updated the projections based on the current accrual rate or event rate to be more precise for early 2027. And this is the time point where we expect to have robust data or area analysis, which is an interim analysis will just guide us to continue the trial, however, will not be the basis for any steps towards based on efficacy data. And what we -- our objective is that we want to be statistically significantly and clinically meaningfully better than the standard of care with regard to DFS.

We will now take the next question from the line of Evan Seigerman from BMO Capital Markets.

I wanted to touch on the upcoming Phase III interim data for [ BNT113 ] in line head and neck [ HNSCC]. Can you talk about some expectations for this interim analysis? Could we potentially see 6- or 12-month OS data? And more importantly, how are you thinking about the potential trade-offs on efficacy and safety here? I know there's been a lot of development in head and neck. So I just want to understand how you're trying to position the product relative on efficacy and safety.

Thank you, Evan. So just to confirm that was a question on 113, frontline head and neck and our expectations for the data from the interim analysis and also any perspective that we have on the trade-off of efficacy and safety. Ugur, would you like to take that one?

Yes. Yes. So this is a PFS-based event-based endpoint and that we expect in the late second half 2026 and patient population is HPV-positive cancer patients. You know that this is a patient population that is increasing in the industrial space. And depending on the headout ratio, this could give us a path towards rigs station and also depending on the further readouts that are [ later tons ] that are based on OS, which could also get us path based on a full approval on OS. So we are very curious about the outcome of this trial. We have accomplished a ton immunogenicity data in the station population and so this is a potentially the biggest patient in time and important readout.

We will now take the next question from the line of Yaron Werber from TD Cowen.

I just have a couple of questions. The first one on [ Rosetta ] lung I see the study was now expanded to 1,260 patients from 986 or so and data is now in fiscal year '29. Was that -- is one histology expanded or are both of them equally expanded? And what was the reason to do so? I think it makes sense given the expansion from your competitor? And also what data should we expect in the Phase II endometrial cancer team this year.

Thank you, Jerome. So to confirm, these are questions from for Ozlem. And so the first people in [indiscernible] on the rationale behind the expanded study across both histologies and then next question was on any potential data from TPAM in HPC this year.

Yes. Regarding the in -- our non-small lung cancer study. We are constantly assessing available data of emerging data from our trials and also from other trials with this very specific antibody class and based on this data, we expanded the [indiscernible] also to increase speed. It was an increase for both for both these policies. We also made trial design with regards to the endpoint, we have PFS now is primary endpoint and a key secondary endpoint, which also helps with the speed. That means the rationale for both amendments is changed statistical considerations and also recalibration of recruitment.

The second question was about TPAM. I guess specifically about TPAM in endometrial cancer, second-line endometrial cancer, our data package, which we plan to submit for BLA this year. we have also, in parallel, initiated a conformational trial, a Phase III trial forward an indication, which is ongoing and our plans remain unchanged.

We will now take the next question from the line of Akash Tewari from Jefferies.

This is Manoj on for Akash. Just one from outside. Are you still planning to take [indiscernible],ADC combo to registrational studies in lung indication? And also, do you expect any revenues from cancer vaccines BNT113 and BNT122 in 2026, any accelerated approval [indiscernible] your base case.

Okay. Thanks, Manoj. So I hope to hear all of that, but what I thought you -- what I gathered from that is on 327, any plans to take a combination into registrational lung study -- sorry. 324. Okay. Sorry, yes, okay. And then a second question was whether we expect any revenue from cancer vaccines in 2026. So Ramon, that one could be for you.

Yes. Hopefully before, as you know, is evaluated with 327 in multiple [ candaindication], including lung cancer. We are expecting data here in the second half of 2026. And of course, we are prepared if we see a strong signal to transition from Phase II and Phase III plant.

And then in terms of potential revenues from our cancer vaccines, they are still -- all the orders are still in the clinical development stage. So we do not expect revenue from them in 2026. The value from assets such as VP, they are affected in the clinical milestones and potential approval as we are working towards rather than near-term revenue contribution.

We will now take the next question from the line of Asthika Goonewardene from Truist.

This is Kari on for Asthika. Just a couple of questions from my first [indiscernible] Sinise company. Would there be any milestone or royalty economics tied to the IP by lung to BioNTech? And second, on COVID sales, how large do you expect a step down to be versus 2025? And how do you think about the relative pressure coming from U.S. versus Europe and versus Germany?

First question on the new company, would there be any milestone or royalty economics tied to the IP that belong to BioNTech? And second, on COVID sales, how large do you expect a step down to be versus 2025? How should we think about the relative pressure coming from the U.S. versus EU versus Germany?

So thank you for the questions. So on your first question, we are not providing any specific financial guidance related to the potential answer of the related rights and RNA technologies onto the nuclear at this stage. The terms of the transaction, including any potential IP-related consideration is still in the negotiation and will be defined as part of the binding agreement expected by the end of the first half of 2026.

What we can say is that we do not expect a material short or mid-term financial impact for BioNTech. Now in relation to your question around [indiscernible] and COVID-19. So we do expect lower COVID-19 vaccine revenues in 2026 compared with prior years as the market continues to normalize and demand becomes more seasonal. And so saying that Kogenate remains an important franchise for us. We continue to generate meaningful cash flows. We continue to have very meaningful market shares. And it's helping us to fund ongoing R&D investments in a big way.

So our focus is really to managing this transition where we continue to advance our oncology pipeline and prepare for potential launches. If I will go a little bit more in detail so -- to your question, the United States, the U.S. is a competitive and a very dynamic market where we expect lower revenue this year as a result of all of this. And then in Europe, as I was mentioning during the presentation, we expect lower revenues as we are defending our market share and begin the managing on the transition for multiyear contracts to now a more seasonal demand of utilization and revenues.

And then this year, specifically, Germany is an important effect because we recognize direct sales of our improving COVID-19 vaccine directly. So the anticipated declines in our top line in the country will have a direct impact on our overall top line, whereas the revenues that are coming outside of Germany only hit our top line as part of the 50% gross profit split mineral size. And that's as much of time I can give you.

We will now take the last question from the line of Mohit Bansal from Wells Fargo.

I have 2, if I may, one from the science side. So -- for the [ Revitalon 2 ] trial, does this make sense to do separate trials for squamous and [indiscernible]? And do you think that there is a lower bar to be successful in squamous trial? That's the first question.

And second question for Ramon, how -- what is your thought process here to do a buyback or some kind of special dividend there given that cash position and your cash requirements going forward.

Okay. Great. Thank you, Mohit. So 2 questions in there. One for Ramon on any thoughts around the potential buyback -- share buyback given our cash. And then another question for Ozlem on [indiscernible] on squamous [indiscernible] to be successful in squamous.

So I'll let Ozlem to answer the question about [indiscernible].

Yes, Mohit, thank you for the question. We -- you asked whether it makes sense to have 2 different studies. We, in fact, have in this study that allow both histologies separated. So it's technically like 2 studies in 1, which gives us the best problems between speed and probability of success.

Thank you, Ozlem. And then on the capital allocation strategy and priorities. So this remains focused on advancing our late-stage oncology pipeline, preparing the organization for potential launches and defending our comment [indiscernible]. We believe our pipeline on all of these programs can drive the next stage of rules for BioNTech as they so get the resources they need, what they need it.

And outside of that, as we have done in the past, if there are assets or technologies that could help our late-stage programs be best positioned we may look at ways to access those assets or technologies through strategic inorganic transactions, but to strengthen our early science pipeline. So no change in there.

Thank you. That's all the time we have for questions. I would like to hand back over to the speakers for closing remarks.

Well, I think -- so it was, of course, an important day of announcements. I would like to, of course, thank all of you for your continued interest in BioNTech. As you hear today, we are entering an important phase for the company with multiple late-stage programs progressing and key readouts ahead to meet a [indiscernible] the backbone of all of these efforts our strong collaboration with BMS on late-stage execution, the next combination with Pumitamig and then continue our strong -- our strategy of having this strong balance sheet focused strategy and the strength and the strengthening of our teams, our partners and our governance so that we remain confident in our ability to advance our pipeline and most BioNTech over becoming a multiproduct oncology company by 2030. We really appreciate your time today and really look going forward to updating you on the progress in the quarter ahead.

This concludes today's conference call. Thank you for participating. You may now disconnect.

Great. Good afternoon. Welcome, everyone. My name is Jess Fye. I'm a biotech analyst at JPMorgan, and we're continuing our 44th Annual Healthcare Conference today with BioNTech.

First, you're going to hear a presentation from the company, and then we're going to go into some Q&A. So if you have a question in the room, just raise your hand and someone will bring you a microphone or alternatively, you can submit it online, and I'll read it off the iPad up here.

So with that out of the way, I'm pleased to introduce BioNTech's CEO, Ugur Sahin.

Hello, everyone. It's a pleasure to be here. I would like to welcome everyone here and online. These are the typical disclaimers. And I would like to start with the -- can we just change what is on the monitor?

Okay. So let me start with the status quo that we have here. While BioNTech has grown and evolved, our vision has remained the same, namely translating science into survival. From the current situation, from today's perspective, we have so far delivered together with our partner, Pfizer, 5 billion doses of vaccines worldwide. We have engaged with our mRNA technology in global health programs, have 6 programs here running. Most importantly, we pivoted back to oncology, have currently 25 clinical trials ongoing and 16 clinical programs. We have also integrated AI and advanced GMP manufacturing capabilities, providing us the opportunity to rapidly translate science into clinical trials.

So our key achievements in 2025, it was a strong, eventful year. We maintained the COVID market leadership and launched our variant vaccine. We advanced our oncology programs with, in the meantime, 25 Phase II and III clinical trials running, with more than 10 novel combination therapies. And we executed strategic deals. The most important, the partnership with BMS, which gave us the opportunity to strengthen the execution, but also derisk our program. We acquired Biotheus, which helped us to get the full rights for our key molecule, Pumitamig, and we acquired CureVac, thereby strengthening our position in the mRNA field. Moreover, we increased the 2025 revenue guidance and ended the year with a strong financial position, with more than EUR 17 billion in cash and cash equivalents and securities.

So a little bit more into the details. So in 2025, we have delivered in the meantime, our fifth very adaptive vaccine. So this is getting to become routine. We maintained our leadership and distributed our vaccine in more than 180 countries. What is really important and what I would like to emphasize here, the evidence for the clinical benefit of our vaccines is overwhelming, and there are hundreds of studies about this. But recently, there is a study published which is really worth mentioning. It's a study performed in France. It's a real-world study in about 27 million adults, 22 million who received the vaccine, 5 million who did not receive the vaccine. And we see that COVID vaccines are really saving lives even in the population of 18 to 59-year-old individuals. So the overall mortality was reduced, the COVID-associated mortality was reduced more than 70%, which is really impressive and shows that we can translate our mRNA science into survival.

With regard to clinical execution, we advanced our late-stage programs. We have now 25 late-stage programs running. More than 4,000 patients have been enrolled in our clinical trials. And the result of this is that we are expecting now 15 Phase II/III readout in the years of 2026 and 2027.

So we continued also to run the company with financial discipline. We continue to use the revenues from our COVID vaccines to finance our oncology programs. We strengthen our P&L through the partnership and cost sharing with BMS. We have a very strong cash balance, helping us in the oncology execution, and we maintained disciplined resource allocation, with an active portfolio management focusing on late-stage programs that really provide a clear value increase.

So 2026, what are the key objectives in 2026? We have 3 priority goals: first, accelerate late-stage development of our first wave of oncology assets; second, build momentum by engaging into multiple combination strategies; and third, shift our approach from a platform-centric to a tumor-centric clinical development approach. So I will show you in a few minutes what this all includes.

This is just a view on our current pipeline. We have the ambition to become a multiproduct oncology company. Today, we have more than 17 late-stage and pivotal trials across multiple high prevalence solid tumors. This includes, of course, areas of unmet medical need where we believe we can make a meaningful difference for patients. In parallel to this development, we are pursuing multiple combination trials, and we are building commercial capabilities in selected tumor types where we anticipate our first launches.

So we are executing for our oncology portfolio a synergy-driven development strategy across 3 modalities. These are next-generation immunomodulators, targeted therapies with a key focus on ADCs and mRNA cancer immunotherapies covering fixed combinations as well as personalized mRNA vaccines. The core of this strategy is the rationale that combination across these modalities can help to prevent or overcome resistance and create conditions for more durable responses, ideally translating into better outcomes for patients. The additional advantage of the strategy is that execution risk is mitigated through diversification across assets, indications, modalities and partnerships, reducing the dependence on any single agent.

So this is the pipeline that we built. It's a differentiated oncology pipeline. We have for our immune modulators, 2 late-stage assets, Pumitamig, PD-L1 VEGF antibody, bispecific antibody partnered with BMS; and Gotistobart, CTLA-4 targeting Treg depleting molecule partnered with OncoC4. Both molecules are designed as next-generation immuno-oncology backbone. We have a growing ADC portfolio, including programs directed against HER2, HER3, B7-H3, TROP2 and other targets, allowing us to execute ADC-based pan-tumor strategy. Finally, our mRNA immunotherapy programs include individualized vaccines that are currently tested in 3 indications: CRC, pancreatic cancer and bladder cancer; and FixVac approaches where we focus currently on lung cancer.

Today, I'm going to walk you through 3 programs, which are visible on the slide, Pumitamig, Gotistobart and BNT324/DB-1311. So starting with Pumitamig, it's a PD-L1 VEGF bispecific antibody, which we believe has the potential to become the next-generation IO standard across multiple cancer indications. The molecule is an Fc silence bispecific antibody, which is engineered to deliver dual blockade of VEGF and PD-L1 as a single molecule. Our preclinical work supports a differentiated mechanism, which includes complex formation by the bispecific molecule that increases the internalization of PD-L1, a mechanism that cannot be done by the combination of the 2 antibodies.

Clinically, Pumitamig has been studied broadly and has shown impressive efficacy signals across indications regardless of PD-L1 expression. Our development momentum is strong. With BMS, we anticipate 8 pivotal studies will be ongoing by the year-end 2026, along with a broad portfolio of combination therapies. This is just one example of efficacy data, which we have recently announced in a global Phase II study where we have observed 85% objective response rate in first-line small cell lung cancer and 70% objective response rate in triple-negative breast cancer with activities reported irrespective of the PD-L1 level. These response rates offer best-in-class potential for this difficult-to-treat patient population.

So to fully realize Pumitamig's potential, we are executing a staggered 3-wave development strategy. Wave 1 establishes chemotherapy combinations in priority indications such as SCLC, NSCLC and TNBC. Wave 2 is expanding into more than 10 additional indications, mostly in combination with chemo. And Wave 3, we'll explore novel-novel combinations to further deepen and broaden the activity across tumor types. Our goal is to meaningfully expand the number of patients who can benefit from effective cancer immunotherapy. We believe that Pumitamig has the potential to both replace first-generation immuno-oncology drugs in multiple indications; and second, expand the reach of IO into new settings. In both settings, there is a substantial unmet medical need driven by different high incidence cancers.

Coming now to Gotistobart. While Pumitamig is intended to improve on checkpoint blockade, Gotistobart is an antibody which goes beyond checkpoint blockade. It's an antibody partnered with OncoC4. It has a differentiated mode of action by selective killing of regulatory T cells in the tumor microenvironment. And I would like to clearly state this is not just another CTLA-4 blockade. This is a mechanism of killing of regulatory T cells that allows to address multiple mechanisms because regulatory T cells not only use CTLA-4, but they act as an IL-2 trap. They have inhibition by cell-cell contact. So by eliminating regulatory T cells, we accomplish a much more potent mechanism.

Together with our partner, OncoC4, we have generated data in more than 1,000 patients so far. The program is now in pivotal phase development in second-line squamous non-small cell lung cancer and has received recently from the FDA orphan drug designation. In parallel, we are evaluating Gotistobart in proof-of-concept combinations, including radioligand therapies, mRNA immunotherapies and ADCs.

These are data from an unblinded Stage 1 analysis of our Phase III study in squamous non-small cell lung cancer, where Gotistobart was evaluated in a randomized fashion against docetaxel, which is the standard of care. It showed a 54% reduction of the risk of death versus chemotherapy, which represents a clinically meaningful benefit in a population with very limited treatment options. The first interim data for efficacy readout is expected in 2026.

So these data are extremely relevant in setting of a major unmet need. Patients with squamous non-small cell lung cancer who progressed after prior checkpoint therapy have limited options. And multiple Phase III programs in this population have failed in the recent years. Standard of care is typically chemotherapy, with response rates generally below 20% and overall survival of around 10 months. So if the Phase III results from Part 1 are confirmed, we believe Gotistobart could offer a transformation chemotherapy-free treatment option for this patient population.

The third molecule that I would like to introduce here is our B7-H3 ADC partnered with Duality that has shown pan-tumor activity and favorable safety profile. B7-H3 is a target expressed across multiple solid tumors. It is one of the most interesting ADC targets. This includes tumors like prostate cancer, lung cancer, small cell lung cancer, GI cancer, gynecologic tumors, making the tumor interesting as a pan tumor target. In partnership with duality, we are evaluating currently BNT324 in more than 10 cancer indications. While several B7-H3 ADCs are in development, we believe that BNT324 is differentiated based on its emerging safety and efficacy profile.

Across the studies today, we have observed encouraging clinical activity in multiple tumor types with a low single-digit rates of Grade 3-4 treatment-related adverse events and low rates of ILD/pneumonitis. What is important is that patients with sustained clinical benefit have been treated now for up to a year and longer, supporting the potential of this molecule as a longer-term targeted therapy across multiple cancer indications.

In terms of efficacy, we have seen encouraging clinical activity across multiple tumor types and combination studies with Pumitamig are ongoing. One area of particular interest is metastatic castration-resistant prostate cancer, where we have observed strong activity in heavily pretreated patients. With our goal moving into earlier lines of treatment, we have designed a Phase III clinical trial in first-line mCRPC. The IND for this Phase III trial has been cleared, and we expect the recruitment to start in the coming months.

Here again, this is an important indication within the prostate cancer spectrum. Metastatic castration-resistant prostate cancer remains one of the highest unmet medical needs. The treatment paradigm is currently shifting towards more intensive therapies earlier. For example, docetaxel is increasingly used in the first-line setting for appropriate patients. At the same time, many patients are not eligible for chemotherapy or prefer to avoid it. We believe BNT324 is well positioned to address the need for an easily accessible, well-tolerated treatment option with the potential for more durable responses.

We view this single agent development as an important first step to establish clinical value and build market presence. Over time, we expect combination regimens to unlock the additional benefit to therapeutic synergy and support broader adoption. As our oncology strategy is built on synergy-driven combination approach, we are systematically testing each modality in combination with each other, which means we combine ADCs with IO, mRNA with IO or mRNA with ADCs.

This is here a table showing our matrix approach across priority tumor indications, including lung, breast, genitourinary, gastrointestinal and gynecologic cancers where we paired our ADC candidates with Pumitamig. Our goal is to identify the ADC combinations and dose levels that deliver the best balance of efficacy and tolerability in combination with Pumitamig. We established a strong preclinical validation of this combination approach and anticipate multiple Phase I/II readouts in 2025, and this data will help us guiding the decision and inform Phase III trial decisions.

In addition to ADCs, we believe that mRNA cancer vaccines are ideal combination partners for Pumitamig and our ADCs, driving robust tumor-specific immunity that enhances checkpoint inhibition. Our individualized therapy, Autogen Cevumeran induced both neoantigen-specific T cell responses across multiple tumor types. And furthermore, we have shown that combination of our FixVac, BNT116 lung cancer vaccine with anti-PD-1 demonstrated encouraging survival in patients with first-line and second-line non-small cell lung cancer indications, which usually have OS, second-line non-small cell lung cancer with OS in the range of 12 months.

So this combination strategies will become the foundation of our tumor-focused approach. We have built tumor area strategies, which are centered around high incidence cancers like lung cancer, breast cancer and other tumors. And we are developing now a strategy where we address several lines of treatment with different combinations. Let me illustrate that with an example in lung cancer. Lung cancer is a heterogeneous disease and remains the leading cause of cancer-related mortality. Many patients are diagnosed with late-stage disease and long-term outcomes remain poor despite advances from checkpoint inhibitors and target therapies.

We are building here a durable position in this indication. Across lung cancer indications, we use Pumitamig, Gotistobart, our ADC portfolio and BNT116 either as monotherapies in combination with chemotherapy or as combination partners in different setting. Our first registrational studies for Pumitamig are in combination with chemotherapy. In parallel, we are advancing this molecule with ADC combinations in second and third line to increase the optionality for patients.

So let me turn to what we expect to deliver in 2026. 2026 is packed with value-creating catalysts in late-stage readouts. We have 5 late-stage readouts, including our HER2 ADC TPAM, our anti-CTLA-4 Gotistobart, our HPV vaccine for head and neck cancer BNT113, Pumitamig from data from a Chinese study in TNBC and our personalized vaccine for treatment of high-risk colorectal cancer. This late-stage trial readout will be complemented with early-stage combination therapies, including Pumitamig plus different type of ADC trial readouts. In total, we anticipate 15-plus data readouts here. This should result in a steady news flow throughout the year, which will support us in rapid decision-making and meaningful value creation.

So our ambition is to build a fully integrated multiproduct oncology company and 2026 is just the beginning of our road map. As of today, we expect 17 late-stage and pivotal trial readouts across different tumor types. We are entering a phase of sustained clinical data output from 2026 to 2029, and we will provide regular updates on execution and progress. This pipeline supports multiple approval opportunities, and we are building launch readiness now, developing indication-specific expertise and advancing market access capabilities in the tumor types where we anticipate first launches.

Zooming out, 2025 provided the foundation for our strategy. We advanced our pipeline and derisk development. We progressed key programs into pivotal stage, established partnerships with BMS, increased our balance sheet and to fund our pipeline. 2026 and 2029 will be on the focus of driving our execution at scale and speed with advancing combination therapies, accelerating pivotal trials and building indication-specific oncology portfolios. For 2030, we expect that BioNTech will be a diversified multiproduct technology company. And we will use this data-based approach to ensure that we can fully leverage our pipeline and generate the data required for registration.

Thank you for your attention.

Great. And as a reminder, if you have a question in the room, just raise your hand, someone will bring you a camera. Great. So I guess maybe starting with Pumitamig. In what settings thus far do you see Pumitamig's data as most differentiated?

Okay. I think we have generated data now in multiple cancer indications. And what we see is what impresses us most is really the activity of the compound in cancer indications and patient populations, which have a rather low PD-L1 expression. So what is really impressive is, for example, the clinical readout in the TNBC patient population below 10%, where we have seen from China readouts that we have a PFS of 13 months. OS is going more into the direction of 20-plus months and we have been recently able to reproduce the data in a global trial. And we see this pattern now occurring in different indications.

Overall, what also impresses me is that in more than 10 clinical indications where we combined Pumitamig with chemotherapy, we have almost clinical benefit rates in the range of 80% to 100%, which really shows that this molecule provides an opportunity even for aggressive tumors to provide stable disease and strong objective responses.

And maybe, Jessica, if you allow me, we would move one step forward. I'm Ramon Zapata, I'm the CFO of BioNTech since mid last year. The other thing that I believe is very exciting is, of course, not just about the assets, but the approach that we are taking in the combination strategy that we have. I think, of course, there are other companies that now they have PD-L1s or PD-1s, but nobody has the breadth and depth of the ADC and the assets where we can combine novel approaches with Pumitamig, as well as the strength that the collaboration with BMS gives us in terms to go in a more efficient way after the development approaches in this late-stage program clinical studies as well as the conviction to go commercially very deep and wide with the asset.

What about relative to ivonescimab, how much of a read-through do you see from the data for that product to Pumitamig's potential? And where do you see the products differentiating so far?

The general rules seem to be class-class specific. So the rules are improved disease control rate, improved objective responses, improved PFS in multiple indications and I would say a clear trend for improved overall survival. And this is something that we are also seeing. Of course, we are not doing any head-to-head studies. Therefore, it's difficult given the heterogeneity of the patient population to draw conclusions.

But the mechanism of action of our compound provides a mechanistic advantage because our compound is not only an immune modulator, but also a targeted therapy. It is directed against PD-L1. PD-L1 is highly expressed in tumors. It's highly expressed on tumor-associated macrophages. So we are targeting with this molecule, the tumor microenvironment directly and have the opportunity to bring in VEGF inhibition and all the related immune suppression blockade into the tumor microenvironment. And we believe that this actually could translate to a prolonged control, but this is just a hypothesis, and we have to see.

So kind of sticking with that, what metric would we look to kind of see that differentiation play out?

I think we will not see in the next 5 years head-to-head studies of one bispecific against the other bispecifics. So this would be the most fundamental metric doing it. But at the end of the day, I believe there are 3 ways of differentiation. It's the engineering of the molecule with the mode of action, then the breadth of development in different indications. We will have -- at the end of this year, we will have 8 pivotal studies in multiple indications. And it, of course, counts in how many indications you are positioning a molecule.

And the third, and this might be the most important is the combination approach. So we are using an unprecedented combination therapy assessment approach. with multiple ADCs with our mRNA vaccines and with other molecules. And the partnership with BMS allows us really to use -- to fully exploit our portfolio of compounds and the BMS portfolio of compounds, and we are completely open also to consider other compounds.

So I believe at the end of the day, while entering with chemotherapy combinations into and providing clinical benefit early on with chemotherapy combinations, the overall trend will be combination with this type of molecules. And this is something that will go on at least the next 5 to 8 years because we really believe that we have for the first time in oncology, the opportunity to get do control in patients in first line, but also open up clinical benefit for patients who are not responding to checkpoint blockade. So this is a huge opportunity. And there is, in my view, enough room for several players in the field.

So maybe turning to the ADCs. So with multiple ADCs in development, can you talk about how you're prioritizing which to move forward as monotherapy and which to study in combination with Pumitamig?

Yes, sure. So we have generated ADC data in more than 1,200 patients so far. We are systematically analyzing multiple cancer indications, different doses. We very well understand the safety efficacy profile of this molecule in different indications. Our strategy is if we really see that a compound has a differentiated mono single-agent activity in an indication with high unmet medical need, we are ready to go with the compound alone. So this is an example with prostate cancer, where we have seen that even in late-stage patients, we see significant objective responses, excellent PFS trend. So this was the decision to go into these indications. We have other indications where we see similar signals in advanced cancers where we might come up with additional mono activities.

For combining with Pumitamig, we are actually doing really the side-by-side studies. So we are evaluating, for example, our HER3 with Pumitamig and we are evaluating our B7-H3 with Pumitamig. One of the learnings of this bispecific antibody is that actually preclinical work is not always able to predict what is going to happen in the clinic. So having an experimental approach and comparing the data head-to-head in different indications is the best way so that we not just jump on top 2 ADCs, which are at the moment used for combination therapies, but also consider other options.

So you laid out a bunch of upcoming milestones. But with respect to your ADC pipeline, just over the next 12 to 24 months, what are the highlights that you would orient investors to really look forward to in terms of data readouts?

The combination data. So we will have combination data with Pumitamig mid to end of this year. And we will present this data on conferences. And this will be, of course, exciting because with Pumitamig, we have a next-generation immune modulator. We believe that some of our ADCs are also very much differentiated. So combining those and excelling to the next level of disease control is the most exciting thing.

I would add Gotistobart as well in the next 12 months.

It was an ADC question. Yes, but you are right. I'm also excited about Gotistobart.

It's a combination, that's why I responded.

So maybe to pull you in, Ramon. So the company has got a large cash balance. What are your plans for it?

Yes. So I think at least until we have more line of sight on where the actual revenues of TPAM, Gotistobart, Pumitamig and all of these combinations are coming, which they should be, of course, TPAM and Gotistobart a little bit earlier, but then with Pumitamig more towards the end of this decade, beginning of the next. I think the first and utmost priority would be to keep developing the pipeline, the internal pipeline that we have, grow organically. And then I would say the second priority is if there are any inorganic acquisitions or collaborations that we can do, that would be then the second place where we will be deploying these resources. And then as we get more certainty on how and when we are going to be launching and the size of opportunity, maybe we can start discussing other types of utilization of this cash.

I think we're just out of time. So I'll stop there, but thank you.

Thank you.

Thank you.

[Audio Gap] by Annemarie Hanekamp, our Chief Commercial Officer. So with that, I'm excited to hand over to Ugur and welcome him to the stage to talk to you about our unique approach to innovation.

Yes. Good morning, everyone. I would like to welcome you also on behalf of my colleagues here in the time square. What we would like to do is before going really into the details, I would like to discuss -- give you a status quo of the company and discuss the scientific concept, the medical concept and how we translate that and execute on that. And it will be a data, there's the presentation today, and you will have the opportunity at the end of the presentation to ask questions.

So this is just an updated company start. We are a multi-platform oncology company with more than 20 ongoing clinical Phase II and Phase III trials. We have an intact disease pipeline. And we have capabilities that we built around drug development and AI manufacturing and research.

