Really excited how things have worked out. I will not waste a lot of time, I'll just quickly go through the agenda and housekeeping items. Let's start with the morning sessions. First speaker again will be our Pharma CEO, Teresa Graham, providing an update on the overall pharma strategy and life- cycle options for our rather fresh and well-established on-market portfolio. The second speaker, as you can see, will be Levi Garraway, our CMO and Head of Global Product Development. Levi will provide the R&D excellence KPIs, which we started to share with you three years ago, and which will allow you to track our progress in that area. Third presentation, then, is given by our special guest, Aviv Regev, our Head of Early Research and Development at Genentech, gRED. Aviv will talk about our industry-leading adoption of AI in early drug development, and she will also share some KPIs tracking this impact. I think that's, at least to my knowledge, an industry first. The morning session then will close with Alan Hippe, CFO of the Roche Group, providing us an update on financial excellence and digitalization in group finance, all prerequisites for thoughtful asset allocation, which will allow us to optimize our investments and keep innovating for patients. Following the morning presentations, we have a 30-minute Q&A with all the speakers of the morning sessions on stage, including Karst Jung, our Head GPS, and Wendy Cheung, our Global T herapeutic Area Head for Oncology and Hematology. Afterwards, we're heading into a 50-minute lunch break, and after lunch we will continue with the pipeline presentations covering all our therapeutic areas. The first session will be given by Levi Garraway, covering oncology, solid tumors, and you might have seen the exciting news of today, we will announce new trial starts here for giredestrant in the adjuvant breast cancer setting. Second speaker, then, will be Matt Hellmann, who just joined us from AstraZeneca, where he led the early oncology development. Matt will be our Global Head of Oncology and Hematology Product Development, and he will take us through the malignant and non-malignant pipelines. Third speaker is Hideki Garren, our Global Head of Neurology Product Development. Hideki will provide us an update on our development efforts in Alzheimer's disease and Parkinson's disease, and he will also provide some important phase II transitions. Next, Larry Tsai, our Global Head of Immunology Product Development, who will touch on Gazyva in kidney diseases, the upcoming readout for afimkibart in IBD, and of course on sefaxersen in IgAN, where we had positive top-line news out last week. Following Larry, we have Chris Brittain, our Global Head of Ophthalmology Product Development. Chris will provide us an update on NMEs currently filed with the regulators, but also cover some early development efforts. Finally, we will have Manu Chakravarthy, our Global Head of C ardiovascular, Renal, and M etabolism P roduct Development. Manu just has some intense weeks behind him with the opening of our new research and development center in Boston. He will present one of the highlights of today. I think for the first time ever, we show a data snippet at a Pharma Day which has not been previously presented at a medical conference. I'm talking about the phase II data for enicepatide in type 2 diabetes patients with and without obesity, where we had a top-line release out last week as well. He will also cover the further development of our CT-996, which we now will take into phase III development. After a closing remark by Teresa, we will have our second 30-minute Q&A with all the speakers from the afternoon sessions, including, again, Karst Jung and Wendy Cheung. The event then will close at 2:40 P.M., but you're invited to stay around for the buffet reception, giving you an additional opportunity to talk to management. Let me also mention one additional housekeeping item. We have, again, prepared a short 11-question feedback survey. The link to the survey will be shown to the participants in the webinar 15 minutes before the end of the event. Participants in the room will receive an email with the link to the survey about one hour after the event. We would very much appreciate if you could, again, give us your feedback, as we always strive for improvement. With that, it's my great pleasure to hand over to Teresa. Teresa, please. Good morning, everyone, and welcome to Pharma Day 2026 to everyone who's joining us in the room and online. I wanted to start by just quickly taking a step back and thinking through a little bit the journey that we've been on as a company over the last number of years. Let's start in 2026. I'm sorry, 2022. Not much of a journey back if we're starting today. 