After the COVID pandemic, in 2023, we pivoted back to oncology. And this is our original focus. And we are driven by the fact that oncology cancer is still 1 of the biggest challenges in health care. And the challenge is depicted here in a very simple manner. We are in principle, in drug development, dealing with 2 types of challenges. One is the inter-individual variability. Every patient is different. And the second challenge is the intra-tumoral heterogeneity, which we size often in patients responding for a certain time to the treatment, objective responses, sometimes complete responses that then the tumor relapse thereafter because of pre-existing heterogenic tumor cells.

The concept that we are addressing here is based on using -- exploiting the power of the immune system by a combination treatment, addressing complementary potentially synergistic mode of actions. So we have a class of molecules, immunomodulators for which we focus on critical IO pathways. And here, the idea is to target complementary pathways in cancer immunity cycle that promote durable antitumor effect. The second element is our targeted therapies, particularly ADCs.

Here, the goal is to have a precise and potent approach to kill tumor cells without harming healthy tissues. And the emerging emergence of third-generation ADCs allowed us here to make enormous progress, not only us, but also the field. And we started with the well-defined targets for targeting tumors. But in the meantime, we are extending beyond the current targets with our own approaches. And the last piece in this concept is mRNA cancer immunotherapies. And here, the concept is regardless what kind of treatment is done, residual tumor cells will stay. And even if the residual tumor cells are just a million of tumor cells, they will result in relapses. And the concept of personalized vaccines or cancer immunotherapies based on mRNA vaccines is to induce a poly-specific immune response that can deal with this heterogeneity and enable removal of the residual tumor cells.

Let's start with immunomodulators. So our focus is critical IO pathways. And today, I will really discuss about the PD-1, PD-L1 pathway and anti-CTLA-4 pathway. And the question is, of course, if you go with PD-L1 and anti-CTLA-4, where is the innovation? And I would like to show you how the engineering of the molecules can help to exploit an existing pathway and do something completely different or significantly better. So the 2 molecules that I would like to discuss today is our bispecific VEGF and Gotistobart, our anti-CTLA-4 molecule. So you are all aware about the bispecific PD-1, PD-L1 VEGF antibody class. And here, several candidates are existing and that combine VEGF neutralization with PD-L1 blockade. And these are actually well-characterized targets. So PD-1 pathway and VEGF.

So it's about the VEGF pathway is not only inhibition of neovasculature, but it is also an immune modulator. VEGF is an immunosuppressive molecule. And by targeting VEGF, we can reduce hypoxia, we can increase immune infiltration, dendritic cell maturation and Treg activity. And the activities of PD-1, PD-L1 are well known. So we -- this bispecific, of course, has these 2 functions. On the one side, neutralization of VEGF in a dose-dependent manner. And on the other side, checkpoint blockade, PD-L1 blockade and thereby accomplishing T cell activation. So the antibody Pumitamig for both arms has a very high affinity so that even low doses -- with low doses, we can accomplish complete inhibition. And this, of course, translates into immune activation. So we measure T cell expansion, cytokine secretion, tumor cell killing, activation. And in all functions, we see a robust activity of Pumitamig.

So the question here is, is Pumitamig just the combination of 2 antibodies. And our statement is very clear. Pumitamig is more than the sum of 2 monospecific antibodies. So we have, on the one side, a VEGF function -- anti-VEGF function. We have on the other side, an anti-PD-L1 function. And Pumitamig has certainly both. But it has, in addition, functions that are not mediated by the monospecific antibodies. This is the link function or the conditional function by combining the VEGF binding and the PD-L1 activity. So this is a slide showing -- we know that VEGF is in natural configuration a dimer. So that provides the opportunity to form immune complexes and immune complexes are dose dependent. If we have a high concentration of the antibody and low concentration of VEGF, we have antibody alone. If we have a ratio of 1:1, we have maximum complex formation. And if VEGF concentration increases unproportionally, we have more antibodies bound to VEGF, but forming less complex. And this is an assay actually measuring VEGF Pumitamig complexes.

So what happens if we form complexes? What we observed is the formation of a complex of pam with VEGF increases the on rate. You can see that on the left graph here, this is a faster on rate, and it reduces the off rate of the binding. And improved on rate and reduced off rate translates into a much higher binding and a lower dissociation constant. So we see here in the optimal concentration a more than 32-fold increase of binding affinity of Pumitamig to immobilize PD-L1. So this is one result of a bispecific antibody. Let's continue with that. The second is that this not only translates into a better binding, but a dramatically increased internalization of Pumitamig. So this is an experiment on a cell line, which expresses PD-L1. It's a tumor cell line with PD-L1 expression. This is a variant of our antibody in which we replaced the anti-VEGF-A arm with a control.

And here, the internalization of the antibody is visible by cells becoming red colored. And you see here even with longer incubation, we don't have internalization with the PD-L1 binding antibody alone. But if we take now the original molecule, then we see rapid internalization of PD-L1 and of the antibody within a few hours, and this is also depicted here. So we believe that the improved activity that we are seeing here for the antibody is mediated in part by the increased binding, the lower off rate, but more importantly, by the stronger internalization of PD-L1 mediated by VEGF Pumitamig complexes. So this is one differentiation factor. And this differentiates a bispecific antibody from the sum of 2 monospecific antibodies. But there's another differentiation factor. So anti-VEGF or VEGF and PD-L1 are both tumor antigens. They are increased in the tumor microenvironment. And this is staining with anti-VEGF-A of the tumor microenvironment. These are samples from non-small cell lung cancer, very strong VEGF staining in the tumor microenvironment, and this is PD-L1 staining.

So we are now targeting with our bispecific VEGF, PD-L1 antibody with both arms and the tumor microenvironment. And this means that we have the opportunity to target PD-1, PD-L1 to VEGF high tumors or you can see it also in a different way to target VEGF-A neutralization to PD-L1 high tumors. So if you sum that up in a simple table, then what we observe is here Pumitamig has following function, blocking of PD-1, PD-L1 signaling, neutralization of VEGF. And this is something that is accomplished by other bispecifics as well as the monospecifics or the sum of combination of anti-VEGF and anti-PD-L1. Then we have the cooperative effect of linking and VEGF binding, and this is not accomplished by combination therapy with monospecific antibodies, but it's also a feature of the other members of this bispecific class. And then we have the mechanism of tumor microenvironment targeting by anti-PD-L1, which is specific to the PD-L1 bispecific class.

So we believe that this is one of the key reasons why we observed with this antibody immune responses in tumors independent of PD-L1 expression. So we see response in PD-L1 high tumors, PD-L1 low tumors and even PD-L1 negative tumors. And clinically, this provides the opportunity that Pumitamig has the potential to become a backbone independent of PD-L1 expression and for the ability to introduce IO into different patient populations. And this is an observation that we made in TNBC, and we are making similar observations in other tumor types as well. Here is the example for EGFR mutant non-small cell lung cancer, second-line tumors. Here, again, strong objective responses in PD-L1 high, PD-L1 low and PD-L1 negative tumors. So this opens up the opportunity to bring this IO to use Pumitamig as an IO in the existing PD-1, PD-L1 positive space that means replace pembro, nivo and other existing molecules and then also expand beyond into indications in which IO is not yet approved.

Our clinical strategy for doing this is based on a 3-wave approach. So we already started the first wave. It's foundational registrational trials in 3 priority indications, Özlem, Ilhan and Michael will talk about this. The next wave is expanding this combination into multiple other cancer indications in combination with standard of care. And the last step is to combine Pumitamig with novel, novel combinations. And we will hear today about what kind of comb combination trials are currently executed. So this is a mode of action summary for Pumitamig. Next year, we will provide more details why we believe or how the action of Pumitamig is accomplished in PD-L1 negative tumors. So the presentation provided a mode of action, but not really an explanation why PD-L1 negative also works. This will be part of the presentation next year. So there's a motivation to come also next year.

So I would like now to talk about CTL4, CTLA-4. And this is also -- this was the first IO target and was introduced as checkpoint blockade. Cat is an antibody monospecific antibody with an optimized Fc that targets CTLA-4 the main function is not just checkpoint blockade, but depletion of tumor infiltrated regulatory T cells. It is currently evaluated in multiple cancer indications either alone or in combination with anti-PD-1. We have running Phase III clinical study, and I will provide information about what we are evaluating here. And we have also clinical studies running in indications that are usually not responding to checkpoint blockade. So anti-CTLA-4 is target for cancer immunotherapy. What is new, why Pumitamig is new.

So the prevailing view on anti-CTLA-4 is checkpoint blockade. So the idea is depicted on the left side, an antibody inhibits the interaction of CTL4 with its ligands and thereby enables that T cells can be activated. And the prevailing view is that this happens in the lymphoid tissues. And this activation of T cells, so that means inhibition of Treg activity results in an antitumor response. The problem with that is that this antitumor response is intrinsically linked to concomitant autoimmunity with a constrained therapeutic window. And there are dozens of clinical trials that try to identify some sort of a therapeutic window for ipilimumab by having different dose schemes by trying to identify better timing.

But more or less, all these clinical strategies ended up either by lowering the efficacy or by having the same efficacy but the same toxicity profile. The concept of an Gotistobart antibody that we partnered with OncoC4 is a different one. Here, the idea is not that much to block CTLA-4. It is more used to use an anti-CTLA-4 effect with effector mediated mechanism to deplete regulatory T cells and do it in a way that is tumor selective. So idea depletion of regulatory T cells in the tumor microenvironment and this provides an opportunity for better therapeutic index. And this was the mechanism was one of the key reasons why we partnered this antibody. And so how can -- how is this accomplished with Gotistobart? It's a little bit complicated and this slide has been presented many times, and I would really like to show you the mechanism, which is really interesting because the mechanism that the antibody wants to accomplish is via the improved Fc receptor function to induce lysis by ADCC and ADCP to remove Tregs in the tumor microenvironment.

The problem is that the antibody still binds and inhibits CT4, which would then be expected to induce autoimmunity. And this is solved for this antibody by PH sensitive binding. So that means when the antibody with the CTLA-4 complex internalizes, it is in the lysosome, the antibody -- the CTLA-4 molecule is -- the antibody stops binding and CTLA-4 can recirculate to the cell surface. And thereby, the antibody does not reduce the CTLA-4 expression on Tregs, but still accomplishes depletion. So the antibody is designed to allow regular recycling and enrichment of CTLA-4 thereby due to the Fc receptor function to selectively kill Tregs in the tumor microenvironment, improve the therapeutic index thereby. And by improving the therapeutic index, the argument is not it is safer, but it provides a therapeutic index so that it can be dosed higher.

So that means the side effects that we are observing are very similar to the CTLA-4 class, but the dose that we can provide to patients is much higher. And we can provide this dose in a prolonged fashion. So that's the way how this works. That means it allows prolonged and repeated dosing. So this is supported by preclinical studies in mice, where Gotistobart has improved therapeutic efficacy as compared to controls. And it also results in a higher, stronger activity of immune cells recognizing. These preclinical data. In the meantime, we have unpublished clinical data in which we clearly see that the antibody does what it should do in patients. So that means reducing the Treg density increasing CD8 T cells, reducing the amount of Tregs in the total population of CD3 T cells, reducing the amount of Tregs in the total CD4 population. And most importantly, and this is a factor really shown in many, many biomarker studies, increasing the CD8 to Treg ratio.

So we are running Phase III clinical trial in patients with squamous non-small cell lung cancer. The study started with an initial part where patients, both patients squamous and non-squamous non lung cancer were treated and the control is in second line squamous non-small cell lung cancer or non-small cell lung cancer as well docetaxel and we have now based on the data already started the Stage 2 of the pivotal trial, which is currently recruiting well. And what we will do is we will present data from this initial study, which is randomized in December this year in -- so which congress is that? ASCO, North American ASCO December 7, right? And we look forward, of course, in presenting the data.

So coming to the third part, mRNA cancer vaccines, I will keep it short because Ozlem is going to share more information on that. The mechanism of action that is desired is here to eliminate residual tumor cells. As you know, we have both mRNA technology toolbox and reported in the recent years a number of clinical trials, including antibody encoded monoclonal antibodies, bispecific antibody cytokines. But here, I would really like to focus more on the nanoparticulate mRNA vaccines. We have the mRNA LP technology, sorry for that for the mRNA-LPX technology. And here, we have the -- we can take several antigens, including shared tumor antigens and encapsulate that in lipoplex nanoparticles. And we have shown that this approach, particularly when injecting the nanoparticles intravenously, results in lymphoid targeting of the mRNA and expression of the mRNA in lymphoid tissues. And this is associated with 2 key mechanisms of action.

One is the release of cytokines type interferon the innate immune response. The innate immune response is dose dependent. So you see here a study where we treated an individual patient with increasing concentrations of the RNA lipoplex. So this is week 1, week 2, week 3. So this is weekly increasing dose. And what you can nicely see is that this is associated with pulsatile secretion of interferon alpha that is starting in the first 2 hours and going down within 24 hours. And this is pretty consistent, okay? So this is the innate immune response. And we, of course, know that innate immune responses can improve the outcome of cancer immunotherapies but this is a transient approach. And then we have a second function, which is the induction of T cell-specific immunity. And this does not happen in the first hours or first days, but this is something that for de novo immune responses is requiring a time frame of 6 weeks to get them to come to a maximum level. And then by repeated vaccination, the immune responses can be kept high.

So we is the full mode of action in our study in 2016, in which we really evaluated also what happens if we use RNA LPX with relevant antigen or with irrelevant antigens, you can download the paper and check for what kind of controls have been done. But the key aspects here is the mechanism is driven by TLR7. It's driven by type 1 interferon. And we have shown that this is something that can be universally applied to any type of antigen.

So, ADCs don't require much introduction. What I would like just to state is that we focus when in-licensing our first set of ADCs to ADCs with a newer linker technology, allowing us not only to get potent antitumor activity, but also bystander activity. And we, in parallel, started to work with our own targets, generate antibodies and generate ADCs against these targets and ensure that we use the ongoing innovation cycle because what we believe is ADCs are just emerging technology, and we will see a number of improvements in the coming years with regard to their potency, the selectivity and their immune effects. And we built an in-house ADC capabilities aiming to improve linker technology, but also aiming to come up with new mode of actions.

So this is the list of currently addressed targets. So we started with HER2. We have a B7-H3 ADC, TROP2 HER3. One of our new ADCs is CA99 special antibody, special antibody with an incredibly high affinity against CA99. And we also generated recently in-house an EGFR HER3 ADC with our own linker approach. So aiming here is combining target tumor cell killing by standard tumor cell killing, immunogenic cell death and the opportunity to synergize. So what kind of synergies are we addressing? So one obvious way to ask for synergy is combine Pumitamig with the different ADCs. And these are all preclinical studies, either in Xenograft models or in syngenic mice. In any model evaluated, we see a synergy that means the combination of the ADC with Pumitamig results either in a complete tumor rejection or improved tumor control over time.

So this is a slide that should come later, but it's okay. I will come back to that. Yes, this is the combination treatment with our ADCs. But as you know, we have also other immune modulators. Ongoing studies also address the synergy of our bispecific anti-4-1BB [ anti-CA ] antibody, which BNT314 or GEN1059, which is partnered with Genmab. Here, we see a clear increased activity of T cells, T cells improving beyond Pumitamig's activity with regard to cytokine secretion and tumor control. We have our bispecific ADC anti-EGFR HER3 topoisomerase inhibitor linker technology. Here as well, we see a very strong synergy of Pumitamig with our ADCs. And this was all supported last year already with preclinical studies.

Now we are in clinical evaluation of our next-gen IOs with ADCs, Pumitamig and ADCs targeting different entities. We combine Pumitamig with next-gen IO. These are IO-IO combinations, combinations in different indications. We combine our anti-CTLA-4 and our bispecific also with mRNA vaccines. These are running clinical trials. And last not least, we do also combination trials where we combine mRNA with ADCs. And these are all addressed by the concept of identifying different types of synergies of these different treatment approaches. So the vision remains the same. This is aiming to bring -- to combine our capabilities on IO, ADCs and mRNA therapeutics for personalized treatment approach and ultimately, not only address off-the-shelf drugs, but also proceed to the personalized space with tailored on-demand immunotherapies. We are combining that with capabilities in automated in-house manufacturing and our AI capabilities to support drug design, neoantigen prediction as well as manufacturing.

So 2026, the key areas of focus is to use the combination therapy momentum. This will be a year of combination therapies where we will generate data and share end of this year, data from multiple indications and Ilhan and Michael will show you the ongoing clinical trials. We want to proceed and we started to do that from just modalities to really disease areas. That means the optimization approach is now ongoing for specific type of tumors, and Ozlem will provide and share the concept how we are doing this to get really the best type of the combination therapy in disease areas. And of course, 2026 is also a year of late-stage acceleration, aiming to start a number of additional registrational trials.

So I would like to ask Ozlem to come here and continue. Thank you.

Also a heartfelt welcome from me. It's a pleasure to see so many new faces and also friends who have been accompanying us for quite a while. So Ugur has shared with you our vision and he has shared with you our science, which typically defines the boundaries of the space within which a company can generate innovation and value. And what I will do in my part is that I want to describe our strategy from getting from one to the other from what we have in terms of value foundation to our vision, and this is about our clinical development strategy for our pipeline. You all know that in -- over the past decade, the paradigm, the treatment paradigm for many tumors has shifted towards combining 2 modalities. IOs immunomodulatory molecules with highly cytotoxic agents, meaning chemotherapies. And this approach without doubt, has brought progress.

However, it has also shown that it is still limited by durability of response -- by short durability of responses by limitations in depth and resistance mechanisms. The next paradigm is already unfolding in front of our eyes. And this next paradigm is based on improving both components of this dual. So what we see happening is that next-generation IOs are developed and compounds with higher cytotoxicity with higher precision are being developed. Our deliberate strategic decision here is that we want to do both. with our pipeline and our multimodal, our multi-asset pipeline allows to do exactly that. This is also the reason why we have immune modulators, and you can also count actually mRNA vaccines or immune therapies as immune modulators on the one side. And on the other side, have brought in into our pipeline over the last 2 years, ADCs as highly precise and potent cytotoxic agents.

The IO we have chosen for positioning as a pan-tumor backbone is Pumitamig. And our pipeline and our strategy is anchored in this molecule. As Ugur has already pointed out, it is also a design choice for our strategy that our ambition is to develop Pumitamig broadly, deeply and fast, meaning across indications, disease areas and across lines. We do this in 3-waves, which are shown sequentially here. However, in principle, are happening in parallel, and I will talk about the way we execute in a couple of minutes. One part of this 3-wave strategy is that because speed is important for us, we are developing -- we are conducting foundational registration trials in combination with standard of care chemotherapy, which is well known to clinicians and well established in 3 indications in non-small cell lung cancer, small cell lung cancer and TNBC. So indications with large market size. You also have heard in our earnings call that we are adding 2 more indications, first-line CRC and first-line gastric cancer, and we will talk more about this in a couple of minutes.

Then the second in parallel conducted way is that we are extending into additional indications, which are listed here and with registrational intent, but also with a large number of signal-seeking trials across indications to ensure that our next decisions towards registrational intention and our selection of indications is well informed. And the third line and the third wave and Ugur already talked about that, is very close to our heart, namely novel, novel combinations in sort of first initial wave with our in-house ADCs, but also beyond. And you already see this happening that we are also starting to combine with our IOs and Ugur has shown some of the preclinical data for this. So this is the Pumitamig part of our innovation strategy. Then we have a second part in this door. As I said, we also want to improve and optimize the cytotoxic component in the IO cytotoxic agent dual and for this, we have built an entire panel, an entire portfolio of ADCs. These are all topoisomerase inhibitors with specific linkers, which also deliver bystander activity.

We have diverse targets, which we are addressing here. And the reason is very clear. There is not a situation that you can address all biologies, all disease areas, all indications with one single target. And what we want to make sure is that we select for each indication the right target and the right ADC. And these targets which we have selected cumulatively, as you can see on the bottom of this slide, cover with their expression and also with their biology, all these multiple indications we are interested in. As you can see, we have, on the one hand, targets like HER2, which means we have an ADC here, which addresses an already very well entrenched target defined market, right? We also have with TROP2 an ADC where -- which I would say is more semi entrenched. So there are already approved TROP-2s, but the market is not as well defined. And then we have newcomers, B7-H3, HER3, where there are no approved ADCs around yet. And the way we have to develop them forward in order to make them fit for purpose for our novel, novel combinations will also differ a bit with this entrenchment, which is given.

So what about the execution? And the last 24 months were actually only about execution. And I would like to summarize on the next slides what we have done and what we have achieved. This shows you the studies and the settings in which we have tested Pumitamig as a single agent and Pumitamig in combination with standard of care chemo. This is data which is worth more than 1,400 patients and was essential for us in order to define across indications, the safety profile to inform dose decisions and to understand with our signal seekers, the mono and also chemo combo activity in these different indications. And as you can see here, we -- and this is also a design choice. We are interested in all these disease areas, lung cancer, breast cancer, women's cancers, GI and GU cancers and have created data in all these indications. This data set has led to decisions to activate our 3 Rosetta trials in non-small cell lung cancer and in TNBC and recently in the next wave with first-line CRC and gastric cancer.

This is the execution which happened for our ADCs, where we tested single-agent activity of all our ADCs, again, to understand the safety profile to inform dosing and to prepare for novel, novel combinations. Again, across indications. And as you can see here, these are massive data sets, which really allowed us to understand which indications are the best, which settings are the best for each of these ADCs, targeted ADCs. Just for clarification, the primary aim is not to develop as monotherapies, these ADCs as monotherapies or in combination with standard of care, but to combine them as early novel, novel registrational pathways with Pumita. However, that does not mean that we are not pragmatic. Should we see in any of these indications compelling single-agent activity. And should it make sense due to commercial opportunity and other factors to proceed in this mono setting, we would do that. And admittedly, there are a couple of signals which make us think. Then we have already expanded based on this mono data or single-agent data from our ADCs in combinations with Pumitamig. And you can see here all the indications we have selected for individual ADCs. This was again informed by the single-agent activity. In some of these indications, as you can see, for example, non-small cell lung cancer, AGA-negative, we have still 3 of our ADCs in the game. And these early novel novel combo studies will define which one of those, for example, in this indication to proceed with. So this is the execution, which is going on.

As Ugur has already said, we -- this year will also -- or next year will also be about moving modalities, the modality-centric view into the disease area view, and this also starts to unfold. And here, we are benefiting from the fact that with this ADC pipeline, with this ADC panel and with Pumitamig as a pipeline and a product opportunity, we have created optionality. And this, again, allows us to think broadly of those disease indications we are interested in. So this, for example, means we are not in the situation that we have only one ADC, and we have to drag it across all the lines in a certain indication. We can ask a question which ADCs, which targets make sense for this disease and also diversify across settings lines and, for example, subsegments of that disease. We are pretty strong in lung cancer, and you can see the individual assets we are exploring/developing in lung cancer. I will not go too deep into our disease area canvases because you will hear on a very granular level from Michael and Ilhan about each of these boxes, which are shown here.

So for lung cancer, you can see that we are active in all of these segments, which are depicted here. We have registrational trials in first-line non-small cell cancer without AGA very broadly. So histology-wise, all comers and also all PD-L1 scores, which we are addressing here with our Rosetta Lung-02 trial, Pumita plus chemo. In small cell lung cancer, Rosetta Lung-01 combines Pumita with chemo as well. Both Phase III trials are recruiting patients. However, we already prepare for the next step. We have tested our ADCs here in non-small cancer, all three B7-H3, TROP2 and HER3 ADC in second-line plus settings, which, as you know, is the sandbox to address and understand single-agent activity. And that already has informed us to and motivated us to start proof-of-concept trials of Pumita with each of these 3 ADCs because all of them look compelling in the first-line metastatic setting to inform the next step. What I also would like to highlight is Gotistobart, Ugur has already mentioned that positioned here in the second-line plus metastatic setting of non-small cell lung cancer. This is a very dire post-PD-1 heavily treated patients setting. And typically in this setting, patients have resistance mechanisms such as high Treg levels in the tumor microenvironment, MHC loss. And this is a perfect place to position a Treg killer, Gotistobart in this case.

We are also building our canvas for breast cancer, and these are the assets we are testing here, Pumitamig plus chemo in our global ROSETTA Breast-01 trial is addressing the segment of patients who are not served with first-generation PD-1, PD-L1 compounds, namely CPS lower than 10. We have an ongoing China Phase III, which looks into all comers in TNBC. And as you can also see, in the second-line metastatic setting, we explore Pumita again, with different ADCs in the HER2 low -- HER2 expresser segments together with our new HER2 ADC [ TPAM ]. And also this will be depicted in more detail by Ilhan and Michael. And this is our GI cancer pipeline, which we are also building with multiple assets, which are positioned here. You will hear about our ROSETTA trials a plus standard of care in CRC first line and in gastric first line. But also here, we are already combining with ongoing trials Pumita with ADCs or other IOs. And in the adjuvant setting, we have positioned in CRC and PDAC our iNeST, our individualized vaccine, autogene cevumeran with ongoing Phase II potentially registrational trials.

So with this, let me quickly wrap up our strategy. We think that we have a number of differentiation factors, which are the tailwind to the strategy, which I have described. One is the molecule itself and Ugur has pointed out the differentiating factors on the molecular level, having the PD-L1 versus the PD-1 component. The other is the convergent and deliberate development of both Pumita as well as our ADC portfolio as a platform. The third is our clear pan-tumor aspiration. And we think that with the tailwind of a strong cash position and of having a partner who shares our vision for Pumita, we are set up for success here. And with this, I would hand over to Michael and Ilhan for more details on the studies.

Good morning, good afternoon, good evening. My name is Ilhan Celik. And together with my colleague, Michael Wenger, we will guide you through this section and going into more details about our pipeline and our indications. As already introduced by Ugur and Özlem, in an overarching approach, we want to share with you where are we at the moment in conducting these trials, results which we can share with you and giving you a perspective what is next coming. We start with thoracic cancer. And as you already have seen on this slide, we will talk about some of the modalities in thoracic cancer, starting with our immune modulators, but also -- sorry, not only with Pumitamig but also with Gotistobart.

So we have an ambition to establish a broad presence in lung cancer by introducing multiple new standard of care changing regimens across both NSCLC and small cell lung cancer stages and subtypes. We will begin to build that through the ongoing pivotal trials listed here focused on introducing new checkpoint inhibitors either as a monotherapy or in combination with chemotherapy. These trials will form a basis for our novel combinations in lung cancer that we think will position us well against other competitive agents being developed in lung cancer. Zooming now into non-small cell lung cancer. So this is a tumor type with significant incidence and where the majority of patients are diagnosed with already late-stage disease, leading to poor long-term survival despite the treatment advances of checkpoint inhibitors.

So this slide summarizes all the data we have generated in non-small cell lung cancer so far as proof-of-concept trials in different indications and lines of treatments in AGA positive and negative indications and also looking into the perspective for future combinations with chemotherapy and ADC. And all these data were already presented at ASCO 2024 or ESMO 2024. Of note, on the right-hand side, you can see some data in EGFR mutant second-line NLCLC, which I want to point out specifically, which is the efficacy you can observe here independent of PD-L1 expression level. So even in the PD-L1 negatives, you can see encouraging data. And overall, this setting is known to be not sensitive to IO checkpoint inhibitor treatment.

This is our ongoing Phase II/III trial for Pumitamig plus chemotherapy in first-line NLCLC AGA negative. We are testing this against the current standard of care, pembrolizumab plus chemotherapy. Since the beginning, and this is really something which I want to emphasize clearly, we have 2 independently powered substudies for squamous and non-squamous histologies. The Phase II dose optimization portion of this study has fully enrolled, and we are recruiting the pivotal Phase III portion now. We expect it to publish the Phase II data next year.

With that, I hand over to Michael.

Yes. Hello. Good morning. Quick refresher on squamous. You heard squamous now a few times. What is squamous versus the other non-small cell lung cancers. Most of the others are adenocarcinomas. Squamous make up roughly 30%. In some areas, it's actually a little more. Squamous cell carcinoma usually develops due to some exogen NOxes such as smoking or asbestos or things like this. It's a pretty nasty tumor. What it means is that not many patients make it out to second, third, fourth-line treatment. Actually, most of the studies stop at second line, and that's also the case with our trial that most of the data will be in second line. It's also a high medical need tumor, right? So there's the standard therapies, IO therapies. And it has been named the #1 medical need for lung cancer primarily as the existing treatments are so bad.

Most patients, as said, get either pembro or ipi/nivo in frontline, but for second line, there's really, really not much. Now you heard a lot already about Gotistobart, so I keep this fairly short now, mode of action, et cetera. But important to understand is that from our Phase I, we actually saw a nice signal that was conducted mostly by OncoC4 alone and the later stages with a little bit help from our side. And again, for squamous, it shares some similarities with small cell lung cancer in that you do see some responses, but then patients relapse fairly early on. And the signal we saw here in this Phase I was quite substantial prolongation of the responses with monotherapy, right, which is from a CTLA-4, which is not the most fancy drug, as you can imagine, that has been used in this setting.