2022 was a tough year for Roche. We had a lot of significant pipeline failures, and that caused us to take a big step back as a management team and actually look at the operations of our business from end- to- end. When we came together in 2023, we rolled out a very transparent look about where we were, particularly with our R&D pipeline. We shared with you the beginnings of R&D excellence, and how we intended to invest end- to- end across our entire R&D value chain to ensure that we were able to change the trajectory of where we were headed. In 2024, I stood on stage and I shared with you the One Pharma Strategy, which was our view of where we wanted to play and how we were going to win in an increasingly competitive set of market spaces. Then last year, in this room, for the first time, we talked about our ambition to be a top three obesity player. Today, I want to share with you a simple reality: that all of that hard work to build a great pipeline, and all of the work that we've done through One Pharma Strategy to build the right capabilities to bring that pipeline into the world, only matter fundamentally if we can turn that into patient impact at scale around the world. What we're going to talk about today is exactly how we intend to do just that. I'm going to start by giving you just a quick snapshot of where we are since last Pharma Day. We're then going to take a quick look at our strategy and how we've really invested, from AI to manufacturing, at enhancing and building the fundamental things that are going to be required for us to ensure that we can reach as many patients as possible with our incredible pipeline. We're then going to take a look at commercial excellence and exactly what we believe it is going to take to succeed in the therapeutic areas in which we're entering. Finally, we'll close with a look at what you can expect in terms of growth drivers over the next couple of years. Let's dive in. We have made significant progress since last Pharma Day. We shared with you at half-year that sales remained strong at 6% growth. We're continuing to grow COP at double digits and increasing margin. We've had 10 positive phase III readouts since last Pharma Day, including, as Bruno mentioned, the sefaxersen results in IgA nephropathy that were positive at interim just last week. Since last year, we've advanced six new promising molecular entities into phase III, including the decision last week to advance our oral CT-996 into phase III, which Manu will talk about a little bit later. Tremendous progress on the pipeline. Of course, that progress didn't just happen; it was the result of very hard work. I'm not actually going to spend too much time diving into this, because you'll hear a whole update on R&D excellence from Levi, but I think it's very clear to see that we have made measurable progress against every part of our R&D engine. From the number of best-in-disease medicines that we have, to the value of that portfolio, to the rapidness with which we're able to advance programs. You can see over time, since the introduction of the bar, that the number of programs in our portfolio have gone down, and that's because we systematically removed medicines that we did not believe had the opportunity to be transformative. But the medicines that we have replaced them with are of significantly higher value. We actually have the highest value pipeline that we have ever had at Roche today. That is because, fundamentally, we are focused in areas that we believe can make the biggest impact to global health. The famous donut slide, I think, tells this story really well. Believe it or not, since you last saw it in July, we've actually had a number of really important updates. The first one is the top-line number. We have gone from 19 new NMEs with a potential launch by 2030 to 20, which, as you will recall, was our ambition for the end of the year. When we introduced this ambition to you in 2023, it was not a foregone conclusion that we would hit this. In fact, we had quite a gap. It's pretty incredible over the last couple of years that we've managed to refresh this pipeline so significantly. Let's talk a little bit about the puts and takes within each of these programs. Let's start what's not on this slide anymore, and that's emugrobart. We recently made the decision to end development of emugrobart, which is our anti-latent myostatin with the Chugai's sweeping technology, from development in obesity. This was based on a thoughtful look at the interim GYMINDA results, which just showed us that we did not believe that it was going to have the potential to reach the TPP that we had in place for it. We have made a decision to discontinue development. We are returning that molecule to Chugai, and Chugai is looking to further development in SMA, as well as considering potential out-licensing opportunities. There are some exciting updates to this slide in the other direction. Two recent additions, Tam-Peli in small- cell and non-small cell, as well as other solid tumors, which just had some really impressive data presented at World Lung, and which we'll discuss later this afternoon, is an incredibly exciting new addition to the portfolio, as is our VEGF x IL-6 DutaFab in DME, which you'll hear more from Chris about later. I also wanted to be true to my word and let you know that we have updated our divarasib revenue expectations from CHF 1 billion-CHF 2 billion to CHF 2 billion-CHF 3 billion, and that's based on the strength of the KRASCENDO 1 data, which, as you know, is our phase III head-to-head against standard of care. We are extremely excited about the molecules that we have on this chart and the opportunities that they represent for patients. Ultimately, what makes that great pipeline possible is strong financial performance. This performance is clearly driven by our best-in-disease on-market portfolio. You can see that we've continued to have strong sales performance and continue to increase the margin. You might notice the sales growth volume in 2026, which drops off a bit. That is largely related to Xofluza and the fact that we had a weak respiratory season at the beginning half of this year. Xofluza clearly is a high sales volume product. When I have talked to you in prior Pharma Days about rigor in the science and discipline in the business, this is what discipline in the business looks like. We are carefully considering every investment that we're making, every expense that we're bringing on board, to ensure that it's delivering the greatest possible return both to the portfolio and enabling our commercial infrastructure. It's because of that strong financial discipline that we believe we have a strong base to grow through the coming years. This is a group slide, which you've seen many times, so I'm just going to spend a little bit of time on some of the individual pharma dynamics here. The on-market portfolio will continue to provide growth right through the end of the decade. We have no patent cliff that we are looking at, either up to 2030 or beyond. We have some exciting things coming into that on-market portfolio, whether that's Gazyva, which is already approved in lupus nephritis, and is now getting its other immunological indications approved, the giredestrant launch, which we're expecting by the end of this year, and the fenebrutinib launch, which should come early next year. These are exciting new best-in-disease opportunities within our on-market portfolio. By the end of the year 2027, we have the opportunity for six more launches: divarasib in second line, non-small cell, vamikibart in UME, Enspryng in two new indications, thyroid eye disease and MOGAD, which is actually generating a lot of excitement in the neurology community, sefaxersen, and, of course, the new kid on the block, afimkibart in UC, where we get data at the end of this year, beginning of next. We have a tremendous base to build from, and we are incredibly excited about the opportunities that are presented. Our confidence in this growth outlook is really based on two critical things. One is the strength of the pipeline, but the second is our confidence in our ability to deliver that pipeline where it matters, and that's to patients. How are we going to do that? We do that through the implementation of our strategy. This is how we bring exciting science from the podium to patients. That journey ultimately begins by making sure that you have the right foundations in place. When we introduced the One Pharma Strategy, we introduced our five target therapeutic areas: oncology, hematology, immunology, CVRM, ophthalmology, and neurology. These are the five therapeutic areas that represent 60% of the cumulative disease burden that our global healthcare systems are facing. They're the five areas that also represent about 80% of the growth that will happen in the pharmaceutical industry. Currently, within these five therapeutic areas, we are focusing on 11 disease areas in an end-to-end way. What this means is that all the way back from early research through commercialization, we want to know that we have a pipeline that will allow us to continue to lead in these areas. This focus has created a significant momentum in the company, because everybody knows where they're going and how they need to get there. Last year, we shared with you the deep dive, or the sort of overview, of all of these five therapeutic areas and what we believe the criteria for success are. Neurology, immunology, ophthalmology, and CVRM are unchanged. Based on the significant movement that we've had this year in our oncology pipeline, we felt it was worth a look at what our aspirations and ambition ought to be in oncology. I wanted to share that here with you today. We want to achieve leadership through transformative anchors, and those anchors are things like giredestrant and divarasib. We want to establish and expand our position in key patient segments, and we want to make sure that we're continuing to invest in that next wave of innovation. When you have Matt and Levi up here a little bit later, I hope you will similarly be excited about what you see, because we have made a lot of advancements in our oncology pipeline, building off the historical legacy of success that we've had, but with eyes absolutely pinned on the future about where treatment is going and what we need to do to be successful. Of course, that great science just doesn't magically turn into a great medicine in the clinic. There are a number of core capabilities that we know we need to invest in to succeed. These capabilities are where I have spent a tremendous amount of my time over the last couple of years, ensuring that we are well-positioned in every single core fundamental area that we will need, not only to deliver the portfolio of today, but to seamlessly deliver the portfolio of tomorrow. Let's start with modalities, because I think this is one of the more exciting areas for us. You'll probably think of Roche and Genentech as a biology-driven company, and you're right. That's always been true. But increasingly, we have more modalities at our disposal to make sure that that biology is exquisitely targeted to