Now you also saw this slide already from Ugur on the Phase III, and I'll get tiny a bit more into detail here what you're going to see also at the North American ASCO in case you're interested in a few days from now. The study had 2 doses were tested. That's what you see on the left-hand side. And this is also the data that we will be showing in a few days. The results from this also prompted the continuation of the study, which is on the right-hand side. And you heard Ugur saying that this study is recruiting. It's actually fairly advanced in recruiting. So we actually will hope to see some results early to mid of next year internally, and then we'll see what we'll do with this.

Importantly, again, it's one of these rare cases where Phase II data is almost done in -- almost identical setting as the following Phase III. You probably are all aware that the main reason why Phase III fail for positive Phase IIs is that something changed between the Phase II and the Phase III. Usually, the indication was broadened or maybe a standard of care changed or prior treatment changed or stuff like this. This is done with the same size, the Stage 1, almost at the same time as Stage 2, right? So we think we pretty much derisked that Stage 2.

Again, data will be shown in a few days from now. We're quite excited about this. And yes, we can't show you a glimpse right now, but you hear from this. With this, over to you.

Thank you, Michael. So we want to close this cycle for lung cancer and looking into another important indication of high unmet medical need, which is small cell lung cancer. So small cell lung cancer remains a challenging immunological cold disease in which responses to immune checkpoint therapy tend to be very short. Nearly 2/3 of patients who are diagnosed with small cell lung cancer have extensive stage disease, which means surgery, for instance, is not possible any longer. As you can see, these patients face much worse prognosis than those with limited stage disease.

Based on data from a single-arm trial Phase II in China, we started a global Phase II dose optimization trial last year to help identify the right dose for the pivotal development. Patients were randomized to receive either 30 milligram per kilogram or 20 milligram per kilogram as indicated on the slide, Pumitamig every 3 weeks in combination with standard of care chemotherapy. So this was a fast way to justify the optimal dose for the pivotal already ongoing trial in first-line and in second-line small cell lung cancer.

The data from the global Phase II trial were presented recently at the World Lung Conference some weeks ago, confirmed that the data which we have generated earlier from the China trial are also valid in a global population and helped us really to pick the dose for the now ongoing and enrolling global Phase III ROSETTA Lung-01 trial. You can see the waterfall plot and overall response rates for both doses on the right-hand side of the slide and an overview of data we have generated across our studies in small cell lung cancer on the left-hand side. And again, the confirmation of the data generated in China now also in the global patient population is very encouraging.

In addition to the waterfall plot and the overall response data, I want to share with you also the data regarding duration of response and PFS. Both indicates also here, we have encouraging data with a median duration of response of 4.9 months and a median PFS of 6.8 months. We look forward to continuing observing this data as it matures and to get more insight and additional insights of the potential of Pumitamig in first-line small cell lung cancer. So this is the study design of our ongoing enrolling Phase II/III extensive stage small cell lung cancer trial. And as already mentioned, the Part 1 of this study, the dose optimization part is completed, data were presented and the Stage 2 part is ongoing. So as you can see here, we are randomizing in 2 arms against standard of care, atezo plus chemotherapy. And in the treatment arm, we have Pumitamig 1,500 mg flat Q3W plus chemotherapy.

With that, I want to hand over again to Michael for the next part, sharing with you some data which we are generating in proof-of-concept trials in combination.

Okay. Strategy again. Strategy is rather simple, right? You heard Ugur talking about it earlier. For lung cancer, we're doing large head-to-head studies against checkpoints, right? Backbone of chemotherapy. That's the current wave. Now what we're doing with the ADCs is basically preparing for the next wave. We add the ADCs to Pumitamig, assuming, of course, Pumitamig is going to make it, and we'll replace then the chemotherapy, hopefully, right, with these studies. So basically, the first wave is replacing the checkpoint. The second wave, if successful, would be replacing the chemotherapy. And ADCs are already replacing chemotherapy with checkpoints, right? And we're just coming from the other side. So it actually makes a lot of sense to prepare for this.

And in the next couple of slides, I'll show you where we are with this and which agents we favor right now from the ADC portfolio. So the first one is B7-H3, right? Interesting target -- why is it interesting? Well, it's so broadly expressed on so many tumor types. You see them listed down there on the left. And I would say 60%, 70% or so match with the expression of PD-L1 and VEGF. So this makes an ideal combination partner for a drug that is so broadly expressed. Now why was it so late in drug development? You're aware there's not so many B7-H3s out there. That's a relatively simple explanation. First, the ligand for B7-H3 wasn't really known. It's still not really known. There's some hypothesis what the ligand is.

And then in the early experiments, when B7-H3 was targeted, it had both lymphoproliferative and lymphodepleting effects, right? So it was a bit confusing. But we're over this right now. So it's now an established target, again, for many, many of these tumor types. And we have, specifically for lung cancer, a nice program, both for non-small cell and for small cell. But I don't want to miss out on showing you all these other indications that we have Phase I studies planned or ongoing as a monotherapy. And again, to emphasize what Ozlem said, we're not necessarily planning for a massive single-agent program for approvals, right? There may be pockets here that are quite interesting, but typically, we think this is also a combination play. Notable exception is CRPC, and I'll come to this after the break because prostate is a disease that is not -- that wouldn't make a lot of sense to use a checkpoint inhibitor.

Anyways, we've done together with our partner, DualityBio, a Phase I, typical Phase I with extension cohorts and already found a broad therapeutic index for this drug. The data for squamous and non-squamous looks encouraging, but perhaps not super outstanding for many of the other agents that are in this field, such as TROP2 or others, right? It's clearly competitive, but it's not like massive in that sense that you would jump to a Phase III alone with this drug alone. And also notable, squamous is always a little worse than non-squamous for the reasons alluded to earlier.

Now it gets more interesting with small cell. Again, what alluded to earlier, you can throw at small cell lung cancer pretty much everything and get a response. But the responses are usually very short-lived. And when we see something that has responses of 3 months, 4 months, 5 months or even longer, then it becomes interesting, and that's exactly what you see here on this slide. What you also see is that between 6 and 9 mg per kg, where we had these expansion cohorts running, there's not a ton of differences from the efficacy side of things. So it wouldn't make a lot of sense to go further. And one of the clearly differentiating factors for this particular B7-H3 is that it's very safe. Why is that? Well, you see the numbers of grade 1s and 2s here in the light colors. This is for 6 mg per kg. This dose will go forward in most or perhaps all indications. This is a dose of 9 mg per kg. You see a little higher hematologic toxicity, but generally, it's really well managed. And so again, an ideal combination partners for doublets, but we also dream already of triplets that could also be done in our pipeline with a molecule like this.

Now where are we with the monotherapy with a single agent. We've done a lot of experiments. You just saw some of the small cell and non-small cell data. We have in several other indications also data, of course. But this is now the combination Pumitamig and while it may look a little complicated, you'll see similar schemas for many of the other combinations with Pumitamig. Need to remember, this is novel, novel combination, so it's not with chemotherapy. So there's always a question of is it safe and what dose do you use? And you can imagine that the agency, the FDA specifically is asking us to also run randomized comparison of different doses relative to Project Optimus.

So this is what we're doing here, both in the AGA-negative frontline and second-line small cell. And then we have several cohorts for pretty much addressing all the relevant indications in lung cancer that can be targeted with this agent. So again, from our point of view, the play in lung cancer for this molecule is not necessarily the monotherapy, even though the data that you've seen are competitive. But really to make a difference, we want to combine it with Pumitamig. We're finishing up the Part 1 right now. So we'll move to the second part very soon. And then it's not a static like you need to complete this step to start the next step. We have at any point in time, also the ability to talk with the agency and go to a Phase III right away. Yes, I think that's for the ADCs in lung.

Now also talked about the mRNA vaccines. You are aware that -- I'm sorry, this is -- no, this is the other one still, 326, HER3. HER3 ADC also falls in the bucket of -- an ADCs, there are some, but not so many like for TROP2. Again, why is this same or similar idea. HER3 targeting isn't so super easy. The initial HER3 antibodies didn't work so well also because HER3 only works through dimerization, right? It only works if it's bundled with another member of the HER family, either EGFR or HER2 or something else. So creating an antibody not super easy. What we did for this beautiful molecule in-licensed from Medilink DAR8 TOPO1 toxin, so also very modern is just a standard Phase I. These are the results for patients with advanced disease, most of these bars here in lung cancer, you see it down here. Some are in breast cancer, and I'll come back to the same slide when we talk breast cancer after the break. But what you also see here, we went up to dose of 5.5, but we don't have to because we have quite nice activity around 2 and 3 mg per kg.

Main tox of this molecule is also hematologic, specifically neutropenia. And so we need to pay attention when giving this molecule. But I think we learned how to do this now also over the course of last year with the help of GCSF, if required, for providing a safe dosing of this molecule. For lung cancer, specifically, we are having expansion cohort in second and third line EGFR mutated and in squamous and non-squamous NSCLC, we go through the dose levels just as you would in any normal dose-finding experiments.

I'll stay on this slide just a moment. Again, it looks complicated, but it really isn't. It's quite logical. We'll start here with the combination. Remember, same as before, HER3 ADC alone probably would find [indiscernible] for a small indication or so or where checkpoints really don't play a role, but we are in here for the combo with Pumitamig. And so we'll start the combo studies right now. So far, so good, a couple of patients treated, no particular safety issues seen. And from this, we'll move on to the dose expansion and later dose optimization and contribution of component, right? All the things that the agency typically wants us to show, and this is designed in a way that we would both have the individual contribution component solved as well as contemporary control, which is standard of care, which would give us an idea of the efficacy relative to the control and potentially could be even expanded to a larger study with registrational intent. We're still here, right?

So next year, we will basically move this program to here and potentially to here and hopefully be able to show then that the drug together with Pumitamig can be standard of care. This runs in parallel to the 324 efforts, as said, at any moment in time, we can pick the winner from one of these programs in case we choose to and in case we feel comfortable, something is outstanding. Yes, mRNA, last point here on the lung cancer [ 224 ]. You heard Ugur and Ozlem talking about our mRNA vaccine, BNT116. Just as a recap because we talked about this a few times before. The idea of the FixVac is it's off the shelf. It's readily available, can be used to get chemotherapy or as its own drug. It doesn't require a person's RNA or DNA to be processed.

And interestingly, also the agencies were happy with this approach thus far when we run these studies, right? We are able to demonstrate with preclinical not clinical [ that will actually do don't have ] patient before studies. We have one big study ongoing. It's called LuCa-MERIT. We constantly meet this study cohorts of 20 or 30 patients. In gray, you see the part that is complete and also that we have reported to you with our friends from Regeneron. We started these experiments.

Some of these cohorts have cemiplimab as a second drug. We're currently repeating some of the more interesting data that we found with Pumitamig. Same idea, right? Like we think Pumitamig is probably better in this setting than also the checkpoint inhibitors. But we want to be sure. And so we repeat some of these experiments even though the experiments with cemiplimab were quite nice. I'll show you some of the data we had. This was the very first entry into human cohort on the very left-hand side that with monotherapy vaccine and if patients were interested and fit enough, they cemiplimab was added. You again see this pattern. You see a response and then it becomes durable, right? This is what you want to see. Like obviously, you want to see a lot more patients responding. But if you can keep patients in response, that basically shows you the vaccine works.

Now there was a lot of skepticism that you can compare the vaccine with blunt chemo and the experiment with docetaxel changed this for us and for perhaps also the scientific community that it's not that the chemotherapy affects the -- not at all the generating T cells, but not to a way that would be detrimental to the success of the experiment. So same thing, we saw some nice efficacy later where cemiplimab were given upfront together and the 2 most recent published data we have were in frail patients. Remember, about 10% or so of lung cancer patients either don't want to receive any type of chemotherapy or are simply not fit enough for that. It's a tough regulatory pathway, but it's also a pathway that is wide open. There's a few approvals atezo in Europe, for example. But otherwise, it's pretty open. And then there's earlier stages, Stage III. So after resection and radiotherapy or without the possibility of getting resection, but after chemo or radiotherapy, we also saw the vaccine showing some nice results. Anyway, this concludes the vaccine section, and over to you.

Okay. So want to summarize this section regarding thoracic cancer with some take-home messages regarding our development in non-small cell lung cancer, small cell lung cancer with the different modalities. So overall, we want to execute the initial pivotal trials to establish the presence in multiple thoracic cancers and tumor types and treatment settings to enable future combination studies. So this is our aim. And with that, we are generating now a break for you after all the science information, a lot to digest. And after the break, be curious and stay with us for the further indications, and we will move on with breast cancer.

After 15 minutes break. Thank you.

[Break]

To our session as mentioned before the break. So we will move on with the breast cancer indication and share with you the different modalities, which we are developing in the breast cancer indication. And so this is, again, our diverse breast cancer pipeline. And as you can see, we will talk about the immune modulators, Pumitamig and some of our targeted therapies.

So this slide was already presented by Ozlem in the beginning. And again, a short recap, indicating that we are developing Pumitamig in TNBC first-line metastatic with 2 trials. One is a trial in China in all-comers in combination with chemo. The other one is the combination with chemo in CPS less than 10. By definition, this is the PDL 1 negative population with the highest unmet medical need. And in addition, we are developing in hormone receptor positive HER2 low. And regarding this, Michael will share with you some data.

So looking into the TNBC indication, these patients are facing outcomes due to limited therapeutic options. As you know, the majority of the patient population is in Stage III and IV and the treatment option depending on the PD-L1 expression level is either chemotherapy or above 10 in combination with pembro. Still, the treatment outcomes vary and they are limited and so the requirement for additional and innovative treatments is high. This slide, again, summarizes the data which we have generated in TNBC with Pumitamig in combination with chemotherapy. These are data from our proof-of-concept China trial. And as you can see here, this is already presented several times. And last time you have seen the data, which indicates, again, as already mentioned for the EGFR mutant second-line setting that Pumitamig in combination with chemo is effective independent of the PD-L1 level. And this is nicely summarized on the table on the left-hand side. And you can see different PD-L1 levels in the ITT population, an encouraging median PFS of 13.5 months. On the right-hand side, you can see here an interim overall survival. And for the different PD-L1 levels and the ITT population, you can see that we are ending up in a range of 20 to 24 months.

Important to note is that compared to the benchmark data in chemo in the below 10, you can see that we are really meaningfully above what we see with median overall survival 15.2 months and median PFS of 5.6 months. So based on the data from China, which I shared with you before, we have initiated a dose-finding study also for this indication. So this is a study which we have started last year in summer to investigate different doses of Pumitamig in combination with different chemotherapies. And the Cohort 1 is going to be presented data out of this Cohort 1 will be presented in a few weeks at the San Antonio Breast Cancer Conference. We are investigating also other chemo combination. Treatment choice, as you know, in TNBC is different depending on the region and the preferences. So this is our study design from the combination trial in TNBC, first-line TNBC PD-L1 negative. And so this study is open and well on track, recruiting patients within 2025.

With that, I hand over again to Michael to give you more details about our combo.

This clock says 0. I don't think I have any time left, but maybe that can be changed. Anyways, so for breast, our main play, at least in the next few months and perhaps a year is trastuzumab pamutecan or short TPA, perhaps also known by some of you under its old code BNT323. This was a molecule that was the first of the ADCs that we've in-licensed from our partner, DualityBio in China and was back then also the most advanced. So what Duality did with a typical Phase I, Phase I/II study exploring different dose levels. And you see here also 2 cohorts in HER2 low and in HER2-positive disease in breast cancer. And this is the outcome of this Phase I data, a fairly large sample for Phase I, different doses. And there's perhaps 2 things to take note or to take away from this.

One, the higher doses yield better responses unsurprisingly. And number two, we also saw some really nice efficacy at the lower expression levels at HER1 at [ IgG 1+ ] specifically. We went up to 10 mg per kg, then went down to 8 and 6 and had a discussion or I should say, Duality had discussion with the agency and 8 mgs were taken forward to this Phase III study. Now before I go through where we are with the Phase III study for TPAM, just to note that this molecule is pretty much derisked in a way that, number one, it's -- some of the studies have been modeled after existing studies from other companies. But more importantly, if you've heard the press release from Duality recently, the HER2-positive study was positive, right? So they declared the study that was conducted in China, which was a randomized Phase III study just in China was positive in HER2-positive disease for this drug.

Now why is HER2 positive, not our Phase III study, very simple, and you probably all know this, T-DXd is approved in most countries, so we couldn't do a study against chemo as Duality did in China. But again, point being that our drug versus chemotherapy in HER2-positive was positive. Data have not been released or published yet, but we know they're positive also from Duality's press release. And in parallel to that study, a global study was started. This is this study in HER2 low. And I guess, in FDA speak, this study is well underway, but we will need to wait for results perhaps a few more months to know what it would like. It's a very simple design and with a PFS endpoint.

Now coming to the assets that are a little earlier. So same theme, right? Like we're doing for Pumitamig plus chemo Phase III trials, and now we're exploring in parallel Pumitamig plus ADCs. And for doing this, we first need to know what the single-agent activity of the ADCs. Is it even sensible to do combination with Pumitamig? And if so, then do the combination rather quickly. This is basically what this slide shows. So we'll start, as you would for all novel in later-line setting. And we do this with TPAM. So again, TPAM plus Pumitamig in a variety of starting doses here. But as I said, for BNT323 or TPAM, we know a lot about the dose. And we don't think we'll run into too many challenges with the combination of Pumitamig again for the reason that one, Pumitamig is a fairly safe drug, more than 1,200 patients are now treated. So we know the safety profile very well. And number two, there's also a ton of combination data already. And so we still need to do this, but we do this fast, and we move quickly also then to the expansion cohorts, which would then address the same segments that you know from other breast cancer studies.

So TNBC, but we will also look, of course, with TPAM targeting HER2 and everything between HER2 ultra-low to HER2 low to HER2 positive in an effort to really think if data are outstanding relative to existing data that we would be able to replace existing HER2 ADCs, right? So this is the goal of this. And again, similar to what you saw for HER2 -- sorry, for HER3 and for B7-H3, we're at this part dose escalation right now, but this one should be open kind of summer-ish or so, so that we will be able to get into tangible data sets for efficacy and safety.

Yes. With this, going on to BNT326. This is our -- again, our HER3, you heard about it already in the setting of lung cancer. HER3, by the way, is not as broadly expressed as B7-H3, but it's expressed in the biggest tumor types that are out there, right, in -- specifically in breast cancer and in lung cancer. So where HER2 targeting ADC wouldn't work TNBC because TNBC is obviously then HER2 negative. HER3 ADC could very well work. So that is what we're doing right here. We're having cohorts, again, as monotherapy in a variety of indication and then move fast to the combination, and we started the combination now with HER3 and Pumitamig. Early days, right? I alluded to this in the lung cancer part. But again, in about 6 to 9 months, we hope we have something that can guide the way whether we go into several or some of these breast cancer indication. And the data here from the combination will help us to show that. With this...

So also wrapping up this section for breast cancer. So again, the take-home message for you is we want to execute our initial pivotal trials to establish a presence in multiple breast cancer subtypes and treatment settings. So grow presence through novel combination. This is our aim. And so this is a theme overarchingly in all indications with all the ongoing trials we are following. So it's a central piece of our strategy. And with that, I want to move on into the gynecological cancer types, handing over to Michael again.

I hope you don't get too nervous about this swapping all the time. But yes, bear with us. Okay. Gyne cancers. One of the features is that checkpoints play at least questionable role, but are definitely not engraved that much in gynecological cancers as they are in several of the other tumor types, right?

You might have heard the results of the B96 study in ovarian recently, which was after, I think, 11 failed studies in ovarian cancer for various applications of checkpoints and various checkpoints was the first one that was positive. And now the community is debating whether that's actually real or not and whether that will be a major success or not. I don't want to go into this. But for us, we're following a very simple strategy, right? Like we basically have our ADCs that will work most likely on their own. I'll show you some data in a second in ovarian, cervical and also endometrial. But if we can, and we are very ready to take a little bit of a risk here, we'll combine them with Pumitamig. Same idea. Pumitamig may -- is not your run-of-the-mill checkpoint, right? It has also the VEGF effect. And we know already from the early days of Pumitamig that it works also in PD-L1 negative patients.

All right. So several combination studies in several disease indications. You see on the left, cervical platinum-resistant ovarian and endometrial cancer. Starting perhaps with the most obvious one, TPAM or BNT323 in endometrial cancer. This slide we've been showing now for 2-plus years because it's the only published data we have on BNT323. Rest assured, we're going to change this. We'll have data presented at a conference in half 1 or I should say, maybe even Q1 next year on a much larger sample than these handful of patients here. This was, again, the drug that we in-licensed from Duality as a first. And back then, the idea was there's nothing out there for endometrial cancer that is HER2 positive. And we wanted to jump on that idea and generate some single-agent data. And once we did that and also had this much larger sample size, we also embarked on a Phase III study.

So just before I show you the Phase III design, we occasionally get the question, what's happening with your HER2 ADC? Is it still alive? Well, it's very alive and kicking. We have one positive Phase II. This is the China study, right, that I alluded to earlier. We have one Phase III study that is nearing completion on HER2 low. And we have this study that we started fairly recently in endometrial cancer, right? So this is a third Phase III study we're doing for this [ very efficacy therapy in cancer, all these patients need to have frontline therapy, cemiplimab were approved 1 or 2 years ago in this setting ]. And so when discussing the design with several agencies, we came to the conclusion that it would be best to single out this particular group of patients that have received frontline treatment with a checkpoint inhibitor because once the study reads out, it needs to be relevant for the -- specifically the U.S. population. And we always see a very high uptake of first line both in the MMR high and MMR low patients. So all of these patients will be treated after failure of the first-line checkpoint plus chemotherapy treatment.

We hope to be able to report on these results a couple of years down the road from now. As said, the study has just started. And part of a potential accelerated approval, as you all are aware of, for the single-agent study that I just alluded to earlier is that the confirmatory trial needs to be well underway or even better fully enrolled or near fully enrolled. And so that's why we are currently making a lot of efforts [indiscernible] study recruited. We haven't -- we talked much during this few hours about one of the ADCs that is also in our pipeline for a while, BNT325 or TROP2. Why is that? Well, quite waiting for [ single agent therapy, some donation of random session ]. And also, we -- this is also an area that is interesting for TMC where we have combined this as our first ADC with Pumitamig. So we'll -- hopefully, we will have advanced and mature data of this drug together with Pumitamig in TNBC around mid next year.

So this is also quite exciting for us because as you saw for the HER3 and for the B7-H3, we're just starting these experiments, and we don't have a ton of data to present. But for the earliest cohort that we started, which was 325 or the TROP2 plus Pumitamig, we will have some mature data in TNBC asset like around mid next year. For the monotherapy, this is what we had, right? So nice response rates, also duration of response in ovarian is interesting. But we also admit that the space of ovarian cancer is super competitive with several TROP-2s ahead of us. So I don't think we will pursue the monotherapy necessarily here based on these data. But as you heard earlier, this is anyway not our play. We think the combo specifically with 327, but perhaps also with other agents is the most interesting thing that we can do.

Now very briefly in smaller or not smaller indications, cervical cancer is, as you know, the biggest women's cancer of them all. It's just small in a sense in Western countries due to the prophylactic vaccine, it's a little bit less prevalent these days.Thanks God. But we are generating a little bit of data in cervical for this drug, actually also for 326, not depicted here. So for the HER3, that looks interesting, and we have some more data in other indications that are not women's cancer, but just didn't find another place in this presentation.

So again, for 325, this is what I alluded to earlier. This is the early data we had, but we will have with quite a few more patients mature data on this combination soon. So this will be the most mature data of one of our ADCs plus Pumitamig. I don't want to oversell this, right? So the individual drugs also still have their own individual toxicities. But at least for us, it validates our strategy that combining the ADC with Pumitamig is feasible, that the additional toxicity is manageable if it's even relevant for -- in the case of Pumitamig and that it gives us enough line of sight to decide if we pursue this further or follow a different path. And we do this strategy with all of our ADCs in Pumitamig, as you probably have figured by now. So this concludes the section on gynecologic cancer. I think -- well, I'm sorry, the cervical cancer and the [indiscernible], we still have a few experiments ongoing. I tend to forget these because they are -- not because they are not relevant, but because we are typically talking a lot more about ovarian than about other indications. Now in ovarian, as I said earlier, for -- if I go back for a second. In ovarian, as I talked earlier, it's a super competitive field. right? So for us to pursue monotherapy in this area would require outstanding data. We may have some data that is really interesting. If it's outstanding also is a function of the duration of the responses that we're seeing. So we're not quite there yet declaring anything. But for Pumita as monotherapy, you see here quite some nice responses and also the duration of response is interesting.

Remember, B96 really the first study that validates or that suggests, I should say, that checkpoint can work. Also VEGF, many studies in this field in ovarian cancer have plus/minus bevacizumab, right? So bevacizumab is not fully entrenched in ovarian cancer. It's used by some, not loved by many also because of bleeding complications. For us, it's -- these data are clearly encouraging for ovarian cancer for Pumitamig alone in the sense that it is -- could be a valuable addition to an ADC or to plain old chemo, but we want to see these data mature further to take a decision because, as I said, super competitive field. Similar things apply to cervical cancer. As we said before, it's still a very important and very, very prevalent cancer, mostly in formerly called third world countries. But from a pure incidence, it's even still quite interesting. And if you talk to people that work in the field, it's still a sizable problem in the United States and in other countries and other Western countries.

These are our data in cervical cancer. We think, again, from a combination perspective, this drug would easily lend itself to combine with the little data that you saw for HER3 or for B7-H3. And this is probably a path we will be thinking about further over the course of the year once these data mature a bit more. That concludes the gynecologic cancer section. Again, from a take-home message, if you wish, we have mature data for single-agent TPAM. We haven't shown you these data because they are not published, but they are mature. We're very confident on the data that we know both the side effect profile and the efficacy profile. We had discussions with the agency, and we have started the confirmatory -- potentially confirmatory Phase III trial. For this to mature a bit further and have a few more patients enrolled will also determine if and when we will file the single-agent data.

Now for Pumitamig and ADCs, not much data with the combination of ADCs in these particular indications, cervical, ovarian. We're working on this. We want to see the individual data mature to know if we want to play in these areas, specifically ovarian competitive, cervical, not so competitive and could be something for us to think about more.

Yes, with this, we'll get to GI.

Thank you, Michael. Yes. As mentioned by Michael, so again, our take-home message is clearly, we want to execute and accelerate the work on the monotherapy and combination data in the proof-of-concept trials. But for the next section, so we want to focus on GI cancer. And on this slide, again, we repeat what we are going to talk about. So we will talk about Pumitamig data. We will talk about a combination with a bispecific antibody, which was introduced by Ugur already at the beginning, bispecific, which we are developing with our partner, Genmab, some data with ADCs, but also with our messenger RNA cancer immunotherapies.

So this is, again, our [ candles ] indicating in which settings in GI cancer we have studies ongoing. And as you can see, again, the solid line boxes indicate in which indications we have registrational trials announced and open. So in colorectal cancer, MSS and in gastric cancer. The dotted line boxes indicate really that these are proof-of-concept trials ongoing either in monotherapy or in combination with Pumitamig or with an adjuvant setting with our individualized vaccine treatments.

So MSS CRC. So this is a significant global incidence of colorectal cancer with an unmet high medical need and the requirements really for new therapies. And as you can see again on the left-hand side, so the majority of patients are in late stage, Stage III and IV. Important to note that only 3.5% to 5% of Stage IV colorectal cancer are MSS high, so sensitive to checkpoint inhibitor treatment. The majority is not sensitive to that. And so especially the MSS CRC indication requires additional treatment option despite the established standard of care, as you can see on the right-hand side. So currently, beva or cetuximab plus chemotherapy is the standard of care with 30 months of overall survival and a 5-year survival rate of 15 to 25 months. Is this enough? Most likely not. And therefore, some treatment options increasing this level are highly required and demanded.

So this is a brief introduction into what we are doing in colorectal cancer. This is a Phase II study, signal-seeking trial, which we are running 40 patients in total. We want to generate data for safety and signal seeking with different chemotherapy regimens. And so this is an ongoing trial and the data are not yet available and will be shared in time. In parallel, as you have seen most likely announced recently, we have really initiated our Phase II/III study with Pumitamig in combination with chemotherapy in colorectal cancer first line. And this is a study schema, Phase II part to -- for dose optimization followed by a Phase III part.

The other indications I mentioned already for the GI section of importance is gastric cancer. And also here, again, you can see that there is a high unmet medical need for metastatic gastric cancer patients as long-term survival outcomes are very poor and treatment options remain limited to chemotherapy in biomarker-negative patients. And as you can see in the CPS less than 1, the chemotherapy option is leading to 12.5 months median overall survival. In the above one, you can see in combination with nivo and chemo, median OS is 13.8 months. Also this indicates really clearly the need here for better treatment options and additional treatment options to elevate the efficacy of treatment and the perspective for patients in need. So again, I want to share with you also our study design of our Phase II/III study, Pumitamig in combination with chemotherapy in first-line gastric cancer, similar approach, a Phase II dose optimization part followed by a Phase III registrational part. So these studies are open and will enroll in due time.

In colorectal cancer, you have seen in the scientific in the preclinical section, the combinations, which we have really investigated with also an IO drug, which is a bispecific drug, which we are developing with our partner, Genmab, it's EpCAMx4-1BB. So in this setting, also MSS CRC, we want to see the additional contribution of this bispecific in a dose escalation part, which is ongoing, as you can see on the left-hand side, then we want to do a dose optimization part in second line and in first-line MSS colorectal cancer, followed by a pivotal study in Phase II, which will come a little bit later in the sequence.

One of the most difficult to treat indications in the GI area is pancreatic cancer. You know that the life expectation of pancreatic cancer patients is low. The treatment options at the moment are chemotherapy, leading to really absolutely low numbers in median overall survival and 24 months overall survival, for instance, with different chemotherapies as indicated here, with 10% only. The 5-year survival rate is 3% only. No need to really indicate here that there is a high unmet medical need. And therefore, we are also focusing on this indication and want to really develop treatment options with Pumitamig in first-line PDAC patients. So this is a signal seeking a proof-of-concept trial, which we have initiated testing different chemotherapies as indicated here with Pumitamig, and we believe that this will be the foundation for combination trials in this indication, also with ADCs and all the other treatment options which we have in our pipeline.

So this is, of course, really absolutely needed to indicate. So the combination with chemo is not the only thing we should evaluate. So there is also a need to combine with ADCs. And we have developed an ADC, which is based on the known expression patterns of CA99. Most of you know this is also known as a tumor marker usually sialylated Lewis A. And you can see the expression of this target is very high in different GI indication in pancreas, in bile duct, in colon, in ovary, in endometrium. So this is a perfect target to be really investigated with our CA99 ADC candidate, which we will move into the clinical dose escalation setting soon. And so we will investigate different dose levels, followed by expansion cohorts for PDAC focusing on second-line PDAC here and a basket trial in gastric and colorectal and others.

So I'm summarizing here also again. So we want to really execute again here, execute and accelerate our first pivotal trials to establish Pumitamig in colorectal cancer and gastric cancer, which is really in focus for us in the GI cancer space. And we want to also further evaluate and investigate as fast as possible combination with novel approaches, either with bispecific antibodies or with ADCs.

Moving to GU. Okay. So cancers, as indicated on this slide, again, you will see some data for the immune modulators, Pumitamig and Gotistobart, but also for our ADC B7-H3. Do you want to do that?

Yes. So for Pumitamig, there will be data or there are data from -- in renal cell in China, both for the clear cell and the non-clear cell. And for prostate, as you just heard, we have some data for Gotistobart and for the B7-H3, both of which would not necessarily be with PD-1 or PD-L1 because they simply don't work in prostate cancer. Maybe I'll let you talk about Pumitamig first, and then I'll go to you.

Yes. So these are data from our clear cell and non-clear cell study conducted in China, first line and second line. In 53 patients, Pumitamig was investigated either in Q2W or Q3W regimen. And the results of this are indicated on this slide. So we see for the clear cell population, second line plus overall response rate of 25%, disease control rates or over 80%, median duration of response, 19.6%. All this is very encouraging, in line with a median PFS of approximately 11 months. So this is definitely encouraging enough to further investigate Pumitamig in combination with additional novel approaches with ADCs or other drugs out of our pipeline. Same you can see here in first-line non-clear cell RCC across subtypes, same pattern, high overall response rates for this disease, disease control rate over 90%, 25 months duration of response and 15 months of median PFS.

And now this is your part, Michael, to talk about the ADC.

If you're wondering how many other disease areas we're going to show you, this is the last one, but you're not quite done yet. There's a few other slides from Ozlem and, of course, from Ramon. Okay.

So 324 or our B7-H3 again, right? If you remember the expression patterns, there was probably somewhere in the middle or a little higher than the middle. So pretty highly expressed in this space. And again, not to hammer home that message because most or all of you know this anyways that here, we would, for once, not do combination with Pumitamig for the simple reason that neither VEGF nor PD-L1 plays a role. And while we are bold in many of our clinical experiments, we're not so bold that we combine things that really don't make any sense. So we have for BNT324, several cohorts ongoing as a monotherapy in this setting. Why? Well, it's a slightly confusing area, at least for me, how this field develops with the radioligand therapies, specifically, but there's also newer ones, how they come into various states, first line, second line behind or together with the newer anti-hormonal therapies, et cetera.

Our approach, again, is relatively straightforward because we don't know exactly where this field is developing. We started off by multiple refractory patients, third line plus, right? That's where our initial data are. And that's where we saw also that this drug as a single agent works better than any other available chemotherapies that are out there, at least if you believe the published literature. But we also think this is such a high area of need where also very -- relatively benign therapies in terms of their toxicity profile and also therapies that are easy to give off the shelf and don't require special institutions or shielding or anything like this would find its role.

So this is our initial data from BNT324. Again, this is a third-line plus area, right? So these patients have exhausted all the available hormonal therapies and have either 1 or 2 and in many cases, more than that treatments received from the chemotherapy spectrum. And what you also see is here from the endpoint, it's RPFS, right? So R is the radiographic progression-free survival. So it's basically what we need to see is that the tumor doesn't progress. We obviously also look at the PSA and other factors in the blood, but this is the established standard for these type of patients. And similar to what you saw before, 6 and 9 mg per kg both work fine, and they don't seem to have big advantages over each other, which also then points the way for a potential Phase III trial that we're thinking of. So we are currently actively discussing what to do with these data. And we think there's some really interesting potential avenues, both in these later line settings where these data were generated, but perhaps also in an earlier setting.

Yes. So this is the last slide of our little part here, and I would hand back to Ozlem to drive us home before Ramon.

So you have heard about our disease area strategy and in some disease areas, we are more broader and in others a bit leaner. And you might ask what is the reason for covering all these disease areas and the answer is pretty clear with having the pipeline and a product opportunity with Pumitamig, we can go into all these disease areas. You have clearly seen that Pumitamig triggers with opportunity, but not only because we can, but also we must, if we look into other development architectures, for example, of the first-generation PD-1, PD-L1 compounds, it's very clear that those development streams, which have gone broad and first have been the successful ones. And having created the opportunity with our ADC portfolio to match multi disease area strategy given by Pumitamig, we are even more encouraged to do that.

With this, we now move to the next modality. We have discussed this morning a lot about PD-1, PD-L1 blockade. The mode of action for first-generation PD-1 PD-L1 checkpoint inhibitors is that they invigorate T cells directed against tumor mutations, tumor antigens, which are already around. So they do not induce new specificities. They reinvigorate those which are already there. And in tumors which have a high tumor mutational burden, the diversity multifront nature of these T cells is higher than in tumors with low tumor mutational burden.

And this is also why checkpoint inhibitors work better in non-small cell lung cancer and melanoma because you have this diversity of preexistent T cells there very broadly. In principle, and this is a hallmark of cancer disease, the efficiency of spontaneously generating T cells against tumor antigens is very poor. So you find only low numbers of tumor antigens recognized by spontaneously occurring T cells. And here is where this new -- this next modality comes in, namely cancer vaccine type immune therapies, which have a purpose of generating a new diversified poly-epitopic T cell pool. And on this T cell pool, checkpoint inhibitors can act again to reinvigorate.

This is also the reason why, in principle, in patients which are checkpoint inhibitor refractory and resistant, the combinations of checkpoint inhibitors with vaccine type immune therapies work because you have created another T cell pool. So the purpose of this type of immune therapies is to generate new T cells. And one of immunology tells us that in order to proficiently generate T cells against tumor antigens, which is not trivial at all, you need 2 critical components, one component is the very antigen, the tumor antigen against which you want to generate these T cells favorably or preferentially these multiple tumor antigens for a multipronged approach.

So it's the antigen, but what you also need is an innate immune -- signature innate immune modulation and activation, which ensures that after by the tumor antigen, the right T cells have been selected and start priming. Priming expansion, differentiation of these T cells is supported by interferon type 1 dominated innate immune signatures. And this is actually one of the reasons when we, the scientific founder started our journey, we selected the mRNA platform as our core vaccine type immune therapy strategy because mRNA as a format as well chemistry comes with an intrinsic innate immune stimulatory activity.

And this also explains the data you might have seen last year at ASCO and more prominently this year at ESMO from our friends out of MD Anderson, which shows that in tumors with high tumor mutational burden, non-small cell lung cancer and melanoma, adding the SARS-CoV-2, which does not bring the antigen, the right antigen here, but which we bring for innate immune stimulatory property of mRNA leads to a significant overall survival benefit. So what this in principle shows is the proficiency of mRNA to bring the second component of inducing potent immune responses, namely the innate immune signature.

That means what you see here in principle as an effect is like combining a checkpoint inhibitor with a TLR agonist on steroids, but it does not address another important component, namely the right tumor antigen. So our mRNA platform, which we use for our cancer vaccine type immune therapies is using mRNA in a nanoparticulate form. The specific platform version we are using in our ongoing clinical trials is based on uridine mRNA, which has a higher propensity to bring in the innate immune stimulation and interferon type 1 effects.

These are nanoparticulate mRNA formulations based on lipoplexes, which you can administer intravenously and they target specifically to lymph nodes compartments where immune responses are primed and triggered and bring the antigen and the innate immune signature to the right place. We are using this platform for 2 purposes: we deliver tumor associated antigen mixtures, which are indication specific. These are off-the-shelf vaccines or FixVACs to patients. And the other version of this is our iNeST, our individualized approach where we use neoantigens based on cancer mutations, not tumor-associated but cancer-specific antigens which are unique to every individual and therefore, require true individualization and on-demand production.

And this platform has been extensively tested in both variations. We know from studies we have already published and preclinical work as well that both components of proficient immune antigen-specific T cell response induction are delivered, the proinflammatory innate immune signature, which you see on the top from one of our studies with all the cytokines, and Ugur has also reported that delivered in a [ positive ] manner and also high magnitude T cell responses against the encoded antigens on the bottom, including self antigens, which shows the propensity to also break immune tolerance.

This is the summary of ongoing and recently completed clinical trials. On the FixVac side, we have, at the ESMO, reported that our BNT111 trial, which is a melanoma off-the-shelf vaccine has met its primary endpoint in PD-L1 refractory and resistant melanoma. The primary endpoint was objective response rate of the combination of BNT111 with cemiplimab, Regeneron's PD-1 compound.

And this combination showing an objective response rate, which was significantly above historical control, which is 10%. We have observed 18% objective response rate with 60% of these responses being complete responses. Duration of response was not evaluable yet, and we will get this data later next year.

We have an ongoing clinical trial with our BNT113 FixVac, which is being conducted in HPV16 positive, PD-L1 positive head and neck cancer. This is a Phase II/III trial for which we will see first readout in 2026. This is a randomized trial where we compare in first line BNT113 in combination with the standard of care pembro against pembro alone.

And then we have BNT116 ongoing and Ilhan -- no, Michael has extensively talked about this compound. You have seen that we are very broadly looking into different settings in non-small cell lung cancer and segments and combination partners and the plan is that next year to identify the right setting.

On the iNeST side, which is our individualized vaccine called Autogene Cevumeran, we have 3 clinical trials in the adjuvant setting ongoing, randomized trials, which the adjuvant setting is the one in which we see the best positioning of iNeST of our individualized vaccine. The reason is that in this early setting, tumor burden is low. We have minimal residual disease and a vaccine-based approach where you induce T cells is in principle, men against men fight, so on the cellular level, which means that low tumor burden is of benefit.

This is also a setting where immune-suppressive mechanisms of by the tumor disease are not established yet. Tumors are not heterogeneous but clonal and immune proficiency of the patient is still intact.

The colorectal cancer trial where we have selected the ctDNA post-surgery positive Stage II, high-risk Stage II population and where we compare iNeST to observation only after standard of care surgery and standard of care adjuvant treatment has finalized enrollment of patients, and we will have readout next year, in 2026 with final readout expected at the end of 2026.

The 2 other trials in adjuvant setting of pancreatic cancer and bladder cancer are more early in their patient enrollment status and are ongoing, both as well potentially registrational trials which are Phase II and randomized.

What I would like to do on the next couple of slides is share with you some insights from another trial, which is also on this slide, our first-line melanoma trial with iNeST, where we have compared iNeST in combination with the standard of care or one of the standards of care pembro in comparison to pembro alone.

This was a study which was not positioned in our favorite indication simply because, as you know, the drug development paradigm is that you start with advanced settings before you with a new compound can work yourself into earlier settings or adjuvant settings. This trial read out and has been presented at ESMO a couple of weeks ago. As I said, this is a Phase II trial in patients with first-line melanoma, unresectable locally-advanced melanoma and metastatic melanoma who have not received treatment for the advanced setting yet.

We have compared against pembro, have combined in the investigational arm iNeST with pembro. The key findings are that this trial did not meet its primary endpoint of significant improvement in PFS. We have observed in the pembro arm PFS -- median PFS of 7.9 months, which is a bit lower than expected from benchmarks and in the combination arm, 8.3 months.

What we, however, have observed is a difference in the 12 and 24 months OS rates in favor of the combination. This data is confounded by the fact that we allowed crossover of patients from the combination -- from the pembro arm only into the combination and 1/4 of the patients used this opportunity. We have observed good safety. The vaccine was also in combination with pembro well-tolerated and in line with what we see in general also in other trials with this platform.

Immunogenicity was strong and showed that the vaccine in principle also in this metastatic setting does what it is expected to do. We achieved multi-antigen or multi-neoepitope immune responses in the majority of patients. Immune responses were of high magnitude in the 10% to 15% of circulating T cells range and immune responses were also of durability.

We looked closer into the PFS data in a retrospective analysis in order to generate insights from this study. What we observed was, and these are Kaplan-Meier curves for PFS where we distinguish or stratify patients who have a narrower or broader breadth of immune responses, meaning patients who have developed an immune response against only one or two or three or more of their vaccine antigens of their mutations.

And what we have observed for the combination arm was that a trend of incremental PFS improvement in patients with higher neoantigen response breadth, which indicates that optimizing the breadth of immune responses and of the capability of our vaccine to induce these will benefit the outcome in our clinical trials.

Another observation is shown on this slide. As I pointed out, the secondary endpoint overall survival showed a numerical trend favoring the combination for the 12-month and 24-month OS rates. This was a secondary endpoint and no formal testing was performed. However, we looked into markers which might correlate with better OS.

We made 2 observations here. One observation was a trend of improved OS in patients with immune cell PD-L1 high as compared to immune cell PD-L1 low. Interestingly, this was only the case for the combination arm, not the pembro arm only. And interestingly, the tumor cell PD-L1 status did not play a role in predictive -- in terms of predictive value for both arms. This again indicates that our vaccine might benefit from combination treatments, which increase the hotness of the tumor microenvironment and create also in tumors which are per se, not PD-L1 positive or high a situation where we can trigger the effect here.

Another observation we made was that a trend of improved overall survival in patients with tumors with low mutational burden treated with combination versus the pembro arm, which again shows that tumors which basically don't benefit from checkpoint inhibitors are the ones which could benefit most from combining with vaccines, vaccine-type immune therapies. And this again supports our choices we have made for our adjuvant setting with tumors like PDAC, pancreatic cancer and colorectal cancer, which are more on the very low to low middle tumor burden side.

So with this to summarize our ongoing and next steps for our mRNA cancer immunotherapies in terms for Autogene Cevumeran, our individualized vaccine, our goals are all about bringing this home in the adjuvant setting. The next readouts to look out for will be 2026 for our colorectal cancer adjuvant trial. The other 2 trials in PDAC and bladder are going on. And here, it is about accelerating the enrollment.

With regard to FixVac, Michael has already pointed out all our activities with BNT116, which we are combining with different compounds, novel-novel ones and also assessing in different lines and subsegments of non-small cell lung cancer. And you will hear more about data from these efforts also next year. And then for our BNT113 head and neck trial, we expect data next year from our Phase II/III trial. So this year was silent for the FixVac and iNeST, but next year, you will hear more.

Thank you very much, Ozlem. I'm very happy to be here with you guys. And despite the fact that I haven't been with the company for long, it's great to see several familiar faces amongst you with all the interactions we have had over the last months.

So we will start this morning with our mission, "Translating Science Into Survival." And you think about it, this puts the patients at the center of everything that we do. And I do believe that value creation delivery is exactly having the same starting point. So today, I will be covering mainly 3 topics, what we have been executing this year, the principles and financial levers that we are applying on the R&D part of our P&L and hopefully exciting you about what's to come for 2026 and the years to come.

So 2025, I think, has been a year where we have successfully defended our COMIRNATY franchise. We enjoy very healthy market shares, stable pricing, strong brand recognition. And I think that has been reflected into our results that we talked about in the last quarter. As we advance our key oncology pan-tumor and clinical execution, you have seen already from Ozlem, from Ugur, Michael and Ilhan, we have more than 20 Phase II and Phase III oncology trials ongoing and more than 30 novel-novel combination cohorts across tumors.

If we talk about M&A, business development and collaborations, I think this has been quite a good year. So earlier in the year, we completed the acquisition of Biotheus. Then towards the summer months, we completed as well what I would say, a foundational partnership with BMS to maximize Pumitamig. And as we speak, we are in the process of closing the acquisition of CureVac that we are aiming to close still before the end of the year.

And then finally, as we strengthen our financial position and continue to drive innovation, for all of you that were listening to our earnings call last week, we have increased our revenue guidance to EUR 2.6 billion to EUR 2.8 billion from EUR 1.7 billion to EUR 2.2 billion that we had at the beginning of the year. And also, we have further solidified our cash position. We closed last quarter with EUR 16.7 billion in cash and cash equivalents, and that is, of course, supporting all of the scientific efforts that we have been hearing from my peers, but also keeping optionality for us to keep moving forward.

As we drive impact and innovation, I am very encouraged to see that we are expanding our later-stage oncology pivotal trials. As you've seen in the last 3 years, we have significantly increased the number of trials from 10 in 2023 to more than 20 in 2025.

And all of that being done with more or less the same level of R&D -- can you hear me or do I need -- and we have done this more or less with the same R&D level of spending. So 2023, EUR 1.8 billion. We are going to be closing this year around EUR 2 billion to EUR 2.2 billion.

Now if you ask me because I know that most of you will already be asking so what is going to be the guidance for 2026. So we are still working on that. I think it is too early to tell you if we are going to be more or less having the same levels of spending. I think as we expand our Phase III trials and widen our net of targets, this might change a little bit. But anyways, I think this has been good financial steering of the company.

So we have these 3 levers that I want to be talking about. So first is, of course, active portfolio management and strategy where we are really resource allocating our resources on programs that have the potential to really make a difference for patients and deliver value for our stockholders. Then, as I was mentioning before, innovative, tailored partnerships that will not only help us to advance the programs, but will also help us to widen our efforts and at the same time, strengthening our P&L, and I'm going to be talking about it a little bit more in details later on.

And then we are an innovative company. We are an innovation engine. And in order to do that, I think that our lead optimization, early science and discovery programs, they need to be properly funded. And for that, we have now visibility and dedicated budgets for all of them. And we are always in the lookout for opportunities in terms of in-licensing or out-licensing agreements that will not only support the science, but of course, support our financial position.

So if we talk a little bit more in details about how important the collaboration with BMS is for us. So I think it's not only allowing us to accelerate and maximize Pumitamig, but it's also strengthening our P&L and our financial footprint in the short term and in the long term.

So we are advancing more than 10 trials, including registrational studies and plans, I would say, very ambitious plans on CRC and gastric cancer. Most importantly, all of this -- so this is a 50-50 relationship. And what -- how this translates now is that before going commercial, so all of the efforts that we are investing on Pumitamig and all of the combinations that my peers have been talking about, so all of these expenses and investments are on a 50-50 basis.

And then finally, of course, with the anniversary payments and the upfront of $3.5 billion and thereafter potential milestones of up to $7.6 billion as we hit registrational and/or commercial milestones, I think this really bolster, I think 2 things, bolster our R&D efforts, but also it's kind of like a bode of confidence on how solid our platforms and our science is. So that's the path to value creation, right? So strategic portfolio management, now shifting towards later-stage derisked programs that have, as I was saying, the potential to really change the course of BioNTech.

My peers here will not allow me to say anything that is not correct. Optimizing productivity and efficiency is very close to my heart, and I have been having deep conversations with them on how we can better allocate our resources to be faster, better and more efficient, all with the intent of, of course, elongating our cash runway, giving us optionality as we move along and hopefully be in as good financial position as we currently are once Pumitamig hits its commercial stages.

And then again, finally, this will, of course, allow us to be fast, to be efficient and readily scalable by the time we are hitting the market and hopefully reaching as many patients as we can.

So if 2025 has been exciting, I would say 2026 has the potential to be even more exciting as we gain momentum in our combination strategy, we shift from modalities to disease areas as has been very professionally outlined by my peers. And then we gain acceleration into late-stage programs.

I'm not going to end up here quoting more scientists because we already have heard from very bright minds, from Ugur, Ozlem, Michael and Ilhan, but I would maybe mention a management scientist that you guys all know, Peter Drucker. He says the best way to predict the future is to create it, to influence, to shape it. And I believe we at BioNTech are just doing that.

So with that, I would like to open up for our Q&A panel and invite my peers to join me here as well as Annemarie Hanekamp, our Chief Commercial Officer.

Thank you very much for your time, for your engagement and also for analysts for your analysis and questions that are always challenging, interesting and will always help us to be better and improve.

So with that, we'll open up for Q&A.

I have two. Okay. I have 2. One, on your ADCs, you're still focused on topoisomerase 1 while you're innovating on the target. And I wonder why, particularly given the MMAE payloads have some additional immune stimulatory benefit or any other novel payloads? And then I'll ask another one.

So I think the topoisomerase payloads come with an additional selectivity, Daina, particularly topoisomerase is a target which is expressed in proliferating cells. So that means all the side effects are focusing on proliferating cells. Thereby, we are getting two types of specificity, on the one side based on the target, the second based on the proliferative index of the cells. We believe that this is, at the moment, a better strategy for developing safe and effective combination therapies, and we are working on novel payload approaches that follow the same logic.

Daina had a second question, you took away her line.

Okay. One more. I just came from SITC, and we know what's happened this year in in vivo approaches for CAR-T. And in some ways, I think you guys were leading with the LNP and your various approaches of either making RiboCytokines or T cell engagers or CAR-T. And it's not on the priority. And I wonder how you think about nurturing some of that really exciting innovation where you were leading sort of from a corporate strategy perspective.

You go ahead and I'll....

No, you are right. So we have indeed a portfolio of in vivo approaches. But the limitation is not the technology itself, but coming up with a viable approach, multi-target approach. And we see at the moment, really the focus in our core strategy with IO ADCs plus mRNA vaccines. This is something which will go with, I would say, the third wave of personalization where this type of in vivo approaches can be really combined with multiple targeting approaches.

At the moment, the limiting factor is there are not that much CAR constructs targeting tumor-specific antigens. The TCRs are too segmented in too many sub-indications, that means HLA restriction plus the target expression. So this is something which has a great potential, but we think it's too early to drive that with full energy into the clinical application.

Yes. I would echo that there is a lot of space for improvement in the technology further optimization space as well. And this is work we are continuing. And sometimes it's good to wait until you move the most optimized version of your iteratively improved technology into clinic.

Tazeen Ahmad from Bank of America. Let me ask maybe one question on pipeline and one finance question. For pipeline, you've talked a lot today about upcoming data. For those of us who've been on this journey with you since IPO, it's good to see that we're going to see a lot of data sets coming through starting next year.

But you specifically focused on ADCs. This is a class that some might consider to be an older class, but I think Ugur you yourself said that's still in the early innings of looking at ADC. So I maybe wanted to ask about the upcoming Pumi plus ADC joined data that you've guided to expect next year.

Can you just give us a little bit of color on what exactly we're going to see and what decision-making processes will occur as to which ones you may choose to move forward just for modeling purposes, we'd like to get a sense for that?

And then the finance question, I'll just go ahead and ask that now. So the process that the company has talked about in terms of maintaining a strong balance sheet while looking for good external opportunities has been followed through since the IPO. As the company moves forward, you do have a very large cash balance. How do you balance out your increasing R&D expense, let's say, with increasing data sets coming internally versus the desire to continue to look for external assets outside and with your threshold for what you want to pay increase relative to what you were doing before?

I can take the first question. So our approach is really assessing single compound activity for Pumitamig in multiple indications and a single compound activity for our ADCs. We see as a general pattern for Pumitamig and that we have increased objective response rate. We have prolonged PFS. And for our ADCs, we identify particularly doses that are safe, but still are associated with a prolonged PFS. We believe that the combination of both will not only have an improved response rate, but particularly help us to further increase the PFS and translate into OS. That's our strategy.

And I think on the finances, so I think you're right. The way I think about how to really preserve our financial strength is relentless priority on what we are doing. So every time that we are talking about resources, how much these are going to be costing, what are the clinical trials that we are going to be starting, particularly early science and thereafter, it's like they really have to be bolstering and underpinning our strategic priorities, which I think they have been very clearly outlined by my peers.

Then on your second question, if I think this is giving us the firepower to look after even bigger opportunities, I think -- I would say, yes, I think if there is anything that is strategically aligned with what we are pursuing that will give us further insights and ways to get into this market in a way that will even prepare us better for commercialization stages or further expanding our pipeline, we will always be looking after these and very interested in analyzing them.

Is there an upper limit on how much you'd be willing to spend right now?

Not really.

This is Jay Lee with Jay Olson with Oppenheimer. Maybe again, 2 questions from us. First, I'm just wondering about Gotistobart. It seems like it's like some interesting data from the squamous cell, squamous non-small cell lung cancer, but just wondering why you haven't started the combination with Pumitamig, whether it makes sense to combine those 2 agents?

And second question is on your dose selection from the small cell lung cancer. The lower dose is selected, I just want to confirm that the FDA has agreed on this decision and how to think about other ongoing Phase II/III trials on which dose maybe make more sense and also the dose for your ongoing ADC combination?

So I can try to answer the first question on Gotistobart and Pumitamig. It's an intriguing combination for sure, and we would love to pursue this. We probably will, but we are through certain historical reasons, restricted with the collaboration with our partner, Regeneron, right? So we may or may not be able to do this fast to an approval. Otherwise, data generation will be something we will certainly do.

For ADCs -- sorry, what was exactly.....

[indiscernible].

For Pumitamig part or the ADC part?

First Pumitamig part and then you asked also regarding the ADC dose. So you asked regarding the selected dose for small cell lung cancer. And the answer is yes. This is discussed and aligned with the FDA.

[indiscernible].

We are looking constantly in the data evolving and we'll make really data-driven decisions about the dose selection. I shared with you also the designs of the new trials, which we are generating, which has a dose optimization part still included. And this will drive.

And with regard to the published data, you might have seen that 20 and 30 milligram per kilogram of Pumitamig is very similar with regard to the safety profile. We are still assessing whether 20 to 30 milligram makes a difference in the objective response rate and PFS. And this is something that provides us -- will provide us additional information, updated data for TNBC coming in a few weeks, right?

Yes.

Okay. Questions from this side of the room. Akash?

So can you talk about B7H3 and why this seems to be your kind of preferred agent instead of the TROP2 for not just small cell but also NSCLC. And maybe it's working in squamous, non-squamous.

But I think the other question is really, given your response to some of the data we saw at ESMO from Merck and Kelun with the sac-TMT drug, right, you're seeing on the low end of 4-month OS benefit, it looks like it's trending to 6 to 12 months. So I think the question is your B7H3, you guys talk about safety, but could you have an efficacy signal that would look like that?

So maybe first to say for non-small cell lung cancer, we have several options, right? So B7H3 is one of them. TROP2 is another one, but the actual -- probably the least data we have, but also the most promising is on the HER3 side. So taking the competitive landscape into account, when I showed the data on non-small cell, I think I said something like this looks interesting and perhaps competitive, but there's others that are ahead of us with other moieties such as TROP2.

And from a place where we want to win this, we want to see single-agent activity that is superior, right? So for B7H3, the NSCLC data is good, but it's not outstanding. Small cell perhaps, yes. So small cell, also, as I alluded to, the duration looks quite interesting. NSCLC, we'll see the data mature. We'll take an informed decision. We do the combination with Pumitamig. But assuming that we will not see anything magic with B7H3 and Pumitamig in non-small cell, I think we have better options.

Maybe just to put that. So if you were to guess right now, what your novel-novel combo would be in first-line NSCLC, what would those 2 agents be? And when would we see that combo get initiated?

So part of when we need to see that combo initiated is when we will have also a more mature Pumitamig data, right? Like because we will not start the Phase III with Pumitamig without knowing where Pumitamig is actually heading towards, right? We have the Phase III started, but we still want to see that mature.