the places where it needs to go to deliver the best result in a disease. We currently have 15 different modalities represented in our pipeline, and that comes by leveraging both internal and external strengths. I hope throughout the course of the day, you'll begin to see how all of these modalities actually start playing into our overarching therapeutic area strategies and how they really are positioning us to deliver best-in-disease solutions. Of course, having all of these modalities is one thing. Manufacturing them at scale and getting them to patients is quite another. Again, Roche has always had one of the largest, most robust, and global footprints of manufacturing of any pharmaceutical company. We've always been very proud of that. It's been a real competitive advantage for us. But in order for us to have it remain a competitive advantage in the future, it requires investment, and it requires modernization. Of course, we have been looking at things like, how do we improve the yield of our existing products? We've had some great successes in doing that. Next-generation manufacturing techniques have allowed us to increase the yield of Ocrevus by more than double and increase the yield of Perjeta by 5x. We are also actively taking advantage of opportunities for innovation across all of the different parts of the manufacturing process. Here's a great example of regulatory innovation, where through the participation in a regulatory pilot, we were able to get simultaneous approval with the FDA and EMA for a tech transfer for Lunsumio. Then, through the regulatory reliance program, that extended to 14 other countries, which means a line improvement that normally would have taken us 890 days took 240. That's a 73% reduction. Those kind of improvements are critical in helping to make sure that we can move as fast as possible to get medicines to patients. That's cool. What's cooler is what is on the other side of the slide. That is a fully live digital twin of our Holly Springs plant in North Carolina. Working with NVIDIA and their physical-scale AI program, we have built the first fully operational pharmaceutical manufacturing plant using a digital twin. This is helping us ensure that before a single beam was put in place, we knew how that plant was going to operate, how those lines were going to run, where efficiencies could be taken, how we could actually make sure that we were making the best use of every square foot. This is incredibly exciting because it doesn't stop at Holly Springs. The idea is that we start there, and then we take those learnings, and we actually scale them across the entire network. That means any time we need to make a tech transfer, any time we need to move production from one place to another, we know in real time how that's going to work, what we need to do, how we need to do it, and how to do it faster. This is an incredible advance. It just doesn't stop with the physical plant of our network. We're also looking at how we can bring in AI technology to revolutionize all parts of our production, everything from packaging and distribution to final assembly of our medical devices and design excellence across all of the different parts of our manufacturing chain to ensure that we are doing things virtually in a way that allows us to gain scale in the real world. This ambition is quite simple, from one site to the network to every individual product. Speaking of devices are going to be a critical part of our portfolio going forward. We anticipate the vast majority of our medicines going forward are going to need a device. One of the things that we've worked very hard at over the last 24 months is ensuring that we have a scalable platform of devices that will meet every single need we have upcoming in our portfolio. These are the six different devices that we have currently moved to a platform model of. These will cover 90% of our future device needs. This, again, allows us to move much more quickly to ensure that we get cutting-edge devices into the market. Technology is helping us here as well. Last year, we announced that we were creating a device pilot plant in Basel. That plant is now live, and what you're seeing here is a live, working example of a digital workbench where we are in real time putting together devices and figuring out what's working, what is not working, and how that will scale across the actual production lines. This technology will now subsequently be scaled into other parts of our network. These are tremendous advances that help us be more effective, more efficient, and move with much more speed. Having these wonderful devices means they fundamentally need to make it to a patient. Being a commercialization powerhouse is just not simply having differentiated medicines, but it's also making sure that we have a personalized and comprehensive physician and patient experience. Technology is helping us here as well. There are sort of two sides to the coin when we look at digital excellence in our marketing performance. We look at the creation of materials, but then we look at the dissemination of materials. By using AI in different parts of technology, we've been able to cut our content creation by 50% and reduce the cost of each individual asset by 83%. Those are resources that can go back into the commercial infrastructure and into our pipeline. We then combine that with our AI-enabled Marketing Automation and Revenue Simulation, MARS. I'm sure they worked really hard to make sure that had a really cool acronym. Marketers do that. MARS actually allows us, across every single channel that we use to interact with patients and customers, it allows us to optimize what message, what materials, what creative we're putting where, gives us real-time feedback on what's working and what is not working, and allows us to create that. This is live and working incredibly well right now in the U.S., and we have plans to scale it globally. I think what's most important about what I just shared to you, though, is that this isn't just about technology for technology's sake. I think you're going to hear the same thing from Aviv. Every time we make the decision to deploy a technological solution in our business, we are actively looking at what is the benefit that it is going to create. What efficiency is it going to give us? What speed is it going to give us? What comparative advantage is it going to give us? We're working very rigorously across the entire end-to-end value chain to make sure that these incredibly important investments are being used in a thoughtful, ethical, and extremely productive way. None of that matters at all if you don't have the right people working in your organization and pulling that through. If I had the time, I would love to talk to you about the entire workforce at Roche, all 100,000+ people who are brilliant, dedicated, committed, and excited about the journey that we're on. I suspect you might not stay for that. Instead, we're going to focus on the people who are here in the room today and who are the folks that you're probably going to be interacting with the most. Many of these are familiar faces to you. I did want to call out three new additions to our team. Wendy, who is our wave Wendy, yeah? She's our new GPS oncology head. Tanya Monga, who's over there too. Hey. Tanya, who's our new neurology and immunology head in GPS. Matt, who is our new oncology PD head. Matt and Tanya have been here a whopping three weeks. Be nice. We're going to all be hanging around. We'll involve them in the Q&A. We're either going to be here at lunch and at the apéro later, and they're all excited to meet you. These are the faces who are going to help bring these amazing medicines to patients. In addition to this team here back in Basel, San Francisco, and around the world, we have hired hundreds of new people, all of whom are coming to us with the skills and experience that we are going to need in order to bring our pipeline forward. While they're not in the room with us today, I also did want to call out a number of the new additions to our early research and development pipeline, including Mark Dawson, who is our new head of pRED. I'm sure all of you will have the opportunity to meet with these folks soon. Now, let's switch gears a little bit and talk about execution, because a great pipeline and the right infrastructure will help you bring a medicine to market, but alone, they actually won't do the most important thing, and that is to get a medicine to a patient. To do that, you need a robust commercialization engine. Let's talk a little bit about how we plan to succeed in that all-important last mile. When we think about commercialization at Roche, we tend to think of it within three archetypes: entrench, expand, and establish. Entrench is where we are maximizing positions, leadership positions in our key disease areas. These can be line extensions into adjacent indications. They can be devices which allow us to increase convenience. But ultimately, it's about commercial excellence and having a fantastic customer experience. Expand is where we are leveraging our existing established disease areas, launching new NMEs, and expanding indications. This allows us to build on our scientific depth and our relationships and drive commercial synergies while we're leveraging the market expertise that we have. Finally, establishing. This is where we're entering into new disease areas with the intent of reshaping the current standard of care. This requires capabilities and infrastructure building, but it also allows us to jump off of all of those foundations we already have put in place while we are increasing our market understanding, building those relationships, and beginning to tackle those unmet needs. Fundamentally, our commercial execution and our pipeline is about one thing and one thing only. That is about disruption. It is about disrupting the standard of care and bringing new medicines to patients to fundamentally change outcomes and reduce costs from global healthcare systems. Now, I'm not going to go into too much detail on this slide, but this is how we view our existing portfolio through the entrench, expand, and establish framework. Again, I'm not going to go into this in detail, but we will use it as a guide. Let's jump into entrench, which consists of some of our key on-market growth drivers. We'll start with breast cancer. First, let's take a look at our HER2 franchise. As we have discussed at previous events, we heard that see the HER2 treatment paradigm evolving, and it's going to move away from that one-size-fits-all mindset of