Also, we don't want to have like a year after a successful Phase III, another Phase III, which would basically not be very, very cost efficient to doing this. I think we still will see on our 3 ADCs, which one is the best. When exactly, I don't know. Like it will be probably during the course of the coming year, probably more towards the end of the year for taking an informed decision if and if so, which one to take in combination with Pumitamig.

Okay. Terence, next?

Terence Flynn, Morgan Stanley. I had a 2-part question on iNeST. Ozlem, you mentioned that based on the Phase II melanoma data, you were thinking about optimizing the breadth of immune responses as that could benefit the outcome in these trials. And so are there any plans to update the algorithm you're using in the antigen selection process? And how dynamic is that during clinical trials? And what does that require from an FDA input perspective? So that's the first part of the question.

The second has to do with manufacturing scale-up and turnaround time for iNeST. Just anything you can provide in terms of an update there in terms of where that stands and how quickly you could expand that if successful in Phase II?

The first part of your question, we are constantly working on optimizing our algorithm also given the fact that with the data we generate in our clinical trials, we also generate the opportunity also with the use of AI to reflect these new insights into the ways how we predict how we predict our neoepitopes. There is very limited opportunity to introduce innovations in the algorithm into an ongoing trial, obviously. There is some wiggle space which we discuss and get support from regulators, but changing the algorithm entirely during an ongoing study is not feasible. So this is something which then comes with new studies.

What was your second question?

Manufacturing scale and turnaround time.

Yes. We have a manufacturing time depending on trial around 40 days. And there are also trials in the adjuvant setting where you can deliver the vaccine in 6 weeks or 8 weeks because the patients can stay longer. For pancreatic cancer trial, we try to stick to 40 days where the patients have only a limited window.

Okay. Up next, Bill?

Bill Maughan, Clear Street. So early in the slides, you had preclinical data that put Pumitamig on the same graph as an Ivonescimab analog and 6 out of 6, it looked more potent. So just are you comfortable making a claim that you believe you may have -- that you likely have a more potent molecule?

And then secondly, just to get Annemarie involved here. You've talked before about building a commercial organization to support the endometrial launch that can then be leveraged for future oncology launches. So can you just kind of list out your priorities on your to-do list for 2026 and how you can start to set that foundation?

I'll take the first question, Annemarie. So we use, of course, different positive control. With IO being one control, it's indeed consistent and that the affinity appears to be differentiated, whether this translates in any way into a differentiated clinical thing we cannot make a statement here. It's a quantitative difference of the 2 molecules, but that might not play a role in the concentrations of the antibodies applied to patients.

And to your second question, yes, we've guided to endometrial for T-Pam as a strategic launch. And I'm very pleased that we're working towards establishing the commercial footprint in the U.S. We have a team of well-rounded professionals from both big pharma, small pharma with a lot of deep experience. And we're getting -- we're out in the field right now compliantly to ensure that we have the right metrics in place for when we have that strategic launch and that foundation.

And very excited also about this BMS partnership to just throw that in as well because the opportunity with Pumitamig is such that we want to make sure that when it plays out, patients across the globe have the opportunity to get access to Pumitamig. Now we're very mindful that, of course, building a commercial stage organization just from scratch, maybe a little bit to chew off in one goal.

So this partnership, which is also a co-commercialization partnership will allow us to focus on predominantly the U.S., where in oncology, 60% of your revenue comes from followed by the markets that we're getting ready for today as well, other big markets like the EU 4, U.K., Japan contributing to another 20% of commercial value and then making sure that the other markets in our initial start-up phase can be accessed by a partner like BMS for Pumitamig. And as we're then building to our novel-novel combinations where we will have a leverage, we can slowly gradually, but with focus, build up that commercial presence.

This is Jay Park from GIC. I have 2 questions, please. On small cell lung cancer, we have clearly seen a good translation of Chinese response rates into U.S. response rates. Every cancer is, of course, different. So if you could just talk to us about your confidence level in kind of Chinese to global translation in other cancer settings as well? And I have a second question after this.

Yes, sure. So the data presented and which you have seen is indicating really in small cell lung cancer, definitely, it is confirmative, right? So we are seeing same levels even in global, a little bit better from the response rate and duration of response. We believe there should be no difference in other indications, but this, of course, is the reason why we are running trials, signal-seeking trials in both regions, global and in China. And so this will be, again, guided by the data, but there is no reason to believe that there are significant differences regarding the response and the efficacy. Safety, by the way, is also very similar.

That's great. My second question is, I think on early ELCC data, TROP2 ADC plus Pumitamig looks very combinable, looks very safe. Obviously, we don't have other, say, 324 plus Pumitamig combination data just yet. But is there anything that you see from 324 monotherapy data that makes you think that it could be maybe as combinable as 325 or other ADCs for that matter?

Actually based on the large portfolio of indications that we have assessed with Pumitamig and with our ADCs, we know the side effect profiles of the compounds. And so far, we clearly see that the ADCs have a different side effect profile associated, for example, with dose-dependent neutropenia, which is not a challenge for Pumitamig.

Pumitamig itself has a very low rate of adverse events, immune adverse events. We don't expect any influence here. So based on that, what we are seeing so far in the early studies, this looks very good. And we don't expect added toxicity. But of course, this is the game go into the combination therapies, assess the safety of the combination, identify the right dose and then go for Phase III clinical trials.

Malcolm Hoffman here for Evan Seigerman from BMO Capital Markets. In metastatic CRC, can you talk about why Pumitamig may be differentiated versus other PD-1 VEGF bispecifics? I know Pfizer recently highlighted the plan start of their Phase III in first-line trial next year.

And at the start of the presentation, you highlighted the potential benefit of tumor microenvironment targeting with PD-L1 versus PD-1 bispecifics. Is this something that we should be anchoring on when thinking about the potential relative benefit? Or is this more just mechanistic theory at this point?

I think it's early in this indication for CRC to predict any differentiation of the PD-L1 VEGF class versus the PD-1 class. Colorectal cancer is an interesting indication since at least the microsatellite stable indication is so far not open for anti-PD-1 treatment. And this is definitely something where we would like to see the signal, we would like to see the PFS, the response rates. And that's also the reason why we are assessing different doses in this indication to ensure that the findings that we got now in one indication is really can be transferred into another indication. It is quite possible that the dose identification could become even a differentiation factor for the response rate and durability.

Asthika Goonewardene from Truist. I wanted to tag on to Daina's question. So some payload optimization has been described in the literature to increase the potential for immunogenic cell death. I see some groups have been able to show increased DAMP expression, et cetera. Has any of this work done with your payloads, just given that you have such a broad assortment of ADCs to combine with? And then how would you compare your payload versus DXd and MMAE on immunogenicity?

Yes. See, we are doing that intensively. And we clearly see and I have shown you data in -- also in syngeneic tumor models where we clearly see that this is not only additive but synergy based. The immunogenic cell death is mediated by release and a number of well characterized factors. So there is no difference with regard to the ADC. So at the end of the day, it's a cell death that is mediated by necrosis by release of intracellular signals, dead molecules. This is happening independent from the ADC class. But the question that we have with regard to the innovation is can we go beyond this classical cell death mediated immunogenicity approach. And this is something where we believe in the next 2 years, we will see innovations coming, and we are working ourselves in-house on some pathways to further enhance this immunogenicity.

And keeping on the IO them here. So PD-1 historically has not shown great data in MSS-CRC, but there's obviously a lot of interest in pursuing this with the PD-1 VEGF bispecifics. I'm guessing the effect there is more driven by VEGF component, but have you done any work internally to show that there might be some immunological activity as well?

In microsatellite stable colorectal cancer, there are publications and clinical data of combination treatments in MSS-CRC, including, for example, anti-CTLA-4 and anti-PD-1 treatments, combination chemotherapies. So no tumor is actually immunologically silent.

CRC has a special aspect to ensure that this immunological silence is not broken. But biologically, for example, if you look to microsatellite stable tumors and instable tumors, you will not find a different biology. It's just more mutations. So it's, at the end of the day, it's a way to enable tumor T cell infiltration into tumors and get the things running. And I'm sure that we will make in the next years, great progress in making colorectal cancer as -- make colorectal cancer, MSS-CRC as a highly immunogenic tumor type suitable for cancer immunotherapy.

And the question is also that the PD-1s have not been successful in colorectal cancer just means that there was insufficiently powered studies, not a statistically significant difference, right? But you do see in clinical practice that there are MSS-CRCs that -- and this has also been shown in the studies, which -- who seem to have an effect.

So the question is what happens if you concentrate the blocking via targeting via VEGF into the tumor microenvironment and thereby change the [indiscernible] there. And that means we should not be -- we should not shy away from indications in which we don't see PD-1 effects, but would like to try this bispecific.

So it is a nice bridge to what showed at the beginning. So Pumitamig is more than combination of the more specific parts of this equation. So what you see before in combining PD-1 and, for instance, in colorectal cancer is not indicating what you can expect here from this bispecific with an enrichment in the tumor microenvironment. I think that's an important part.

Will Zhang from Wells Fargo. So just we saw competitor data at ESMO that showed basically a diminishing treatment effect as you increase PD-1 expression in subgroups. And just wondering a response on your end to that. And also, can you give us some color on how BioNTech is thinking about a potential treatment effect in checkpoint inhibitor indications versus non-checkpoint indications?

Yes. Yes. I think this is an observation, a trend that we are also seeing. And this needs to be translated into the clinical strategy, how to deal in these indications to have a better hazard ratio also in these indications which are high PD-1 positive.

Okay. So that concludes this Q&A panel and our event today. Thank you very much for everyone for attending. And we'll now open for a networking event, which will be just through here in the lobby.

Welcome to BioNTech's Third Quarter 2025 Earnings Call. I will now hand the call over to Doug Maffei, Vice President, Strategy and Investor Relations. Please go ahead.

Thank you, operator. Good morning and goodafternoon, everybody, and thank you for joining BioNTech's Third Quarter 2025 Earnings Call.

As a reminder, the slides we'll use during the call and the corresponding press release can be found in the Investors section of our website. On the next slide, you will see our forward-looking statement disclaimer. Additional information about these statements and other risks are described in our filings with the U.S. Securities and Exchange Commission, or SEC. Forward-looking statements on this call are subject to significant risks and uncertainties and speak only as of the date of this conference call. We undertake no obligation to update or revise any of these statements.

On Slide 3, you can find the agenda for today's call. I'm joined by the following members of BioNTech's management team: Ugur Sahin, Chief Executive Officer and Co-Founder; Ozlem Tureci, Chief Medical Officer and Co-Founder; and Ramon Zapata, Chief Financial Officer.

With this, I'll hand the call over to Ugur.

Thank you, Doug, and warm welcome to you all as you join us today. As BioNTech has grown, our vision has remained constant, namely translating science into survival. We are building a global immunotherapy powerhouse, a fully integrated biopharmaceutical company with the science, scale, capabilities and the aim to deliver multiple approved therapies and reach patients in need.

Cancer remains a systems problem, heterogeneous across patients and variable within individual tumors. We believe the future lies in rationally designed combinations, pairing potent and precise mechanism of action that create biological synergies.

To this aim, we have purpose built a diversified clinical pipeline spanning mRNA immunotherapies, next-generation immunomodulators, ADCs and other targeted agents that enable development of potent, personalized precision medicines and Novel-Novel combinations across solid tumors. Our goal is to address the full continuum of cancer from resected high-risk tumors in the adjuvant setting to advanced and metastatic disease to treatment-resistant and refractory cancer. Our strategy concentrates capital on 2 priority pan-tumor programs that are designed to anchor various combinations.

One is Pumitamig, formerly BNT327, a PD-L1 VGFA bispecific that unit checkpoint inhibition with vascular normalization in one molecule. We believe Pumitamig is particularly suited as a next-generation IO backbone to combine with chemo ADC and other immunomodulators. The other is mRNA cancer immunotherapy that is designed to activate and educate the immune system with precision.

Our mRNA cancer immunotherapies have advanced in randomized late-stage trials with focus on the adjuvant setting. Both approaches have disruptive potential and align with our vision. We believe these programs could establish new standards of care and improve survival outcomes.

Together, these programs provide breadth, optionality and scalable registrational path across solid tumors. We are investing deliberately scaling clinical development, building manufacturing that ranges from personalized to large scale production and preparing for commercialization in key markets to reach patients in need.

Now turning to how our achievements in the quarter relate to our vision and strategy. We see Pumitamig as a potential standard of care across diverse tumor types spending settings already treated with checkpoint inhibitions and those that checkpoint inhibitors have not demonstrated benefit.

With our partner, BMS, we are executing a both registrational program. This quarter, we made significant progress in advancing Pumitamig, taking concrete steps towards our registrational plan.

In Q3, we progressed enrollment in two global registrational trials in lung cancer and remain on track to initiate the TNBC Phase III this year. This keeps us aligned with our target of first potential launches before the end of the decade. Across the portfolio, more than a dozen signal-seeking studies progressed, either with chemo backbones to expand into additional indications or at Novel-Novel combinations with BioNTech proprietory assets. Importantly, we advanced clinical mono agent profiling of potential combination partners, helping to derisk dose, schedule and safety assumptions for future registrational design.

These steps, including Phase III recruitment momentum, initiation of new combination cohorts and deeper combination partner characterization are all about informing the next wave of registrational trials planned with BMS from now onwards.

Turning to our mRNA cancer immunotherapy platform. In October, we presented Phase II trial updates for BNT111, our fixed candidate in anti-PD-1 resistant refractory melanoma and for Autogene cevumeran, our fully personalized mRNA cancer immunotherapy in first-line treatment of metastatic melanoma. Our data reinforces our view that adjuvant settings where tumor burden is low and immune control is most effective represents where mRNA immunotherapy can deliver the most significant benefit to patients. Ozlem will share details on how this readout sharpening our development focus.

This quarter, we hosted our second AI day. It underscored that we are not only pioneers in new pharmaceutical technologies but a fully integrated AI tech bio company with AI tools that enable discovery and development of innovative medicines. We showcased AI-based approaches designed to convert complex dimensions of data diversity into personalized therapy development.

We demonstrated two distinct strengths of our AI capabilities, addressing inter-patient heterogeneity and [ intra ] tuner variability and driving precision and potency in our treatment approaches. With regard to our COVID-19 vaccine franchise which is partnered with Pfizer, we successfully launched our variant adaptive success for the current season following regulatory approval.

With this launches in major markets and with a strong balance sheet over EUR 16 billion in total cash, equivalents and securities, we have the resources and the flexibility to fund the oncology transition by maintaining a disciplined P&L Simply put, we are transforming scientific advances into late-stage progress in our priority oncology program across indications. In parallel, we are building the capabilities and the financial strength to translate positive data asset into market opportunities and most importantly, into patient benefit.

With that, I will hand over to Ozlem to discuss our clinical execution and near-term data readouts.

Thank you, Ugur. I'm glad to be speaking with everyone today. I'll start with a top line status of programs that are heading our pipeline before moving to specifics. Firstly, with our PD-L1 GFA bispecific antibody, Pumitamig, we are executing a broad registrational program in partnership with Bristol-Myers Squibb.

Second, for our mRNA cancer immunotherapies, we have recently provided two Phase II updates that support and inform our current development strategy. And third, for trastuzumab pamirtecan or T-Pam, our HER2-targeted ADC, known previously BNT323 that we developed with our partner, duality. We continue to progress towards first BLA submission now planned for 2026, subject to regulatory feedback.

We are evaluating T-Pam as a monotherapy into randomized Phase III trials, one in metastatic endometrial cancer and 1 in breast cancer. For both studies, we expect data in 2026. We have also initiated a signal-seeking trial evaluating the novel combination of T-Pam with Pumitamig.

For Pumitamig, let me recap the clinical development framework. Our refined free wave plan that we are pursuing with our partner, BMS. Wave one aims to establish Pumitamig and free foundational first-line indications. Small cell lung cancer, non-small cell lung cancer and triple-negative breast cancer through global registrational Phase III trials. With 2 and 3 aim to expand the opportunity of Pumitamig by amplifying its differentiation. And we do this in two dimensions.

First, for signal seeking studies in combination with standard of care across tumors that inform our indication strategy and prioritization and second, for Novel-Novel combinations notably with our ADCs that enhance efficacy. We have delivered tangible progress on all these three ways waves in Q3.

Regarding Wave 1, in small cell lung cancer, the global Phase III is recruiting and the Phase III dose is locked based on the dose optimization data set with a safety profile consistent with known PD-L1 BGS chemo experience. In non-small cell lung cancer, the Phase II part of a seamless Phase II/III trial achieved full enrollment and the Phase III portion is recruiting.

In TNBC, we remain on track to initiate the global Phase III this year, targeting the PD-L1 low segment where unmet need is highest. This slide shows additional studies. These are supportive studies for dose finding, setting refinement and regional programs that contribute to the body of evidence supporting our three foundational global Phase IIIs.

Wave 2 [ CF1 ] is our expansion engine. We now have more than a dozen chemo-based signal-seeking studies across tumor types and lines of therapy. In Q3, we opened new cohorts and continue to mature data sets that will feed into our pivotal planning. This helps to ensure that the next registrational wave is evidence led and prioritized by benefit risk profiles, patient population size, well-informed study design and commercial opportunity alongside other key factors in our decision matrix.

Spearheading heading this next round of pivotal trials, we are initiating two trials in partnership with BMS, with registrational intent for Pumitamig in combination with chemotherapy in first-line microsatellite stable colorectal cancer and first-line gastric cancer.

Wave 3 elevates the potential of Pumitamig for Novel-Novel combinations to maximize its clinical impact, reinforce class differentiation and set up a multiyear pathway to sustain the value and the longevity of the drug into the new decade. Here are several combo cords of Pumitamig with our ADCs or other novel compounds are already enrolling and have gained momentum in Q3. Initial data over the next year will inform decision-making for our first pivotal combination regimens.

In parallel, we are continuing mono agent profiling of potential combination partners to set clear baseline for dose safety and sequence. Taken together, Q3 was a quarter of strong clinical execution that strengthened our registrational core widened our expansion engine and advanced the Novel-Novel combination rationale that we believe will further distinguish and elevate Pumitamig over time.

Let me now highlight two Q3 focal points. First, our first-line small cell lung cancer registrational program and why the recent updates are catalytic and second, our advances in mono agent profiling for refining our combination strategy. Small cell lung cancer remains a challenging immunologically cold disease in which responses to immune checkpoint therapy tend to be short-lived, resulting in modest gains over chemotherapy alone and low long-term survival.

Over the last 18 months, we have built a cohesive evidence base across multiple Phase II studies in first and second line small cell lung cancer. Initially, in China, and now globally, showing encouraging activity and a manageable safety profile. This quarter, at WCLC, we reported the first global data from our Phase II dose optimization study in untreated extensive-stage small cell lung cancer evaluating two dose levels of Pumitamig plus chemotherapy.

All patients irrespective of dose had disease control at 20 mg per kg, we observed a confirmed objective response rate of 85% and a median progression-free survival of 6.3 months per kg yielded a confirmed objective response rate of 66% and a median PFS of 7 months. Median overall survival data were not yet to mature. Safety remained consistent and manageable with low discontinuation and no new signals beyond those typically seen with chemo and PD-L1, VEGF-A agents. Two points are worth emphasizing. First, dose clarity, which is a critical derisking step for any registrational program, the global dose optimization readout allowed us to lock the Phase III regimen at 20 mg per kg every 3 weeks.

Second, consistent performance across regions. Earlier China data sets in first-line, extensive-stage small cell lung cancer showed robust activity and manageable safety. The global Q-Free data are consistent with those findings which further strengthens our confidence in Pumitamig's benefit across patient populations and practice patterns. Together, these results support our ongoing global Phase III ROSETTA Lung-01 trial, which compares Pumitamig plus chemotherapy against atezolizumab plus chemotherapy and untreated small cell lung cancer.

In parallel, in China, we continue the second-line randomized Phase III trial of Pumitamig plus chemo versus chemo alone. This quarter, we expanded our Pumitamig small cell lung cancer program to include Novel-Novel testing, and we launched signal-seeking studies of Pumitamig plus our B7-H3 ADC, BNT324 in both first and second-line small cell lung cancer.

As Phase III readouts and Phase I/II ADC combination data sets mature, we will be increasingly well positioned to select and advance additional regimens designed to establish long-standing presence in small cell lung cancer. This brings me to our strategy for advancing combinations of Pumitamig with other Novel agents, one of our key differentiation approaches. The cornerstone is establishing monoagent evidence of activity durability and safety before we decide to pair with Pumitamig.

For our B7-H3 ADC, BNT324, our monoagent database has expanded significantly over the last 12 months. B7-H3 road expression profile aligns well with Pumitamig's consumor opportunity. In small cell lung cancer, BNT324 as monotherapy achieved an objective response rate of with deep tumor shrinkage across the waterfall, an unusually strong single-agent signal in this setting.

In non-small cell lung cancer, activity was observed in both squamous and non-squamous disease, including an EGFR mutant subset with an objective response rate of 21%. And in heavily pretreated metastatic castration-resistant prostate cancer, we observed meaningful tumor shrinkage with BNT324 and the durable radiographic progression-free survival with a manageable safety profile.

Recently at ESMO, we reported data for our TROP2 ADC, BN325 in second line plus TNBC with an objective response rate around 35%, is the control rate of roughly 81% and medium progression-free survival of about 5.5 months. Also in Q3 for our HER2 ADC T-Pam, we saw a substantial expansion of a monopropdata base by the DYNASTY-Breast02 Phase III trial our partnered reality bio conduct in China that met its primary endpoint of PFS improvement versus trastuzumab emtansine in pretreated patients refer to positive unresectable or metastatic breast cancer.

T-Pam is another promising combination partner with the potential to expand Pumitamig's therapeutic reach into the HER2 expressing tumor spectrum. Taken together, these data provide a clear monoeropy baseline and help us set the bar for add-on benefit from Pumitamig plus ADC combinations. Across these programs, the mechanistic rationale is consistent. VEGF-A blockade can normalize resecure to improve ADC delivery, while PD-L1 inhibition can convert ADC mediate cytotoxicity and antigen release into a broader durable immune response. Aiming for deeper debulking plus immune control. These represent complementary mechanisms that single agents cannot engage simultaneously.

So operationally, we made two key advances in Q3, continued mono agent profiling to refine dose and sequence and qualification of our add-on benefit threshold. And expansion of Pumitamig plus ADC cohorts across prioritized settings. Of note, our go/no-go decision-making process is driven by a holistic evaluation that goes beyond efficacy, sickness and safety profiles. We strategically assess market opportunity, unmet needs, competitive dynamics and way other key factors to ensure every decision aligns with our mission to deliver transformative benefits for patients.

Moving now to our second oncology cornerstone mRNA, cancer immunotherapy. iNeST is individually manufactured per patient to target personal neoantigens. The biology and our clinical experience point to greatest relevance in earlier disease settings where lower tumor burden allow the immune system to consolidate control. Our ongoing randomized Phase II trials are designed to test that premise in a rigorous way.

Of the shares expected includes BNT111 for melanoma, BNT113 for HPV16 positive head, neck cancer and BNT116 for non-small cell lung cancer target shared antigens and is intended to pair with checkpoint inhibitors and increasingly, our next-gen backbones. We continue to advance execution and evidence generation across multiple tumor settings, while keeping optionality around where and how FixVac is best positioned longer term.

This quarter at WCLC, we presented results for [ BNG ] plus cemiplimab is consolidation treatment in unresectable Stage III non-small cell lung cancer. We also presented data at ESMO from two randomized Phase II trials in melanoma, one with BNT111 FixVac and the other for [indiscernible]. I will briefly walk you through the melanoma readouts and their implications.

Starting with BNT116 FixVac in the high medical need population of patients who had relapsed or not responded to PD-1 treatment. The Phase II study evaluated BNT111 plus cemiplimab against a historical control objective response rate of 10% reported for anti-PD-1 treatment in this setting. The study included two calibrator monotherapy cohorts to characterize the safety of each agent and its activity on objective response rate. The objective of this design was signal characterization, not cross arm efficacy claims.

In the monotherapy cohorts on progression addition of the second agent was permitted. More than half of the patients in each arm opted for this addition after a median duration of [indiscernible] monotherapy treatment of brown formats. The study met its prespecified primary endpoint by rejecting the nalhypothesis of an ORR of 10% and with statistical significance. The ORR of the combination was 18%, including deep and durable responses.

Notably, 2/3 of our responses were complete responses. Supporting the depth of activity. Follow-up showed a positive impact on long-term survival, 37% of patients were still alive after 24 months, 21% were free of tumor progression. Safety was manageable, driven largely by expected mostly Grade 1,2. Cytokine-related events, consistent with the mRNA platform.

BNT111 monotherapy also demonstrated objective responses and a consistent safety profile. Taken together, these results support that BNT111 is active in this difficult post-IO setting and provide us useful footing to guide setting selection and optimal combinations going forward.

Turning to iNeSt. The data presented at ESMO come from our randomized Phase II trial evaluating Autogene cevumeran in combination with pembrolizumab where is the pembrolizumab alone in first-line metastatic advanced melanoma. As previously disclosed, the trial did not meet the primary endpoint of a statistically significant improvement in progression-free survival.

That said, we observed a numerical trend favoring the combination and overall survival. In the combination arm 12 months, overall survival was 88% and 24 months. Overall survival was 74% compared to 71% and 63% in the pembrolizumab arm respectively.

Of note, crossover was allowed in patients randomized to pembrolizumab received a combination at progression. For the overall survival analysis, those patients remain in the originally assigned arm which can dilute for observed treatment effect over time. We observed robust neoantigen-specific T cell responses in the majority of evaluable patients with multi-epitope breast and persistence of T cell clones well beyond induction, indicating that the mRNA ERP is mediating the intended biologic activity that we want to achieve. The translational readouts give us three actionable insights.

First, T cell response breadth correlates with activity. Within the combination arm, patients who mounted a broader neoantigen-specific T cell response experienced longer progression-free survival, supporting our ongoing efforts to maximize antigen reps and to target early and low tumor burden disease with still proficient immune cell priming capacity.

Second, immune cell PD-L1 matters. We saw a trend of improved overall survival for the combination in tumors where immune cell PD-L1 was high while tumor cell PD-L1 did not discriminate overall survival in this data set supporting that low tumor cell PD-L1 should not exclude tumor types from vaccine PD-1 strategies.

Third, signal in IO insensitive biology. There was a trend of improved overall survival with the combination in tumor mutational burden, low patients, precisely, the population that typically gains less from IO. This is consistent with the concept that the vaccine can supply immunogenic targets when endogenous mutation load is limited and further encourages development in setting such as pancreatic cancer and MSS colorectal cancer with low tumor mutational burden and unresponsiveness to IO.

Altogether, these mechanistic insights support our ongoing randomized Phase II trials. Both the specific indications we have chosen, which is colorectal, pancreatic and bladder cancer as well as our focus on the adjuvant setting, where tumor burden and heterogeneity is lowest than T cell proficiency is still high.

Now looking ahead, what comes next. We will continue to generate and present new clinical data across our oncology pipeline. Data directly here late-stage decisions. For Pumitamig, we will share early data from our TNBC program in December, including from our dose optimization cohorts, which are central to defining the Phase III regimen.

From our ADC platform. We expect additional monotherapy updates from BNT324 in cervical cancer and platinum-resistant ovarian cancer from BNT325 in TNBC and from BNT326 6 in HER2 [indiscernible] and low hormone receptor positive breast cancer. These studies explore indications define dose and sequence guardrails and set the add-on benefit bar for Pumitamig Novel-Novel combinations. For the randomized Phase II trial evaluating [indiscernible] monotherapy treatment versus watchful waiting in adjuency TDNA positive stage 2 high-risk or stage III colorectal cancer, we expect an interim update in early 2026.

The efficacy evaluation of the primary endpoint of disease-free survival is projected for the end of 2026, when the data set will have reached the intended maturity. Then later this year, we plan to present data together with our partner on C4 from the nonregistrational first part of our ongoing global Phase III trial, evaluating our anti-CTLA-4 antibody [indiscernible] versus chemotherapy as a second-line treatment for squamous non-small cell lung cancer.

Overall, these upcoming data points advance the same theme. Evidence led prioritization by establishing dose finding and mono ADC baselines to further refine Pumitamig registrational path and leverage randomized setting specific readouts to position our mRNA immune therapies where they are most likely to succeed.

With that, I will now turn the presentation over to our CFO, Ramón Zapata for the financial update.

Thank you, Ozlem and a warm welcome to everyone who has joined today's call. I will begin by reviewing our financial results for the 3 months ended September 30, 2025. Note that all figures are in euros unless otherwise specified. The total revenues reported for the period were EUR 1.519 billion, an increase from the same quarter in 2024, which was EUR 1.245 billion. This increase was mainly driven by the recognition of USD 700 million as part of the BMS collaboration in the third quarter of 2025.

For context, in total, we expect to receive USD 3.5 billion in upfront and noncontingent cash payments from BMS and between 2025 and 2028. We expect to recognize this as revenue in increments annually over the development phase of Pumitamig. For the third quarter 2025, we reflected USD 700 million in our revenues.

Moving to cost of sales. This amounted to approximately EUR 148 million for the third quarter of 2025, compared to approximately EUR 179 million for the same period last year, driven by lower inventory write-downs. Research and development expenses were approximately EUR 565 million for the third quarter of 2025 compared to approximately EUR 550 million for the same period last year.

R&D expenses were mainly driven by the initiation of late-stage trials for our immunomodulators and ADC programs and partly offset by cost savings resulting from active portfolio management towards our priority programs. SG&A expenses amounted to approximately EUR 148 million in the third quarter of 2025 compared to EUR 150 million for the same period last year.

The decrease was mainly driven by lower external costs partially compensated by our ongoing commercial buildouts. Our other operating results amounted to approximately negative EUR 705 million in the third quarter of 2025. compared to approximately negative EUR 355 million for the same period last year.

Our other operating results for the third quarter of 2025 was primarily influenced by the settlement of a contractual dispute. For the third quarter of 2025, we reported a net loss of EUR 29 million compared to a net income of EUR 198 million. for the comparative prior year period. This was mainly driven by the effect of settlement disputes.

Our basic and diluted loss per share for the third quarter of 2025 was EUR 0.12, compared to basic earnings per share of EUR 0.82 and diluted earnings per share of EUR 0.81 for the comparative prior year period. At the end of the third quarter of 2025, our cash equivalents and security investments totaled EUR 16.7 billion, including the USD 1.5 billion upfront payment received from BMS. Our strong financial position empowers continued investments in our late-stage priority programs and preparations for commercialization of our diversified oncology portfolio.

Turning to the next slide. We are updating the company's financial guidance for the 2025 financial year. Our previously issued revenue guidance range for 2025 was EUR 1.7 billion to EUR 2.2 billion, and today, we are increasing it to EUR 2.6 billion to EUR 2.8 billion. This is mainly driven by the recognition of USD 700 million from our BMS collaboration.

Other guidance considerations such as those related to our COVID-19 vaccine business, including inventory write-downs from COVID-19 vaccine sales in Pfizer's territories as well as expected revenues from the pandemic pre-pended contract with the German government and revenues from our service businesses remain unchanged.

Turning to expenses. We are lowering our prior 2025 financial year R&D expense guidance by EUR 600 million to a new range of EUR 2 billion to EUR 2.2 billion. This updated guidance reflects our active portfolio management that has enabled significant R&D efficiencies. As part of that, we follow a rigorous go/no-go decision-making across all development stages as part of this prioritization efforts. This allows us to focus on the programs in our portfolio, which we believe represents the largest opportunities. Consistent with our commitment to disciplined and sustainable growth, we are also improving our full year guidance for SG&A and capital expenditure for operating activities.

We are reducing our full year G&A expense guidance by EUR 100 million to a range of EUR 550 million to EUR 650 million as a result of ongoing cost optimization initiatives. We are also reducing our full year guidance for capital expenditures for operating activities to a range of EUR 200 million to EUR 250 million to better reflect our targeted investments in manufacturing. Aligned with our disclosures earlier in the year, we expect to report a loss for the 2025 financial year as we continue to invest in our transition to become a fully integrated commercial oncology company.

As Ugur outlined, we continue to focus on executing our strategy around [indiscernible] tumor product opportunities, Pumitamig and our mRNA cancer immunotherapies. We currently have multiple ongoing Phase II and III trials across these programs reflecting our strategy to bring novel combinations to patients, we expect to generate additional meaningful data for these programs in the months ahead. As we advance, we will continue to maintain rigorous financial discipline and remain focused on achieving long-term sustainable growth.

Before concluding, I would like to invite you to watch our Annual Innovation Series R&D Day event on November 11. During the R&D Day, we plan to provide a deeper dive into our oncology strategy, including plans for Pumitamig and our mRNA immunotherapy candidates. Thank you for your ongoing support and interest as we continue to create value for cancer patients, society and shareholders.

With that, we would like to open the floor for questions.

[Operator Instructions] We will now take our first question. From the line of Tazeen Ahmad from Bank of America Securities.

I wanted to get a sense about how you're thinking about the market opportunity for MSS CRC and first-line gastric cancer. Can you just talk about how your product can be particularly differentiated from what's currently used?

Thank you for the question. We lost your audio there a little bit. Could you just -- sorry, could you just repeat that question? I just want to make sure we get correct.

I wanted to ask a question about the market opportunity for MSS CRC and for first-line gastric. I wanted to get a sense of how you think about the opportunity relative to the competition.

Thank you, Tazeen. We got it at that time. So that was a question about how we think about the CRC first-line opportunity in gastric and how it compares to the competitive field. So Ozlem, would you like to take that question?

Yes, I can take a question. Both indications as CRC and gastric first line still high medical need indications. And we think that the combination of VEGF-A and PD-L1 blocking from a biology point of view, has a rationale for development and has the potential of improving the clinical benefit for these patient populations.

We will now take the next question from the line of Terence Flynn from Morgan Stanley.

I had one question and then one just clarification. So for BNT323, was just wondering if you can share any more color on the delay in the BLA filing in terms of the gating factor here? And then on the new R&D guidance, just want to clarify that, that reflects the assumption of some of the 327 expenses by Bristol-Myers and that, that was the driver of the change here if there's other prioritizations that fed into this?

Yes. Okay. Thank you, Terence. So two clarifications in there. So maybe if we do the R&D guidance first, and I'll direct that one to Ramon. And then Ozlem, I'll direct the 323 BLA progress question to you after that.

Thank you for the question, Terence. I would say that the lower guidance on R&D is not about reducing spending on 327. We are updating this guidance to reflect the lower expenses for the year. The reduction is mainly driven by the phasing of certain programs and a deliberate focus on our key strategic priorities, meaning 327 as you rightly mentioned.

We demonstrate disciplined portfolio management, but I would say it's too early to say whether this represents a structural shift. Depending on the pace of our late-stage programs, including the expanded efforts on Pumitamig, R&D spending will remain at similar levels or increase again next year. I think what really matters is that we continue to allocate resources with focus and flexibility to maximize long-term value and support our key strategic priorities and programs.

And Ozlem, would you like to take...

The reason why we -- originally, we guided towards end of '25 for 323 BLA submission. This moves now into '26 because we have continued discussions and conversations with the FDA to further understand additional data needs and are generating this information. The plan is still to submit in '26. And in '26 6, we will also get for this program data from our ongoing breast cancer study.

We will now take the next question from the line of Daina Graybosch from Leerink Partners.

Thank you for the question. I have a question on the overall strategy with Pumitamig of established and Elevate as two steps. And why you're taking that approach versus in some indications doing them simultaneously let's say, in multi-arm Phase III studies with ADC combos and Pumitamig on top of traditional standard of care chemo to leapfrog particularly where you have some early data with the ADC in indication and the competition is fierce. Thank you.

Thank you, Daina, for that question. So that's a question about our strategy for vomiting and the various stages, the various steps to our strategy with established and Elevate. So I'll direct that question to Ozlem.

Thank you, Daina, for the question. You are actually right. We have this 3-wave strategy, establish expand, elevate. And even though we call it free base these activities which are going on in parallel. We have a certain focus on the chemo combination or combinations with standard of care because these studies can we simply started much faster, and we have a focus on speed to be really first to market in certain indications. However, there is data generation in combination studies ongoing in these indications with our ADCs, for example, and will come very soon also following this established ways.

We will now take the next question from the line of Asad Haider from Goldman Sachs.

This is Nick Jennings on for Asad and the Goldman team. Given that the BNT327 Phase III trial in triple-negative breast cancer is initiating this year, could you provide any insight as to what we can expect to see in the Phase II details coming up at SABCS. And is there any new information we can expect that provides additional confidence in the Phase III success?

Thank you, Nick, for that question. It's a good one. So just to recap that, front, the Phase III triple-negative breast cancer, which is initiating and Ozlem, the specific question is whether we can provide any additional details on the Phase II results that we'll be presenting SADCF.

So we will present some more efficacy data, safety data and also dose data.

We will now take the next question from the line of Akash Tewari from Jefferies.

This is Manoj for Akash. Just one question. So we recently saw Harmony trial in first line and the CLC making some changes to look at primary PFS and [indiscernible] analysis separately for comonomer populations. So considering these changes, do you still think ROSETTA-02 trial in BNT327+ chemo is sufficiently powered for PFS and OS endpoints in the Phase III portion. Will there be any trial change, any trial-design changes based on these new information?

Yes. Thank you for that question. So it's a little hard to hear some of the details on that, but I heard you talking about Harmony 3 and whether that may have any read-through or effect on the way that we're conducting our trials for Pumitamig. So I'll direct that question to Ozlem.

Yes, we are constantly with upcoming new data, reevaluating our statistical analysis plan for ongoing trials, and we'll also look into this specific trial.

We will now take the next question from the line of Yaron Werber from TD Cowen.

Great. And I had a quick follow-up for Ozlem on 323. Just the need to generate more data to support filing, can you be maybe a little bit more explicit? Do you need to generate -- it sounds like you're going to have more data, as you noted, in breast cancer next year. And so is the thought then file for breast cancer next year. And what was the feedback for endometrial cancer? And do you still plan to file for that? Or maybe just give us better clarity.

Yes, maybe I was misleading for the endometrial cancer discussions with FDA, have nothing to do with ongoing breast cancer study. It's not about generating new data. It's about follow-up data and [indiscernible]. So that pushes the time line a bit into '26, but does not change our submission strategy and our plans for 323 overall.

Okay. And that's for breast cancer. And then what about endometrial cancer? What's the plan there?

No, no, no. Endometrial cancer is our first submission. This is what we said all along. Originally, it was spent for '25. We -- this is pushed out to '26 because, as I said, we are in discussions with -- in BLA discussions with the FDA and providing further data breast cancer, the breast cancer study, Phase III study is ongoing, will read out later in 2026.

We will now take the next question from the line of Mohit Bansal from Wells Fargo.

So again, a question on VEGF PD-1. One key comment we get from KOLs or experts is that with these bispecifics, it does look like that they are better VEGF inhibitors, but it doesn't look like that the PD-1 is -- the component is better.

So I mean, how do you think about that? And in the context of these -- this bispecific showing an OS benefit in lung cancer trials, how important it is for PD-1 to be better at this point, given that -- we are seeing good PFS benefit, but OS is kind of on border line. So I would like to get your thoughts on that.

Thank you, Mohit. So a question generally around how much confidence we or others have in the bispecific class. And you mentioned that VEGF binding is maybe better, but PD-1, you're saying maybe not as good in bispecifics. And specifically, as benefit in lung. So direct that question to -- Ozlem?

Yes. Let's start with our confidence. Our confidence is increasing into this drug class. And the confidence is not based on better VEGF better PD-L1, but what the antibody as a bispecific molecule and we are seeing now that this is getting more and more clinical data that this is not only called on PFS, but also have an impact in OS. And maybe, Ozlem, if you would like to add mechanistic understanding how this could also be helpful.

Yes. Mechanistically, in principle, our preclinical data, and that was also part of develop of preclinical development and selection process for this antibody shows that blocking of PD-1, PD-L1 pathway, as well as the GSA blocking in the respective preclinical settings is robust and it's not inferior to what you would see with the individual antibodies -- having said that, we also think that we have a PD-L1, not a PD-1 arm here as an additional element to the mode of action, namely targeting of this molecule into the tumor micro environment.

And this, again, is a very good condition to amplify both on the PD-1, PD-L1 side, but also on the VGF receptor signaling aside all the effects on canonical and noneconomical effects of these two targets. So this is the preclinical piece and mode of activities, but the clinical data has to have a true from the data we have across tumor indications. This is not a Phase III data. We are very confident that the activity has PFS effect in certain in certain indications and also duration of progression-free survival starts to look good.

We will now take the next question from the line of [ Havan Desa ] from BMO.

This is actually [indiscernible] for Evan from BMO. Thinking about the guidance range for this quarter, could you quantify how much of this reflects the relatively stronger quarter for COVID versus just general updates for the BMS collaboration and U.K. government agreements. I know you mentioned most of this was tied to the collaboration, but I was curious if there were any minor changes on the go front would be helpful to think about the relative contributions there. I appreciate it.

Thank you, Malcolm. So let us talk a little bit about the revenues. And I will refer to your commentate question, but I also think it would be helpful for the audience to understand that bit of the BMS revenue. So [indiscernible] -- so for [indiscernible], we continue to see a stable position with a strong market share and stable pricing. U.S. oxination rates are roughly 20%, which is in line with what we had anticipated. We have always assumed lower volumes versus last year.

So overall, the business is performing within expectations for the year. While the broader market remains uncertain, we continue to lean on our strengths like strong brand recognition, reliable supply and rapid adaptation, and we do expect to close the year in line with our outlook.

Now if we talk about the BMS revenues, the updated revenue guidance mainly reflects the collaboration with BMS, as you rightly point out. And under this agreement, we will receive a total of USD 3.5 billion in upfront and on continuing cash payments between 2025 and 2028. While the timing of cash inflows and revenue recognition deferred revenues will be recognized in broadly equal amounts over the next 3 years, with the remaining balance recognized together with a final payment in 2028. This will provide a clear and predictable contribution over the next several years.

We will now take the next question from the line of Joshua Casado from Evercore ISI.

This is Josh on for Cory Kasimov. On your and your partner's decision to push Pumitamig into gastric cancer, did you see compelling clinical data, not sure if this is presented or not? Or is this push into this new indication based off your understanding of the [indiscernible]

Thanks, Josh, for that question. So it was a question about Pumitamig and our announced decision to move into gastric cancer, what was that based on? Have we seen any data that we can speak to that support that decision. So [indiscernible], would you like to take that question?

The emerging data for than industry cancer and as an indication, which -- for which checkpoint blockade is approved. It's an indication that we have seen responses in combination with tumor therapy and an indication that we see based on the data that we've got in other GI indications. A clear room for improving over standard of care.

And also the mechanistic question are that entire angiogenic and PD-1 targeting approaches are validated approaches in gastric.

We will now take the final question from the line of Jay Olson from Oppenheimer.

We're curious about your collaboration with Bristol-Myers Squibb. And can you talk about the governance structure and which party makes the decisions for new trials and who leads the new clinical trials when you initiate them?

Yes. Okay. Thank you, Jay. Thanks for that question. It's an interesting one about how our collaboration with BMS works [indiscernible] mechanic [indiscernible]. I can't say that word. So [ Ozlem ], I'll pass it over to you, who makes decisions for [indiscernible], who makes decisions on clinical development.

But it's a tactical approach with multiple collaborative arms, the FSC in which we discussed all the indications so far all indicate all decisions that are made are based from interest of both partners, but both partners have the opportunity to do combination cars with their products. Yes, we got that the other partner is interested to join [indiscernible] or not, so we have a lot of flexibility in this collaboration aiming really to do all kind of studies and to exploit the pipeline of the other partner as exhausted as possible.

This concludes today's conference call. Thank you for participating. You may now disconnect.

Hi. Good afternoon, everybody. So welcome to all of you in London in the beautiful venue of the Science Museum, and to everyone joining us by webcast to our second AI Day. And we're really excited today to show you some interesting new things. But before that, the formalities.

So here are our forward-looking statements, which you can see. And you can refer to those in the presentation, if you'd like more details. We do not commit to updating these, and these are current as of today.

So let's have a look at the agenda that you can see here. So first of all, we will have some upfront sections from Ugur and from Karim that will talk about how AI is fully integrated into development and our entire business model in BioNTech. And then we'll move on to some InstaDeep examples where you're going to see lab-validated results and their applications. So we're really excited, and that will show an evolution from what you saw last year.

So now I'd like to introduce to the stage, Ugur Sahin, our CEO of BioNTech, and he's going to present the opening presentation on advancing a disruptive tech-bio company.

Yes. Thank you, Michael. Thanks, everyone. I would like to welcome everyone also on behalf of Karim. And I would like to give you the scientific biological background, why we need AI and for what we are using AI.

But first of all, it's a really great pleasure to be here in this place. Actually, this place here, the Science Museum, is the place where the first COVID-19 vaccines, the empty vials, I saw it here. So on December 8, Margaret Keenan was the first person on the planet who received an approved COVID-19 vaccine, and this vial is straight here, close to the lancet from Edward Jenner, who introduced the very first vaccine study worldwide.

But today, it's about AI. And this is just showing you that our AI approach is not limited to London. It's a global approach. We have sites where we do AI on multiple continents. And actually, we have been doing AI and 2019 was the first time that we met Karim and started to work with InstaDeep, but we are -- we were doing AI before that, but we didn't call that AI. It was machine learning, yes. And what is really new was InstaDeep coming in. While we were developing our tools based on existing technologies, I think what we can say about InstaDeep is really research on developing new technologies, completely new technology. And therefore, we are not only doing research and development in pharmaceuticals but also research and development for our AI tools.

So a few words about BioNTech. We are a clinical -- late-stage clinical company with multiple programs in oncology, which is our core focus, but we have also both pipeline in infectious diseases, particularly for diseases with high medical need, for example, TB, malaria, HIV and others.

And AI is in the meantime, really fully integrated into BioNTech. There is -- there are -- there is -- I would say, there are only a few projects where we don't use AI approaches. And the most important is that we are continuing, continuing to improve our methods and technologies with AI.

I give you a little bit of background that you understand what we are doing and how AI is connected. Our core focus is oncology. And oncology is making a lot of progress in the last 20 years. And one of the breakthroughs in oncology was immunotherapy, the use of the patient's immune system to fight cancer. And there were a number of breakthroughs that we started in improved survival of patients, but still there's a huge medical need. More recently, we've seen new treatments based on ADCs and bispecific antibodies. And of course, we believe in the future of messenger RNA therapeutics immunotherapies that could provide us additional benefit for eliminating tumor cells.

But let's start with immune modulators. So the classical immune modulator, which is the most widely used category is anti-PD-1 antibodies. An example is nivolumab or pembrolizumab, which are used -- have been used in hundred thousands of patients.

And we were developing in the last years bispecific antibodies because we were interested to increase the fraction of patients who can respond to bispecific antibodies. And one of the molecules that came in from a partnership with a Chinese company, Biotheus, is BNT327, Pumitamig. And this is a highly interesting molecule. It combines 2 mode of actions. It's anti-PD-1, which releases immune cells, which are inhibited by the tumor cells and enable immune cells to act and cure tumor cells. And on the other side, it inhibits the generation of vasculature based on a mechanism, which is blocking VEGF, and it has a number of additional magic tricks, which result in immune responses and objective responses in cancer patients across multiple cancer entities.

And this is really exciting development in the whole field. We call this the bispecific anti-PD-L1/VEGF class, which is expected not only to reach the tumor types that are currently used, currently address its entire PD-1 treatments on the left side, but also can go into categories, cancer indications where anti-PD-1 treatments are not approved yet.

And we realized in the last 18 months, this is really a big, big, big opportunity, and it's too big to do it alone. Therefore, we decided to go into a partnership and announced a few months ago a partnership with Bristol Myers Squibb.

It's a global partnership to develop this class of antibody, Pumitamig, in multiple cancer indications. We have data in the meantime from more than 1,000 patients in more than 10 indications, giving us the direction which indications could benefit from that. And this will not only be monotherapies, but also combination therapies.

So we believe that these types of treatments can help us to control tumors and, in some patients, also provide a lasting clinical benefit, ideally cures. But cancer is very complicated. And most of the patients who have initial control progress over time. And the reason is here depicted in a simple cartoon. So cancer is evolving from healthy cells by DNA mutations, and these DNA mutations are just accumulating over 5 to 20 years.

And so that means during the accumulation, the tumor cells because these are all random mutations, generate a heterogeneity and an individual cancer by this reason is really individual, no 2 patients share the same type of mutations. But the bigger problem is even that we have a intratumoral heterogeneity that means every tumor cell carries another set of mutations, which means that we have a situation where cancer can evolve over time and it's an evolvement not only against the treatment, but also an evolution against the immune system.

And so knowing that from the very beginning of the 1990s where tumor immunology really become molecular, we were interested in cancer vaccines, because cancer vaccines come up with the promise that we might be able to induce immune responses against multiple epitopes.

So if the tumor is polyclonal, the idea is here to induce a polyclonal T cell response that go into different directions so that we can combine multiple antigens. And this polyspecific activity is expected to move the last million tumor cells that remain after treatment, for example, with checkpoint blockade. So we have developed and pioneered several approaches to that. Two of them are shown here.

One is aiming really to completely individualize the treatment. That means it's based on identification of the mutations in individual cancer patients by sequencing. And then these mutations because they are recognized by immune response are called neoantigens. And then we assemble a vaccine, which is tailored to these mutations. So the first description of this approach was in 2011. We showed the preclinical approach. And in the meantime, we have multiple clinical trials running in various indications, including pancreatic cancer.

The second approach uses a complementary concept. That means if we say certain tumors tend to have certain type of tumor antigens. So we identified them. And for example, in melanoma, we have identified 4 antigens, which cover 100% of patients with melanoma in lung cancer. We have identified 8 antigens covering more than 90% of patients. So it's a combination vaccine approach where the vaccine is off the shelf and then used directly for vaccination. So there is no need to generate de novo.

So this is the approach shown in more detail, taking the individual sample from the patient and mapping the mutations by comparing with normal the neoantigen prediction, which is computationally done, then on-demand mRNA manufacturing and transportation to the patient. And of course, this all is driven by data and algorithms. And the manufacturing is just in time. We can deliver the vaccine in less than 8 weeks, and our aim is to be able to deliver a vaccine in less than 4 weeks.

And this shows you an example how this works. So the sequencing of the tumor can yield up to thousands of mutations. In some tumors, only dozens of mutations. So there are tumors with high number of mutations, and there are tumors with lower number. And we do the computational ranking of these mutations based on the idea that these epitopes can bind to the human leukocyte antigens, which are presenting this.

And the way how they bind is defined by patterns, binding patterns. And this can be computationally calculated. And on this calculation, many other features, for example, how much the mRNA encoding these mutations is expressed in the tumor and whether we expect heterogeneity of the gene.

So whether there is frequencies or the fraction of tumor cells in the tumor because of the heterogeneity, it is quite possible that you target the mutations, which is only in 75% of tumor cells, so you are inducing a selection of 25% of tumor cells.

So this is our ranking algorithm. And this is based on computational approaches. We used the classical approaches. And in the meantime, we worked with InstaDeep to develop deep learning approaches, including a number of additional aspects of the antigens, for example, in which cell compartment they are expressed, which type of molecular patterns they have? So it's multiple additional features, which come in.

I would like to give you here an example of, at the end of the day, the terminal biological mechanism that we need for this type of approaches. It's about killing of the tumor cells by lymphocytes. And this is an approach in biology in nature, which is among the most deeply quality-controlled biological events. Because at the end of the day, it is about a cell in the body killing another cell. It needs to be authorized. And the authorization is really done by a complex process, and this complex process is also misused by cancer cells to avoid killing by using mechanism and that circumvent that.

But if you go deeper into that, one of the key aspects here is the recognition of the tumor cell by T cells. And this is happening by the T cell receptor. So that means these T cells have T cell receptors. And every T cell has a different T cell receptor. There is a complexity here. So our biology in humans allow that we have around 500 million different T cell receptors in our body. So there is a huge library of T cells that can recognize something.

Then we have the HLA molecule that is on the tumor side, which is presenting an antigen which is inside the cell. So the mechanism is here actually evolved to recognize hidden viruses in our body cells. So it's an immune attack against infections. And our HLA diversity in humans have been evolved that viruses don't circumvent the killing by just avoiding the patterns that are presented on individual HLAs. But this makes it extremely difficult.

And then you have the peptide, which is a fragment of 8 to up to 15, 16 amino acids depending on the HLA class. And we have now extremely complex interaction with the T cell receptor. Actually, these have 2 chains. We have the HLA, and we have the peptide, and only if the combination of all 3 works, we get a killing. And the problem is even bigger because the T cells have to avoid that this interaction does not kill any other cells. So it's also quality control. So -- and this is one of the biggest challenges in AI, to identify the T cell receptors that can recognize an MHC peptide sequence. So this is part of our prediction algorithm. We know the HLA of the patient, we know the mutated epitope and we would like to understand whether there is an immune response.

And nature is doing that extremely well. So the COVID-19 pandemic showed us that 2 different COVID-19-infected patients could develop the same T cell receptors when they have the same HLA. So that means this discovery of hundreds of millions of T cell receptors works very well.

The question is -- we have 2 types of questions. Can we identify T cell receptors that are recognizing tumor antigens? And the second question, of course, is, can we do it better than nature does? And we will have, I think, 2 talks about this, about these problems and how AI is used there. So this is complex and computation is important. But the situation is even more complex in cancer.

So we have several levels of diversity and heterogeneity in cancer. So this is here the cancer heterogeneity and the clonal evolution. And on the other side, we have the immune system, and the immune system can also evolve. So I compare that sometimes like a Go game. So the immune system is playing against cancer, but the situation is even more complex because other factors are also playing a role, HLA molecules, the microbiome of the patient, environmental factors and so on.

So the question is, are we going to understand -- if this is a game, yes, are we going to understand the moves of the cancer cells by reading this? And we believe into that. We believe that if we feed enough data into AI and if we really bring in the biology that we can make the evolution of a cancer cell predictable. And if we know if it's predictable, we can interfere with multi-specific approaches.

So we are translating that into a collaborative approach. On the one side, we are generating data from our personalized vaccine studies, preclinical and clinical studies. And on the other side, InstaDeep and colleagues are developing new tools like DeepChain. We are using these tools, of course, not only for optimization of cancer vaccines, but also optimization of proteins, mRNA structures and so on. And Karim will talk about our in-house supercomputing clusters and how this contributes to obtaining better results.

So in summary, this slide shows our vision. And I think this is more than our vision. It's really the view of our future fully integrated AI tech company, which combines a number of capabilities. Our vision in future is that we can take clinical samples, do a personalized omics, understand what is going to happen, and use our pipeline of molecules to come up with a combination treatment, which is consisting of off-the-shelf drugs, for example, Pumitamig, or our ADC, plus a personalized vaccine. So this is, in principle, we have shown that this is doable, but we need to do it at scale, and we need to do it in an affordable manner.

So I think this is a good introduction for Karim, who now will come and introduce the capabilities that InstaDeep has built.

Thank you so much, Ugur, for this exciting presentation. So I'm Karim Beguir, I'm the Co-Founder and CEO of InstaDeep, the AI unit of BioNTech Group. And I mean, it's a very exciting time to be building in AI, and there is a lot to talk about. But very briefly, I will try to give you a sense of our approach and the opportunities that we are developing and how we work collaboratively with our BioNTech colleagues to get things done with results as well.

So I mean, AI, I've been working in the field for more than 10 years. And if you were to summarize what's been happening since the beginning of the deep learning revolution in 2012, really it is a triple exponential. So you have data growing exponentially, I think, for example, about the cost of like whole genome sequencing, which is now around $100, which is absolutely insane. So massive exponential growth of data, but it's also compute and finally, model innovation.

So the compute side of things, I mean, it's pretty incredible how predictable things are. I mean this is for me one of like the most impressive sort of like graphs in the history of like computing and machine learning more recently. But if you look at since the 1940s, since everything started with the first computers like Colossus, ENIAC and the other. And despite technology changing so much, I mean, at the time, it was vacuum tubes, now we're like in chips and semiconductors, in the future, it may be quantum computing, yet everything is so sort of like linear in log space. So here, the y-axis is actually sort of like increasing by a factor 10 at every point.

So you see like how this exponential keeps going and sort of like this is also famous as Moore's Law, which in general for computing is like compute efficiency for a given budget, you get twice as much compute every roughly 2 years. But in AI, it's actually Moore's Law on steroids, like the amount of compute, which is deployed in ML workflows doubles every 4, 5 months. So this is actually pretty insane. But it doesn't stop there.

And the third point is model innovation because, yes, we have more and more data. Yes, we have more and more compute or at least more affordable compute. But the third point and perhaps the least understood is the efficiency of the models themselves. In 2012, in the beginning of the deep learning revolution, you needed to label literally like millions of data points to get an algorithm to learn. You don't need to do that anymore. With self-supervised learning and recent progress, you can literally feed the entire Internet or entire databases and get the system to learn. And incredibly, actually, the efficiency of those models improves every 8 months by a factor 2.

What do I mean by that is, if you want to get to a certain level of performance, every 8 months, you need just half the compute to get there. So not only do we have a crazy amount of compute coming in and becoming available, a crazy amount of data, but the efficiency of the models is absolutely incredible. So it is a triple exponential. But yet, if you follow what's been going on in terms of like progress and the like, it does feel that progress is almost vertical.

And sometimes one wonder like, is this hype? Is this a bubble? Or is it true that progress in AI is going to be incredible in coming years? And I would tell you like having been training neural nets for more than 20 years, I think that this is real, and we're going to see incredible progress in coming years.

And actually, there's something qualitatively different that is happening that didn't happen before. And what is that? Well, it is that AI is so competent that is actually now accelerating itself.

So if you think of AI as a plane with 3 engines, and these engines like we saw our data, compute and model innovation, AI is so competent now that it is at the point where it's going to accelerate every one of these drivers and hence, push progress much faster and deeper.

And so this new era that Richard Sutton, one of the godfathers of AI named this year as the era of experience is taking us to new heights and allowing progress in a way that was simply impossible before. And here, I want to show you a little bit like what happened since the beginning of the deep learning revolution.

In the very early days here, you had progress, which was coming from games, like you had with technologies such as reinforcement learning, you had a score and you had a system that would learn by maximizing this score. And then we had the big ChatGPT moment, and this is like became the golden age of large language models around 2022.

But then again, we are coming back to a time now where systems based on trial and error are coming back to become very effective. And the reason is, AI is so competent now that what used to be possible only for games becomes possible for a larger class of problems. And so systems now can actually improve data. They can create synthetic data for the problem at hand and using reinforcement learning, improve their own answer to those questions.

So this is what has been driving, for example, like progress in reasoning systems. You could see it also, for example, in a lab to optimize a certain protein sequence, like lab results are understood by AI that's going to use those to further improve its answer.

So effectively, we're not limited by data anymore. We still need a lot of data, but systems now can generate synthetic data and take advantage of it and progress. So AI now is driving data, which is the first engine.

The second engine like we said, is compute. Well, today, I don't know if you noticed, but NVIDIA keeps coming with these GPUs much faster than before, roughly now a new generation every year. This is because those hardware systems are actually co-designed with AI. It's not only very clever engineers. It's also AI systems using technologies such as reinforcement learning that are really accelerating progress in hardware. This is the same for Google with its TPU v7 chips and other that are codesigned by AI. So you see how AI is boosting its own hardware.

And finally, model innovation itself, one of the things which is the most amazing about this period is that we actually reached the gold medal at the International Math Olympiads (sic) [ International Math Olympiad ], at the International Olympiads of Informatics (sic) [ International Olympiad in Informatics ] Computer Science this summer.

To give you an idea, only a few years ago, like 4 years ago in 2021, we thought it would take another 20 years to get to these results. And these were the experts in the field. And yet we are there. And again, technologies such as reinforcement learning are key here because you have a score. You can actually evaluate a chain of reasoning or code. And this means that now AI systems can conduct machine learning research or more precisely work hand-in-hand with experts to push progress forward. So really like this is accelerating faster. And again, technologies like reinforcement learning, becoming more and more important.

So at InstaDeep, we have anticipated those trends for a long time. And I'm very proud to say that we've been active in reinforcement learning research, in particular, for many years, and this work is coming to fruition. So today, I'm very happy to let you know that at the next NeurIPS Conference, which will take place in December 2025, NeurIPS is the largest and most influential machine learning and AI conference in the world, we actually have multiple research RL papers accepted, including for the first time in our history, an Oral excerpt on top of a Spotlight excerpt. So really, congratulations for the team for pushing the envelope in terms of algorithmic innovation in RL. It's really exciting.

But it doesn't stop here. And what's exciting also is that we've had the most productive 12 months in InstaDeep history. If we look at where we were at the last AI Day to this day, we've had actually 6 Nature journal publications on biology and AI. And for me, this is a testimony of the quality of the innovation that is taking place between BioNTech and InstaDeep. This is collaborative work between our different teams and really like super, super exciting, including for the first time in our history, a cover of Nature Machine Intelligence in June.

So as you can see, we've been having fun in R&D and innovation. But in reality, you need a lot more than that to win in AI today. And like Ugur mentioned, you need a full integrated approach. We need to be competitive at every step in the process. And so what does it mean?

If you think about what I was saying earlier about the 3 engines of AI or the 3 engines of the AI plane, where you kind of see them, it means being excellent in compute and model scaling, you need to train those models at incredible large size. We're talking about like hundreds of billions of parameters, trillion levels of parameters. But you need also the AI innovation, model innovation, which we discussed. And you need also to have a great data strategy and an ability also to use AI to accelerate your data acquisition.

And if you do these 3 things, then you get to exciting applications. And this is exactly what we're going to cover in this order and starting with compute and model scaling with Alex, who is going to present our latest results.

Hi, everyone. It's a pleasure to be here. So I'm Alex. I'm the Head of AI Research at InstaDeep. And echoing what Karim said, we had indeed a very interesting summer, a very exciting one. So we have seen that AI has made the headlines with major accomplishment being, achieving gold medal at the International Math Olympiad, winning programming contests with generally capable models and even now stepping into the real world, right, with advances in robotics and physical intelligence.

And I would say, in my view, these are not isolated milestones. These are predictable outcomes of the scaling laws, right? So the scaling laws, let's say, it's an empirical law stating that the performance of modern AI system is a predictable function of the resources spent to train such system being data, time, compute, memory and so on.

And so it's not only a single scaling laws nowadays, actually. It's the pretraining scaling laws as we know, but it's also post-training, which is really enabling agent to actually interact with an environment being simulated or real and learning from these interactions to accomplish even greater task.

We also have the inference time, the test-time inference scaling laws, which state that you can spend more compute to refine and polish the results of what the AI system will produce.

So the question for us is, okay, as a company, how do we position ourselves there to perform in this new environment? And the philosophy of InstaDeep has always been to build an integrated AI ecosystem, starting from the hardware going to the orchestration and software. Because it's only through -- this is our belief, it's only through a tight hardware, software integration that we can gain the performances, the cost efficiency and the control required to achieve our objective.

So how does it work in practice? Last year, it's for this reason that we had the pleasure to announce that we built the Kyber cluster, which is a AI supercomputer made of NVIDIA H100. This contained 14 of these racks, which have been engineered in-house by our bare metal team to optimize the performances for our own AI workflow, for example, scaling large language models training or running simulation for RL training, like Karim mentioned. It brings our total compute capacity to 500 petaFLOPs and is now our major source of compute power for the company.

Now that we have this hardware, which is critical for work, we need to make it very easily accessible by all engineers to empower their work. And so there also, we built our own product, our own platform, which is called AIchor. It's a full product available to our customer, and really enable us to run very seamlessly experiments on Kyber. So with just a GitHub process, git commit, we can run experiment very easily. And that's why our engineers, around 200, 300 of our engineers have actually submitted more than 15,000 experiments a month in average in 2025.

We also keep our GPUs and hardware very buzzy, where we maintain a very high usage of 75% of GPU usage on our cluster.

The next step in building on top of that is obviously the ML software stack that we have to design to, let's say, squeeze the most performance out of each hardware accelerator. And that's why we've been building an entire ML ecosystem that is meant to be very efficient, scalable, modular, such that we can answer the requirement of the research development in which we operate.

So let me give you 2 examples of this in action. The first one is about scaling large language models. So LLMs are part of our daily life and sometimes we train them, quite often actually. And so here, we took the challenge of trying to scale our Nucleotide Transformer models, which we published in Nature Method later last year to a 100 billion parameters model.

The first challenge here is that 100 billion parameters model does not hold on a single device, on a single GPU. So the first thing we have to solve is actually how do we distribute it, how do we shard it across GPUs?

And our answer here is to use fully sharded data parallelism within a single DGX, within 8 GPUs and then horizontally scale that across all the racks of Kyber here, I just depicted 3, but we have more than that. Horizontally scale that using data parallel. If we were to grow a model even more, we could use Tensor parallelism, we could use pipeline parallelism or even like sequence parallelism, if we want to handle a very, very long context length.

So that's the first point. The second one, at the code level, we have to do a lot of optimization as well. We can use advanced CUDA kernels like Flash Attention. We can do mixed precision and quantization or we can try to optimize the XLA compiler and use better network configuration and so on and so forth.

The result is a staggering 66% Model FLOPs Utilization. So just the definition means that basically, we maintain our hardware buzzy, 66% of the theoretical limits of the hardware, right?

And so just to give you a reference point, the large public run of Llama 3.1, which contain like 400 billion parameters was around -- the MFU was around 40%. Of course, it's a much larger network. It's run on thousands of GPUs, which will give you a sense of the meaning of that number and how high it is. And we're actually going to talk a lot more about the foundation model we built in the next section in a few minutes about our AI innovation.

The second example I want to give you is about scientific computing. Traditionally, when scientists are trying to discover and look for a new molecule, a new drug, chemicals, materials, well, they start with thousands of candidates. But realistically, only a few of them can be tested in the wet lab, right? So scientists face not only this kind of discovery problem, but this smart selection problem, right?

The problem is that if you choose the wrong candidate among the many potential, well, you waste time and resources. So how can we do a smart selection here? And lucky for us, let's say, most of this property can actually be accurately estimated using quantum chemistry. The problem here is that quantum chemistry is extremely low -- extremely slow, sorry. It's accurate, but slow.

On the other hand of the spectrum, you have classical force fields that are extremely fast, but really prone to errors. So how do we handle this? Our objective has been trying to combine the best of both world, the quantum level accuracy, but order of magnitude faster.

And our answer to achieve that is MLIP, machine learning interatomic potential. These are a class of machine learning models that are trained on quantum chemistry data, so very accurate data, but that run much faster. And the result is indeed very impressive. In terms of accuracy, we see that there's a near perfect correlation between MLIP and the reference DFT calculation, energy level, whereas classical force field are prone to error as you can see. So it's accurate one. But second, it's also much, much faster and cheaper, actually up to 10,000x cheaper. For any dollar spent on MLIP, you have to spend more than $10,000 of classic DFT calculation. So it's usually a huge improvement.

And in addition to that, as opposed to classic quantum chemistry method that don't scale very well, MLIP does. You can run simulation on tens of thousands, if not hundreds of thousands of atoms very efficiently. So we are very excited about this technology. It's the early days, but we're excited about the potential because of its application to so many different domains and application of interest for us. So I invite you to keep a look on that and we have a booth downstairs about it.

So that, I hope, gives you a sense of what we've been doing at InstaDeep in terms of developing your AI stack going from the hardware level with Kyber, the orchestration level with AIchor in our product there and the machine learning software stack.

Now I want to give some space for Bernardo, who is going to describe how we've been using this stack to develop the next generation of foundation models in genomics. Bernardo.

Thank you. Thank you, Alex, and well, hello, everyone. It's a pleasure to be here. I'm Bernardo, Senior Research Scientist at InstaDeep. It's my pleasure to present our work on AI applied to genomics. So genomics is a study of our genome, our genes and how they play together in our cells. And I want to show you how we are using AI to understand that.

So the first thing we did last year was to publish our first model, our foundation model for genomics called Nucleotide Transformer. And since then, it has become one of the most popular genomics AI models in the field and used in many papers and many -- and to develop many new models. On our side, we have used Nucleotide Transformer, here on the left, to develop new iterations or kind of fine-tuned versions for different applications that were published over this year where we used Nucleotide Transformer to annotate the genome at single nucleotide resolution with SegmentNT, so the second model.

We built Isoformer that combines DNA, RNA and proteins to perform at different tasks. And we even combined Nucleotide Transformer with a conversational agent which made the cover of Nature Machine Intelligence with ChatNT. So all this is published, but we are already working beyond this.

So if I put into perspective the current models that exist in the field, nowadays, we have models that learn from genomes. So Nucleotide Transformer is an example, Evo as well. So they are just trained on genomes and then fine-tuned on different downstream tasks. And on the other side, we have models that learned from functional data. So Borzoi or AlphaGenome from Google. So today, we are very proud to announce the release of our new Nucleotide Transformer version of NTv3, where we try to unify both paradigms into a single model that learns from genomes and in this case, more than 150,000 species genomes, but also at the same time, it's post-trained on thousands of functional data from many different experiments across different organisms.

So what is NTv3? So NTv3 combines a full set of capabilities being multispecies but also multimodal going from genomes to functional tracks, genome annotation, all at once. It goes from human genomics to plants and metagenomics, is now capable to process sequences of 1 million nucleotides, so the longest that exists nowadays. And it's also generative. So you can design DNA sequences with de novo properties and I will show you some validations in the lab as well.

We built a suite of models from small 10 million, very affordable, to 4 billion parameter models and is also designed for efficiency even with this long context and model size.

So now we will dive into the details of NTv3, starting by the main pretraining phase. So we take NTv3, and we pretrain it on more than 150,000 species' genomes, so that's about 8 trillion nucleotides. And we do it in different phases from short to longer sequence lengths to cover the whole tree of life from very small virus, plasmid sequences to human genome of almost 1 million nucleotide sequences. All these through 15 trillion tokens. So the longest pretraining existing in genomics.

So you can do this using a masked language modeling objective where you perturb the sequence, for example, if you mask 15% of the nucleotide and you ask the model to reconstruct it. And if you do it over and over again, the training loss or the error at this objective kind of starts going down, so the model starts learning this objective. And by doing this at different model scales, we can really see the scaling laws of AI in action. So our smallest model gets this performance, and the bigger the model, up to 4 billion, gets even better.

So we have these different models of different sizes, different efficiencies that you can use now for various applications. So starting with some inference time just to show you how efficient our model is. This is the current set of models in the field. So when you compare across different sequence lengths up to 1 million, the efficiency in terms of inference time, you can see that they all suffer and it's very hard to scale these models to long sequences. So that's a common issue in the field. NTv3 was designed for efficiency for this problem. So with our 3 -- here shown, our 3 NTv3 models, you can go to 1 million nucleotide sequences with a minimal loss in terms of inference time. So it's really very affordable and possible to use for downstream task at this scale.

We tested this NTv3 on the first set of tasks, long range, around 44 tasks that go from gene expression, chromatin accessibility, genome annotation across various human tissues. And this is just to show you kind of a busy plot of all the tasks we have been compared NTv3 against other competitive models. But if I summarize all this information and group it by quantitative and classification tasks, we can observe that our models are better than the competitive models and particularly our small model, just 10 million parameters and it's already very efficient. So that's the main message, the first main message. So a very good small foundation model, very easy to use. But if you scale the model size, you can see that you get performance on both types of tasks. So larger models, better performance.

So that's on a set of kind of downstream tasks that are already useful for people, but we wanted to take a step further and bring all these functional data of genomic tracks and genome annotation into an additional post-training phase.

So we take our NTv3 model, and post-train it on genome annotation and genomic tracks experimental data. So this means that we take for a few set of species, all the introns, exons, splice sites, all these elements that matter in the genome, and we try to use NTv3 to predict them from the sequence. And at the same time, NTv3 needs to predict all these experimental data, so around 17,000 experiments from 16 animals and 6 plant species at single nucleotide resolution. So these are the kind of example profiles that NTv3 needs to predict. And doing this with sequences up to 1 million nucleotides long.

So you do the post-training on all these data, and then we can show you how we perform with NTv3 on genome annotation and genomic tracks experimental data.

So we start with genomic tracks. And just to give you an idea of what the actual predictions look like. This is a piece of our genome with 2 different genes. This is 1 million nucleotide window. And here, I'm showing experimental data from K562 leukemia cells. So at the top, you have the experimental data, for example, from RNA-seq, DNA-seq and other different assays. And at the bottom, you see the NTv3 predictions. So it's -- in one go, the NTv3 can predict for a 1 million sequence, single nucleotide profiles that match very well the experimental performance. So you see with NTv3, you can predict and recapitulate these assays. This is an example for 2 genes.

But if we now look across the genome and just compare with the state-of-the-art model, so the Borzoi model across these different experimental readouts in human and mouse, we are showing an improvement over the state-of-the-art across all of them. So we are outperforming the current models on this single nucleotide prediction of experimental data from human and mouse cells. So that's on genomic tracks.

We can also evaluate now our model on genome annotation. Again, this is a busy plot, but that's how our genome looks like. So we have a 1 million window with many genes at the top. And our model has to predict all these different elements, where is the gene, the intron, the exon, splice sites, and all these elements have different resolution.

So I'm again showing you the actual annotation with the predictions of NTv3. And if we zoom in -- to be easier, if we zoom in into a gene, now you can see kind of a better pattern of the gene with all the exons and the introns in these lines, we can see that NTv3 predicts that indeed it's a gene, the locations of all the introns and the locations of all the exons and even the splice sites, which are just 1 nucleotide out of this 1 million context window, and the same for the UTR regions.

So very rich predictions from NTv3. They look like the actual annotation and we can again summarize the performance and compare with the state-of-the-art model, SegmentNT. So the percentage improvement across all these different elements of 14 elements that we train NTv3, again, showing that we outperform the current state-of-the-art on gene finding, regulatory elements like promoters and enhancers and also splice sites. So we take the pretraining, learn from genomes. We take the post-training, learn from functional data, and we outperformed the state-of-the-art models there.

And we can even now bring the model further. And instead of just being predictive, like previous models, the previous version of NT and former, we want to bring this model to the generative space as well. So nowadays, we have models like Evo that are generative, but we don't have models that do the 2. So NTv3 is the first model that can do the 2 in 1 go. So NTv3 learns from these native predictions and functional data, but can also do de novo and conditional sequence generation. And that's thanks to the masked discrete diffusion framework that we implemented in NTv3, where you can guide NTv3 to generate sequences with a given property.

So we are unifying representation capabilities with these fine-tuning approaches with generative capabilities. And I want to demonstrate this using an example that we actually took it to the lab and validated. So we collaborated with the researchers from Vienna from the IMP Institute to design enhancers that are promoter specific. They activate specific genes. And so enhancers are sequence elements that modulate the expression of genes. So they can be very useful for gene therapy, to activate genes in different cell types.

So we wanted to design enhancer specific for promoters, but that are active at different levels as well. So we took NTv3 with this masked diffusion approach, made it generative and generated enhancers for different tasks, I will show you after, and validated them in the lab through reporter assays.

So first experiment that we did was to prompt NTv3 to design enhancers with different activity. So you take a gene of interest and you want to design an enhancer that activates the gene with low, medium, high levels. We train NTv3 to do that, generated a few sequences in the computer, sent them to the lab. They generated the sequences and added them into cells in a reporter assay. And in this spot, I'm showing the experimental results. So in gray is the native enhancers from the cells. So you see that you have enhancers that activate the gene at low levels, medium and high levels. And I'm very happy to say that when we tested the generative NTv3 enhancers, we observed the same kind of phenomenon. Our prompted enhancers for low activity were indeed lowly active, activated the gene less. But we could also design enhancers that activated the gene even stronger than the native enhancers. So this was a success in terms of generating enhancers that activate genes at different levels, again validated in the lab. So these are the first experiment.

Then the second one was to design enhancers that activate specific genes, specific promoters. So you prompt NTv3, for example, with a high activity in one promoter and low activity in the other. And then we test it in the lab, the activity of the 2 promoters with the same enhancer.

So here, I will show you the fold change between the prompted high-active promoter and the low-active promoter. So you want high fold changes, so high specificity. And we tested 2 different promoters. So these are DSCP -- the DSCP gene, and compared with the state-of-the-art generative model from using DeepSTARR, we have served a stronger specificity for the DSCP gene and an even stronger difference also for this RpS12 gene. So this is showing that our models can design highly specific enhancers towards specific genes. And again, in gene therapy, for example, this can be very promising.

So these were 2 experiments validated in the lab. And I just want to come back again to the whole presentation and the different key points that I mentioned today. So NTv3 can be used to predict experimental data that we call genomic tracks from different cells. So think about gene expression, chromatin accessibility, et cetera, can be used to predict the annotation of genomes and can be applied across different species. For example, genes, splice sites, et cetera. By predicting all these properties, we can now infer or interrogate NTv3 to predict the impact of variants on all these different properties. And you can even bring this further and generate sequences with specific properties like enhancers in this case. So very, very happy, I think, for this milestone to present here today this NTv3.

And with this, yes, thanks a lot.

Thank you so much, Bernardo. And really like I want to congratulate you, [ Thomas ] and the entire NT team. NTv3 is a breakthrough. And really like it's so extraordinary to see the team getting to build the largest context window in genomics today, state-of-the-art performance, an order of magnitude faster inference than anybody else in the field, and all this at very reasonable budgets, if we compare to Frontier Labs, it's really like a testimony to the incredible innovation happening at InstaDeep and BioNTech.

But it doesn't stop here, and we've shown -- just shown you like state-of-the-art lab-validated results in genomics, but we are also very active in protein space. And we're going to have Bora introduce our latest cutting-edge results in protein design. Bora.

Perfect. Thank you very much, Karim. So hi, everyone. My name is Bora. I'm a Research Scientist here at InstaDeep. And today, I'd like to take a little bit of time to talk to you about our use of GenAI for protein and specifically antibody engineering.

So I'd like to start by taking a few seconds just to set the scene. So when we are normally designing a protein, we're not just designing for one property. We're actually optimizing multiple properties all at once. And the solution essentially needs to satisfy multiple constraints. Now the traditional way to approach this would be to develop N models for N different tasks and then apply after the other.

The problem with this is that it's very, very inflexible. If the task at some point should change, so maybe your internal pipelines or the actual experimental pipeline changes, then you need to go all the way back to scratch, develop new models, curate new data and so on and so forth.

So we want to flip this on its head a little bit. What we envision is essentially just one big model that has been trained with as much of the data of interest as possible and so is aware of all of these things. Essentially, it's learned a very rich joint distribution over all of the different attributes that we care about. That means that at inference time, the scientists using this model and interacting with it can essentially prompt the model, specifically with only the things that they care about. So one model essentially becomes all of these previously mentioned models.

Another advantage here is that because you're training the model with lots of data, the model can also learn correlations that were previously invisible and that drives our performance.

So we spent a lot of time thinking about what sort of model, so what sort of architecture ML paradigm is the thing to go with here. And we ended up using Bayesian Flow Networks. These are very well suited to different types of data, which we encounter in scientific settings. And we first started by publishing a proof-of-concept paper where we introduced our models, ProtBFN and AbBFN. These are sequence-only models, and we actually showed that compared to leading autoregressive models, BERT style transformers and diffusion models, they outperform them in terms of both sequence naturalness, diversity and all the things that we care about.

But today, I'd like to take this a little step further and introduce AbBFN2. AbBFN2 is our first truly multimodal antibody design model, and it allows a scientist essentially to flexibly interact with the model, design antibodies for any task that they're interested in and optimize them on multiple fronts.

So when I say antibody is what I'm really referring to in this case is the F v region. So that is made up of these 2 chains, the heavy chain and the light chain, and it's actually part of the larger molecule. The reason why we focus on these F v regions is because essentially in the past years, we've seen a massive, massive expansion in the different formats of antibody-based therapeutics. You've got your kind of standard IgG molecules, but also antibody-drug conjugates, bispecifics, slightly more esoteric novel versions of bispecifics or multi-specifics and so on and so forth.

But the one thing that's common to all of these things is that the key recognition of the antigen happens via an F v. And so that's why we need to model this, and we need to model it very, very accurately. The problem is further -- is made even more complex because F vs are highly, highly diverse. So a very, very conservative estimate would be that there's more than 10^16 possible naive antibody sequences as we call them, which makes this a massive needle-in-a-haystack problem. But the issue is also that antibodies are weird molecules. Normally, a protein is expressed from 1 single gene, whereas for an antibody, 5 different random genes are essentially spliced together to produce the molecule, and the biophysics of the molecules are also very interesting. So that means that your haystack is now huge, but it's also multidimensional. So you really need fine grain control over the generative process to actually pick something out from here that works for your purposes. And that's what AbBFN2 does.

I'm not going to bore you with kind of the details of these things, but this is essentially 45-plus different modalities or attributes of an antibody that the model includes explicitly. So any design task that we can express in terms of these modalities, the model can tackle. If we don't care about one property at one time, it doesn't matter. We just ignore it, and we focus on the other ones.

So this includes stuff like the genetics of the antibody, the biophysics of the antibody, but also the sequence. And we're constantly developing new capabilities. So we now can do per-residue energetics to stabilize an antibody. We also look at things like germline families and also genetic information at the residue level. And we're also working on including structure of both the antibody and the antigen as well as quantum accuracy energetics.

So a couple of results here. The first thing that we do is essentially use the model to label known sequences, that is, I have an antibody sequence. I want you to tell me everything there is about the sequence, and I want you to label it very accurately. So here, we've tried 23 different tasks, and we find that AbBFN2 outperforms every other baseline that we've tested on all of these tasks, sometimes by a very large margin.

This is very nice because it essentially means that the model has really learned the relationship between sequence and metadata or attributes. And it also means that practically, the model is essentially a one-stop labeling tool. Rather than using 5, 10 different tools, all of which have different software requirements, you can just put your sequence through AbBFN2 and get all of the information about that you care about.

It also means that we can tackle the inverse problem. So that means I have a specific requirement and I want to design an antibody that satisfies that requirement. So as an example here, I've chosen to show you some stabilization results. So stabilization of an antibody here refers to the interface of the heavy and the light chain. So this is where they bind together to each other. And this is really, really, really important, both in the clinic, but also naturally. If an antibody is very stably bound to its kind of paired chain, then that means that it's more stable, which means it's easier to express in large quantities, which brings down costs. It also means that it's easier to store, and it's just generally something that we're interested in. This is also specifically very important in the case of bispecific antibodies because there, you really need fine-grain control over which chains will pair up with each other.

So last year, we were able to essentially recapitulate natural interface stability. So these are interface stabilities that you would expect to see in natural immune repertoire, so sequences that come from actual human immune system. This year, we've pushed this even further, and we can now actually arbitrarily set the energy that we want. And so we can tune essentially the stability of a given heavy/light chain pair.

Another thing that we're interested in is multiparameter optimization. So this is, you have 5 different properties or 10 different properties that you all want to optimize. And as I said, traditionally, you would use 5 or 10 different tools, one after the other. The problem is that these tools are unaware of each other, so they might undo each other's effect, so to speak. And also, they will introduce more mutations that are strictly necessary.

In our case, we make use of AbBFN2's capability to understand all of the attributes all at once. And we also make use of inference-time compute scaling. So we tell the model, here is the starting sequence. Here are the 5 things that I want you to optimize, so bring into those blue regions. And then essentially, we allow the model to think about its response, edit it here and there and make changes progressively.

And we see really, really nice results with this. So when we look at all of the antibodies that we've tested here, we have an 80% success rate. If we actually look at only the antibodies that you would, in the first place, take a little bit further during preclinical developments or the tractable ones, the success rate shoots up to more than 90%. And the very, very interesting results in this case is that the number of mutations for one objective, for instance, is at 46.6%. This is roughly in line with experimental approaches to doing this. But when we add 4 extra objectives that we optimize for, we actually only need 10 more mutations. So the model is really aware of if I make this change, this actually satisfies multiple things at once. So this is the best one to choose.

Now part of this is also sequence humanization or essentially reducing the risk of an adverse immune response of a sequence. Traditionally, again, with a purely experimental approach, this is often done in a kind of trial-and-error way. You take your starting sequence, you introduce a few mutations, you check that everything still works. You do that again. If something breaks, you revert back to a previous state. Do that again over and over until you essentially find your idealized candidate. This can take a very long time. But you might also, at some point during this process, realize, oh, this antibody was never going to be optimized.

So what we want to do is essentially integrate models like AbBFN2 into the experimental workflow. So rather than having this iterative approach, we essentially use AbBFN2 to optimize the immunogenicity risk. This takes 20 minutes. And then afterwards, you can still do all of the things that you were going to do, including affinity optimization.

And this really is as easy as I make it sound because we've also ensured that the model is usable, right? So we've packaged the model and it's now available on DeepChain, and we've essentially made sure that certain workflows that people might be interested in are easily accessible. So in this case, for instance, we can do conditional generation where I have certain attributes that I want in an antibody. So I could say, for instance, oh, you know, I have this specific CDR-H3, so loop-length in mind, I have the light chain sequence already, and I have most of the heavy chain. So I want you to just generate me the rest of the heavy chain, generate me a library that I can then take forward.

Alternatively, for the humanization workflow, we've actually packaged this as well. So in this case, all we need to do is enter the sequences that we're interested in, set essentially how many times we want the model to iterate on these sequences and then press go.

So to save us the time here, I've actually pre-run one of these humanization experiments. And you can see here that the input sequence is given. And you can see, as the kind of the model works its magic, changes are made progressively and over time, the humanness increases. We can also then, for instance, scroll down and check that the sequence still folds up in the same way, so nothing has been disrupted. And this is really just to make life easier for the bench scientists using the model.

So with that, I'll take you back to the slides because we've actually tested these things in the lab. In silico results are well and good, but you always need to demonstrate that these things work. So in our case, we've taken 4 antibodies, these are clinical stage antibodies, against 4 diverse targets. And these are antibodies that have actually undergone a humanization procedure experimentally. We've also done this with AbBFN2 and tested that they still bind.

In all of these cases, the antibodies still bind with good affinity. But what's really remarkable is in most of them, we actually need far fewer mutations, which allows you much more space to then do further optimization according to your needs, be that what it may be. So this is really, really exciting. We've done the work on a computer essentially, and we can show that it works in the lab.

And with that, I just kind of want to pull it back and say that the aim of the model is really to integrate into pre-existing workflows. No one should have to change their experimental workflows to fit the way a model works, but rather the model should be able to fit to your needs. And this is really possible with AbBFN's kind of, as we like to call it, condition anywhere, generate anywhere paradigm.

And with that, I'd like to thank you all for listening and hand back to Karim.

Thanks, Bora. And it's really exciting to see the progress on our Bayesian Flow Network models. And as you can see, I think one of the differences with last year is this time, we have lab validated results. You saw that for nucleotide transformer, you're now seeing it for our generative protein models. And we are really focused on having an impact. And so where are we now in this presentation? We past the halfway. And as you have seen, we've been looking at compute or Kyber cluster results on scientific computing, then we looked at algorithmic innovation. And now we're going to get closer and closer to applications. And I think a specific point, which is extremely important, is working hand-in-hand with our biotech colleagues on the data front, making sure we can extract as much insight as possible from the data.

And in this context, I'm very happy to introduce Nicolas, the Head of our BioNTechAI team as well as Youssef to tell us more about the work we're doing in data.

Hello. Pleasure to be here. My name is Nicolas. I am the Head of BioNTech AI team at InstaDeep. Hi, Youssef.

Hi, I'm Youssef and I'm Machine Learning Engineer at InstaDeep.

Basically, BioNTechAI strategy is quite simple, as Ugur mentioned, it's driven by data, and there is always potential to continuous improvement of our algorithms. The more and more we generate data, we will show that in the context of the iNeST1 personalized vaccines, but this is all across the company. And we are also aiming to learn as much as possible from the tumor. This is where the information is and this is where we need to develop algorithms to leverage as much as more this information for the design of effective vaccines. So we would like to walk you through 2 examples of how we are designing AI algorithms and tools to actually learn from the tumor and learn from the data itself, one on the Sequence Space and one on the Image Space.

First, let's talk about the Sequence Space. And for that, I want to introduce you the concept of the Dark Proteome. The Dark Proteome encompasses uncharacterized proteins from hidden translation products beyond the canonical proteins and known PTMs. Now those proteins that come from protein coding genes, traditional classical protein coding genes. So there is a whole new world of proteins or peptides that are not born the same way. They came from aberrant splicing events or gene fusions or long noncoding RNA sequences or noncanonical open reading frames.

So how can we look at this? We wish we had like sort of lantern to illuminate the dark proteome. And for this, we developed InstaNovo a tool that -- thus protein -- sorry, peptide sequencing, de novo peptide sequencing library free peptide sequencing. And I will tell you why this is very important. Sequencing peptide is very complex, right? It's not as simple a sequencing DNA. You need to chop your peptide into pieces, into fragments and then accelerate those fragments in a magnetic field, these fragments have a master charge, so they give a trajectory, and then we end up having a spectrum like the one you see here, the MS2 spectra here. In traditional mass spectrometry, what you view is you will have a library, a reference library where you really know what you are looking for, for canonical human proteome that is easy, but for de novo peptide -- sorry, for noncanonical peptide, that's a bit more complex for dark proteome.

And once you have the library, you do a database search. So you try to match this is Spectra with your library to finally get -- in this case, you're a know dark tumor antigens. What InstaNovo does is the problem of having this library, which we don't really know in the context of these noncanonical peptides. Another interesting thing is that these peptides could be very cancer specific. So they are great for designing targets, new targets, target discovery or biomarkers for cancer. Ugur said that in the end, the cancer fighting cells like your immune system fighting your cancer cells, well, you want to kill the ones that are cancerogenous, right? So your target needs to be cancer specific.

So just to give you an idea of how we are using InstaNovo here, we see a table here where we have tumor and normal identifications, and we find a few peptides where you see that the number of tumor identification is much larger. The output of these peptides come from InstaNovo. So it has already shown this potential in detecting tumor-specific epitopes from this undocumented open reading frames. The InstaNovo has been published in Nature Machine Intelligence, and we made it available for the whole community to use it and try it. And it has been also covered by Science Magazine on an article of next-generation de novo peptide sequencing. This is work that has been done in collaboration with Professor Tim Jenkins and DTU. And we are extending this collaboration for introducing InstaNovo V2, and even larger model, 63 million labelled spectra, where you see the increase in the peptide spectrum matches, and it has a higher accuracy like 10%, 15% increase in accuracy in the data set that we have been testing. So we are very excited to apply it in BioNTech for the discovery of new targets and biomarkers -- cancer specific targets and biomarkers.

With this, I would like to leave the place to Youssef to show us a bit of how we are trying to improve our digital pathology algorithms.

Thank you, Nicolas. Hi, everyone. So last year, we showed our AI-Assisted annotation tool and how we increased the efficiency of pathologists fivefold. However, 5x faster pathologists is still not enough because we have thousands of whole slide images to annotate. And, the question we had to answer is how can we reduce the pathologists' annotation efforts while ensuring the best model performance? And the answer to this is data. In computer vision, usually, when you look at your data when it's unlabeled and labeled, it's different points like you see here. And what we do usually is that we take random points from your data to use it for the model training. This works when you have a lot of data, thousands or millions of data points, you can label. But when you have a few data points, and we want to reduce actually the pathologists efforts. And you take your data and you plotted for example, in a t-SNE graph like this one, for example, it's a real t-SNE graph of a data set. You will see that your data points are not covering all of the patterns. So here, each cluster is actually a different pattern in your data set.

And you will be missing the highlighted patterns here, for example. When you test your model after that, you are not sure you will be getting good results in these patterns because the model didn't see them. And what you want actually is that you cover all of your patterns and you don't have to have a lot of data to label. And for that, we actually took the leading open source software in the data curation and the histopathology visualization. And we built our own internal product on that which helped us to explore, understand and work with our histopathology data.

And here, I will show a demo for that. So what you are seeing here is actually the real clusters of data. This is the CRC 100K data set, for example. And when you look at one of the clusters, here, for example, I guess, it's the tumor, you will see the same pattern there. And when you go on the other side, this one, I think it's the adipose or the fat cells. Yes. And you see a totally different pattern. And for this data set, we have the ground truth labels. So if you visualize the labels here, you will see that actually the foundation model is doing really well in clustering the data set. So you can see here that for -- it's different clusters, you have specific colors. For example, the yellow one is the debris and the green one is the tumor, for example. And it's doing even better because for the tumor, for example, here, you see that we have a lot of different clusters. So if you take this part from the tumor here a specific pattern, and you take another part here, it will give you a totally different pattern in the tumor. So we have even subclasses for each class.

And it doesn't only work on these patches. What we made it also -- we made it work on whole-slide images. And you can take a cohort, for example, for the task of the MSI MSS, and you can see all your whole slide images. And you can also see their embeddings and their t-SNE graph. And here we fine-tune a little bit the model on the task itself. And when we visualize the label, you can actually see that the MSI are most of them are grouped together, and you have the MSI low and MSS other ones. You can also see your data to find the outliers, the most unique one. So we can visualize the uniqueness here.

Yes. So the brightest -- the point is the most -- the more unique. Here, for example, if we take this point and we investigate it, let me open this one here. And we can also investigate the whole slide images inside the app. And when we zoom in, actually here, we find it's the most unique because it's out of focus. And that's how we can find the outliers or the wrong data. It's actually the focus is on the market made by the pathologists and not on the sales themselves.

Another picture also you can do here, if you can go back to the presentation. One of the feature also you can actually see your whole slide images and you can zoom up to the cellular level to investigate them. And we actually built a nice module on there where you can test different AI agents from different providers. For example, here, we are just assessing image developed by Google DeepMind and we want to see its answer to the question. So you can select the region and then you can ask the agent. For example, here, we are asking if this, for example, if it can confirm the presence of invasive colorectal cancer in the image. Yes. And here you get the response. Yes, it confirms that. Yes, you can also give it a try after that in the booth after the presentation.

Thank you very much. So you see how we are empowering digital pathologies that BioNTech with these tools. And yes, you are more than welcome to give a try downstairs soon. Now, with Karim for more applications in AI in BioNTech.

Thank you, guys. Thanks, Nicolas, and thanks you, Youssef. It's really exciting to see the progress we're making in terms of like improving the data quality that we have and also quantity. And so if we summarize, if you remember at the beginning, we said we have 3 engines that are powering the AI plane. The first one is compute, and that's what we saw with Alex. And then we looked at AI innovation with Bernardo and Bora. And finally, now on the data front with Nicolas, Youssef and the BioNTech AI team. So this is all very nice, you could tell, but then what can we do with all this? And what is really exciting with having all those capabilities under the same roof at InstaDeep and BioNTech is that we can start now to tackle truly hard biotech problems. And today, we're going to show you our first results in terms of applications, starting with nanoparticle design with Lexi and Cheng.

Hi, everyone. My name is Lexi, I'm a scientist at BioNTech. And I've been working together with Cheng for the last year. One thing that we're really interested in is how to develop the best vaccine. And in order to do this, we look at first, what is our immune system trained to respond to. Oftentimes, that is viruses and bacteria, and these viruses are large, and they have a highly repetitive surface and sometimes that surface is symmetrical. So what we can also do is look at what have some historically successful vaccines looked at. They've actually taken advantage and harnessed this capability of having something that is large, something that has a repetitive system on the surface and is symmetrical.

Some examples of this include the Hepatitis B vaccine against the Hepatitis B virus. The human papilloma virus vaccine that helps against cancer and more recently with the malaria vaccine. Now all of these really harness what our immune system is trained to respond to. They have an antigen on their surface in this large repetitive manner. And so we would like to combine with InstaDeep to be able to do this from scratch using AI-assisted de novo protein design. But that's not the only thing that we want to do with this innovation as we also want to marry this together with the power of mRNA technology, which has been so, so successful for many vaccines.

Now what does this look like practically? What this looks like is we would like to deliver mRNA and utilize the cell to build our nanoparticles from scratch. This begins by starting with a single protein component that must first find its friends and velcro to 3 other components of the same thing in a really oriented way. Once they have found these friends, they need to continue to assemble into up to 20 of these trimers coming together to form these beautiful repetitive arrays to form a nanoparticle vaccine. And this nanoparticle vaccine will eventually hold antigens of interest that we want to tailor to our specific vaccine of choice.

Now what can this look like? We want to be able to design not just one of these, but ideally, we would have a library of these tools that are tailored and fine-tuned to the application at hand. And here is just an image showing how many of these nanoparticle designs that we want to be able to build and bring to life. So just to really drive home how complex of a process we are trying to do here. What we really are asking is to build a protein from scratch that we can launch from mRNA and have this protein really interact at the molecular level with not just 3 other proteins, but come together and form a 6 steamer, up to 6 steamer of these proteins in this beautiful, amazing nanoparticle array.

And so to walk you through some of these details, I'm going to hand it over to Cheng so that she can tell you about the amazing advances that they've done.

Thank you, Lexi. Hello, everyone. I'm Cheng, Research Engineer at InstaDeep. So now let's see how can we build a nanoparticle step-by-step. Just as you can see in the video, it's like building a house. So we start by designing some small pieces of building blocks. In our case, they are the trimers, which is an assembly of 3 identical proteins. So using generative AI models, we can design thousands of de novo trimers as you can see here, always different sizes and shapes. These trimers will form the basis of building blocks to build our nanoparticles, okay?

Now we've built our building blocks how do we construct the nanoparticle exactly? Just as houses have their architecture, nanoparticles will have their symmetries, as you can see here on the left, they can be a tetrahedron, which consists of 4 trimers or in the middle and octahedron, which consists of 8 trimers or even on the right, you can see the biggest one icosahedron consisting of 20 trimers. So all these previously generated building blocks can be computationally assembled to this user defined various shapes, and this leads to thousands of symmetric nanoparticle assemblies.

Until now, we've only designed this 3D structure of the nanoparticles, but in order to make a house habitable, you need to add [indiscernible] to consolidate the structure. So in the case of protein design, we will need to design the amino acid sequences to make the protein really functional and really forms the desired shape. In this case, we use AI models to generate hundreds of amino acid sequences per nanoparticle, which are supposed to really form the desired structure.

Okay. Now we've generated hundreds of thousands of nanoparticles, but it's extremely challenging to -- for these small pieces of proteins to find themselves and really assemble as exactly as what we want. To confirm this, we will need laboratory testing. But it's usually time consuming and very limited by capacity. So the question is that how can we select the most promising candidates so that we can test them more efficiently and achieve a higher success rate. Here comes InstaDeep solution DeepChain Folding Studio. It integrates the state-of-the-art protein folding models and allows large-scale screening within a short amount of time. So just to give you an idea, we can screen 10,000 designs within 1 day.

So now Lexi will show you how this narrow down high-quality designs perform in vitro testing.

Thanks, Cheng. So this is the moment of truth for a biologist is to go into the lab and see how we actually designed these proteins to structurally form what we want them to do. And what I have the pleasure of sharing with you today is that, yes, we can do this. We can build these nanoparticles, as you can see the models on the top of the screen, a variety of different shapes and sizes. And then we can go into the lab and utilize an electron microscope to see that, yes, we are able to build these nanoparticles as Cheng and her team have designed.

But we didn't decide to stop there. What we're really interested in is functionalizing these nanoparticles and placing antigens of interest on to the surface. And so we took it this step further, and we can also show that we can place antigens on the surface of these nanoparticles and they still can structurally come together as designed and intended as again shown by these electron micrographs. So this is really an amazing feat of AI-assisted de novo protein design and structural biology coming together for enhanced vaccine.

So thank you so much, and I'll hand it back to Karim.

Thanks a lot, Lexi and Cheng. And really, I don't know about you, but for me, this is really magic to think that you can design a protein sequence just purely with AI and have it to self-assemble in a trimer and then self-assemble again at much larger motives. And then potentially have this as a scaffold of use to be -- to put antigens and trigger like immune responses. And so that's a significant challenge that we managed to overcome in this project but the applications don't stop here.

And for our last but not least presentation, we're going to show you amazing work done into the other side of developing like an immune response, which is, could you actually fit a particular -- design particular TCRs for a given antigen target? And this is what Mike and Antoine are going to tell us about it.

Great. Thank you, Karim. So my name is Mike. Nice to meet you all. I'll be presenting today with my InstaDeep colleague and, Antoine on our work on T cell receptors also known as TCRs and specifically how to make these into a strong binders as possible, something that we think is critical to unlocking their full therapeutic potential.

So why focus on TCRs? Well, one reason is that TCRs can unlock antigens that are otherwise not available with conventional antibody-based therapies such as ADCs. And the reason for this is that antibodies need to target things that are on the cell membrane. And the limiting factor here is that membrane targets by and large, are not usually cleanly tumor-specific because this means that there's some residual expression on normal tissues, which limits the dose. TCRs on the other hand, recognize antigen in a completely different way. And this is something that we were actually mentioned earlier, but we have a process called MHC presentation where proteins inside cells. The whole protein is subject to this. They're digested into peptides at the end of the life cycle and they're sent to the cell surface on a molecule called MHC. And that's what TCRs can recognize.

And so it's essential here is that's -- something T cells can see basically the whole proteome not just the component that's on the cell surface membrane protein. And because of this, this unlocks antigens that are some of the highest quality cancer antigens we know of like oncoviruses, cancer mutations, new antigens as well as genes that are expressed due to dysregulated gene expression in cancer. The other reason we really like T cells and TCRs is that we believe that they are likely critical to getting durable responses in cancer. Probably the best example of this is checkpoint blockade where we now have data showing how durable these responses can be. So this is data from nivolumab in non-small cell lung cancer showing that 5 years out, we had this tremendous divide between patients who got the nivolumab versus chemotherapy.

And then more recently, we have data from TCR-T. TCR-T is a cell therapy where patients cells are engineered to express the cancer-specific TCR. And the data we've seen so far is that these can also have very durable responses. This is a TCR against a cancer antigen called PRAME. So our thesis is here is that likely to get the most durable effects with cancer therapy, we likely want to be bringing T cells into the fight. But there's a challenge with T cells, which is that their natural binding affinity to their targets is actually quite weak. It's in the micromolar range. And this is okay for their day job, which is going after viruses and bacteria, which are very highly expressed. But when we want to go into cancer where the antigens are more typically weekly expressed or variably expressed, we need these to be very strong binders. So to be in the TCR-T cell space -- the cell therapy space, we probably need our binders to be nanomolar binders.

And to get that, we either need to be very lucky and find the very rare natural T cells that combined at that very strong level or we need to do some sequence engineering to make these into stronger binders. Now if we want to go with an off-the-shelf biologic, avoid cell therapy, we need probably an even stronger binders, something in the picomolar range. And that's going to be a million-fold increase in binding overload you would typically see with a natural TCR. That's a huge increase. You're not going to get there ever with a natural TCR, it's probably going to take 10 to 15 mutations. So that's a serious mathematical engineering problem to solve.

And one thing that we've realized now after several runs that we need to have a really strong computational process. The standard approach to this problem is something called face display. So in face display, it's a fully experimental process that is randomly exploring the sequence space. It's done for each of the 6 CDR loops, the complementary determining loop of the TCR. And at the end of the day, this will typically explore about 1 billion sequence variance, which sounds great. However, the true space, which I said is about 10 to 15-point mutations away from natural TCR is 10 to the 32, that's a huge number. So even with the face display, we're just scratching the surface of all those variations and to find a TCR that is developable, it binds strong, it binds specifically. It'd be quite lucky with just a random exploration of the sequence space.

So what we've developed is a new approach. We replaced the face display with something that's AI-guided and it's rational. It's choosing no variance in this huge space to 10 to the 32, but in a way that we think is much more effective. And because of this, we are having success in finding TCRs that can check all these boxes. But again, like the competition is key. And after having done this multiple times now, what we realize is having a solid understanding of the peptide and we see TCR structure, which varies target to target is critical.

And Tom will talk about our nuances on that difficult problem.

Thanks, Mike. So yes, let's dive a little bit into the structures of TCR MHC complexes. So the reason why we are interested in this structure is because we want to understand the physical interactions between the TCR on one side and the MHC on the other side. So now when we talk about TCR affinity structures, we often mention the CDR loops. So there are 6 CDR loops, 3 on the alpha chain, 3 on the beta chains. And these loops are highly flexible regions of the TCR that are getting in contact with the MHC.

So if you look at this left graph here, you can see that the -- where we represented 12 structures, right, lined on the same MHC. You can see that the loops tend to cluster into the same regions. So that tells you that overall, the docking mode is concert. However, if you want to know the exact shape and position of each loop, then you have a huge diversity. And that's really the key problem when it comes to TCR-pMHC structure prediction. And we have a good example here where we took the CDR2 beta and MHC of 2 different complexes. And these segments were in the structure are actually fully determined by the genome. So you would expect that if they share the same sequence, they will have the same interactions. Okay? But it turns out it's not the case. So this is why we really need to have a very accurate TCR-pMHC structure model if you want to be able to understand these interactions.

So now let's talk a bit about how you can model this in silicon. And this has been a lot of improvements in the field over the last few years, a lot of exciting work in the community. So based on this amazing work, we've decided that we would actually build our own model, right? When we want to benchmark these models, we actually are interested in the accuracy of the CDR loop that I've just mentioned. So here we've benchmarked these models on a set of completely unseen targets, and you can see that our model performs better than our competitors. Now our competitors are generative models. So this means that a common strategy, if you want to boost their performance is just to sample many more structures for each target. But actually, you can see that if you do this, while they don't even match the performance of our internal model.

So now you may wonder, okay, how do we leverage the structure into our pipeline to design these T cells. So we start from the natural TCR that has a low binding affinity, then we obtain the structure, and then we use 2 different AI algorithms. So the first one is the variant sampler. So it's going to propose candidate mutations. And the second one is the affinity predictor. So the affinity predictor is here to rank all of these mutations and help us to select them. So then we go to the lab, we make experimental measurement of the binding affinity, and we repeat this process 3 times until we reach the desired TCR binding affinity. And the nice thing is that we don't need to actually test thousands of mutations. We can just restrict to a few hundreds.

So on the right side, you have an example of a very successful campaign that we had. So on this graph, we are representing the dissociation constant. So the lower the value, the stronger the banding affinity is. So initially, we saw with this with [ WT ] TCR and on 0. And then after just one round, we enter into the nanomolar range. So this unlocks the first therapeutic modality, which is called TCR-T. And then we continue after 2 rounds we -- and for 2 rounds and then at round 3, we reached the picomolar range, which unlocks the second therapeutic modality [indiscernible] TCR. And the nice thing with this pipeline is that we are actually able to repeat this process, and we repeated this on 4 different targets. And on average, we had an average of binding affinity enhancement of 50,000 fold.

So now I'm going to show you something in nicer, which is in vivo results on an animal model. So the experiment is actually quite simple. You take the subject and you implement the tumor. And then every day, you measure the volume of the tumor and you inject the treatment. So if the treatment works, the tumor should not grow and if the treatment doesn't work, then it grows. So we've tested this on 2 different cancer targets. So we have 3 curves here. So the gray one is our first control. So this is what happens if we don't inject any treatment. Then the second one is the red one. So this is our second control. And this is what happens if you inject the wild-type TCR. So without any affinity enhancement. And then the last one, the yellow one is the tumor controlled with our enhanced TCR. So you can see the results are quite striking, I think, here, yes.

Now yes, we are quite happy with this pipeline. We've built this very robust pipeline. We hope that we can move forward and in the long run, make very good progress to elicit durable immune response for patients affected by cancer. Thank you.

Thank you, guys, and really like very exciting results and I'd like to congratulate the joint teams at BioNTech and InstaDeep for the results. And like Ugur mentioned, this wouldn't be possible without like a lot of collaboration. And perhaps like these 2 examples that you have seen really show you the power of combining together AI expertise, compute at scale with lab experiment capabilities, but also the significant biotechnology expertise of our colleagues at BioNTech. And so this is really integral to getting results and state-of-the-art results like we've shown you today. So really, congrats to the 2 teams. And we're going to have a Q&A with Ugur, and we're going to make this part a bit more interactive.

So Ugur, if you'd like to join us.

Any questions from the audience?

Questions? Don't hesitate. Yes, we have one here.

I'm Francisco, Genomics England. I have a question about your approach to understanding tumor biology. So we have excellent insights and applications based on the understanding of the human -- or the tumor genome proteome. I wonder if you can tell us about your strategy to also incorporate those interactions between the tumor and its microenvironment and how that can lead to new treatments?

Yes. This is an excellent question. Of course, the tumor, there is much more information than the genomic information. We do also transcriptomic analysis and the transcriptomic analysis gives us, of course, an understanding, for example, whether T cells infiltrate the tumor, the activation status of the T cells, but we can also decipher more or less all types of cells that are infiltrating the tumor. And in cancer immunotherapy, there is -- there are, at the moment, categorization, the simple categorization of tumors into PD-L1 positive tumors. PD-L1 high positive, low positive tumors. So at the moment, the pharmaceutical industry is running just with a single parameter. But we see that the information in the tumor is much, much bigger.

We can also see evolutionary processes happening in the tumor, for example. For example, we have found tumors related to microglobulin, which is the key molecule for presentation, presentation of these epitopes is lost. So there is much more information, and this will come over time. Yes, we will make use of the information to see how the battle between tumor and immune system is going on.

Hello. I have a question from the webcast here. This is from Jeana Han from TD Cowen. Where are you mining data from to train your models? And secondarily, how do you incorporate data generated in-house at BioNTech, either preclinically or from clinical trials to help train and improve the models?

Ugur, perhaps I can say a few words on the part about data mining and Ugur can mention on the clinical side. So really like what we've been trying to do and in some cases, like managed to do quite well is collect all the available open source data that we can get our hands on. This is what we did, for example, for the nucleotide transformer series NTv3, we reached 15 trillion nucleotides in total in terms of like data to train on. And so that's one part.

And then working collaboratively with all the different biotech teams to have specific data adding to that mix. I want to mention also that data is not all of the same type. We start with pretraining at very large scale. But the value of data that is specific to give an experiment, if you think about, for example, what we have shown on TCR affinity maturation. So lab data, like developed in collaboration with the different teams is essential here, and that makes also like a huge difference in terms of getting results. So get as much data as possible externally and add specific but very high-quality data internally. And we're working more and more with the clinical teams also to unlock these capabilities.

And I don't know if you wanted to say a few words on that, Ugur.

Yes. It is indeed, yes, with regard to the clinical data, we are not yet in the space of big data, but we are in the space of deep data. With deep data, I mean, really, the multidimensional information that we can get for patients, including, for example, the image from the histology. And even with a few hundred patients and without using our sophisticated AI systems, but more or less unbiased statistical testing. We are able to see amazing, amazing correlations between survival and biomarker data.

And actually, Ugur, you told me sometimes even like looking at data manually with your expertise, you could see like patterns, yes.

Yes. This is actually one good friend of mine is using the term AI for actual intelligence. And we can really benefit from them until we have much more data and then actual intelligence could become artificial intelligence. So this is the way at end of the day, the learning systems are really based on our human knowledge, where we need to pay attention. And later on, the power of the systems of the AI systems is to go beyond that and see patterns that we as humans can't see.

Yes. And we see actually even like an opportunity to learn directly from the expertise of the biotech experts. If you think about like processes like RLHF like -- from human feedback, it's really about that. And so modern systems could actually learn directly interacting with the different scientists and this is like one of the projects we're working on. So there is a lot more to uncover.

Michael Pye from Baillie Gifford. It sounds like a lot of the advances you've made in models, so NTv3, the antibody discovery design model. These sound like intuitively to me as a non-specialist could be very valuable outside BioNTech. How do you strike a balance between advancing the field and kind of be monetizing the work that you've done in this space?

I would say -- so first, like we try to bring all this innovation to our DeepChain platform, and you've shown -- you've seen some live demos of it today. I would say, in terms of priorities, clearly, our #1 priority is to work with our BioNTech colleagues to progress the different projects we have in Oncology. But biology is extremely vast. So there are lots of opportunities to develop collaborations with external partners when these do make sense for BioNTech Group, which is very often the case. Not everybody is focused on the same problems.

And like you said, those models are quite generic, if you understand antibody structure, different properties that has like a broad application. And I would say, even broader for nucleotide transformer. To give you an example, we've built -- we built partnerships even in plant genomics. So yes, there is a broad appeal, but our first mission is to support our BioNTech colleagues. And when there is a chance to build win-win partnerships, we will take it.

And a very quick follow-on, if I may, further. Can you help us to understand what can you do with nanoparticle mRNA that you cannot do with today's mRNA constructs?

I think one of the key aspects is the duration of T cell responses. So as we know, we know that mRNA can induce really high antibody titers. But we also know that the titers as dropping with the half-life of antibodies in the range of 21 to 30 days. And with nanoparticles. We hope to see more stable antibody titers because there's nanoparticles remain in the body for a pretty long time.

Maybe another one from the webcast then. Agentic Systems are the new frontier for AI models, such as Gemini and ChatGPT just to name a couple. When do you think BioAI agents will become viable and/or useful to us?

I mean they're already useful. I think as you've seen, like, we're very excited about the applications and some of these are already coming into fruition. But it is true that we're going to see much more in coming years. Perhaps a limitation that we have with the current systems is the fact that -- like if you look at models like ChatGPT, Gemini, Grok and others, they're really focused on learning and training on the Internet as a whole. So when they understand biology, they understand it like from reading articles or web pages, they do not, for most of the time, understand biological sequences themselves.

And when you bring in like deep understanding of a biological sequence at the nucleotide level, you see really like magic being uncovered. And we have that example with our ChatNT work that made the cover of nature machine intelligence. This was really bringing an expert nucleotide biological sequence model with a general purpose language model, and that showed a lot of promise. So I think that's the frontier. And in this particular case, the team was able to push the frontier forward, but we're going to see a lot more from that. And the future of Agentic Systems is our system that really understand the biological data in multiple modalities but they also can read like scientific literature in real time to be able to provide perspective and almost like generate novel ideas.

I remember, Ugur, you had given us this out like a few years ago, and this is starting to become true. The systems are capable of formulating scientific hypothesis and we see this coming and becoming more and more frequent. But if I had to say, like what's the year where you see this really starting to play at scale, I would say probably 2026, so next year.

Great. Maybe one final question from the webcast. Given the broad potential for AI, both at BioNTech in general, which technologies, modalities and applications do you think will be prioritized at BioNTech specifically to deploy our AI tools or AI capabilities?

The way how we develop BioNTech is really solving one big problem, yes, how to improve cure for cancer. And this is not a single application. It is, as you have seen today, really a series of modular task that if combined, provide powerful capabilities to develop novel antibodies, better ADCs, better mRNA therapeutics, better vaccines. And in this setting, I think one of the presenters said, having a system which is universally aware of all the schools has advantage from a very specialized as compared to a very specialized model.

So we have to see how this evolves, but we are we are very confident that this approach -- this holistic approach, understanding immunity understanding cancer, supporting development of therapeutics in the broader scale is the way how future pharmaceutical companies should be built.

Cool. I think this was our last question. So thanks again for everybody who attended either in-person or online. It was a pleasure. And yes, stay tuned for more progress. Thanks a lot.