the past. Our HER2 franchise is extremely well-positioned to remain in a leadership position, and we expect to continue to play a very important part in the treatment of these patients. In particular, Phesgo is expected to remain a key treatment option in the frontline maintenance setting. You can see on the slide that we continue to increase our Phesgo conversion rate. As always, whenever we add new countries into the mix, we see a little bit of a downturn. That's what you're seeing here. It takes a little bit of time for those countries to come online, and we expect to see those numbers continue to increase. We expect to end at around 60% or above 60% in global conversion, with many countries already in the 80%s and 90%s from our very earliest launch countries. While we're to reiterate our HER2 franchise outlook, which I know we've told you many times, but just to ensure that there are no changes, we expect the HER2 franchise to peak at CHF 9 billion this year. As a reminder, that CHF 9 billion was at 2024 constant exchange rates, followed by a steady decline through the end of the decade with a solid tail of around CHF 4 billion. That includes quite a bit of Phesgo, about CHF 1 billion for Kadcyla, and then a little bit of Herceptin and Perjeta. We do not expect a Perjeta biosimilar in the U.S. before 2028, and we do not expect one in Europe before 2027. Once again, let me just underscore that our HER2 franchise, we do not expect to see any sort of cliff situation. Let's move on to one of our cornerstones in NHL and building on leadership and novel combinations. Polivy is now firmly established as the standard of care in first-line DLBCL, with rapidly increasing global patient numbers and firmly taking the lead in inpatient share. At the same time, we are looking to unlock Columvi and Lunsumio in non-Hodgkin's. Columvi is now the leading bispecific in second-line DLBCL, which is supported by its very well-regarded OS data. Then, of course, we have the SKYGLO data in combination with Polivy, which has the potential to really change the treatment paradigm one more time. Lunsumio just had positive CELESTIMO data in second-line follicular, and we expect it to continue to be able to help an increasing number of patients as we make that march into earlier lines. Both Columvi and Lunsumio were very specifically designed to treat different kinds of patients, and they have very unique niches that they serve in the market, with Columvi serving more aggressive types, whereas Lunsumio is more suited to the community setting. We very much expect to continue to see great growth in this franchise. Now, let's switch over to Hemophilia A. In Hemophilia, we have set the standard of care with Hemlibra, but we don't intend to stop there. Let's just be clear, that's actually not a super easy thing to do. 80% of patients on Hemlibra today have zero bleeds. We have real-world data for over 10 years showing our efficacy and safety. We have incredible patient penetration, and our patient share continues to grow as we move increasingly into non-inhibitor and older populations. But we're not resting on our laurels. When you have a chronic condition, convenience becomes something important, and we wanted to make sure that despite all of the options that we have on the market for Hemlibra today, we are continuing to increase the convenience. I'm very excited to say that in 2027, it's stuck in my pocket, we plan to launch the autoinjector. Now, this little guy made its first appearance at ISTH in Paris earlier this year in the medical affairs booth, where it was so popular, we actually had to station somebody there because people kept prying it off. You will not have to pry it off anything because we have some available over lunch for you to actually interact with, so you can get a sense of how it feels. This device is designed to ensure optimal therapeutic exposure. It has improved comfort. You can't actually see the needle. There's no visible needle. Of course, it has time-saving administration. We don't stop there either. We are actively working through the development of the second-generation hemophilia treatment, which I always find hard to say, zemocimig NXT007, which we think has the opportunity to once again fundamentally change the treatment paradigm for patients with hemophilia A. To sum it up, we created a new standard of care in hemophilia. We're improving onto that standard of care, and we're creating the next-generation standard of care. I would be very remiss if I were to leave this presentation talking about hemophilia and not mention that the Chugai team that was responsible for discovering Hemlibra recently won the prestigious U.S. Lasker Award. That tells you just how important this molecule has been. Now let's talk about Ocrevus, which changed the game in MS. It continues to be the number one anti-CD20 therapy with more than 525,000 patients globally, and we are well-regarded within the MS community, ranking number one amongst MS patient populations. Ocrevus set the standard in MS, but again, we're not stopping there. Ocrevus' subcut, Zunovo, came onto the market, and you can see the current patient count number. It is increasing exponentially. Ocrevus' subcut is now the fastest-growing anti-CD20 in the U.S., and that's driven by community practices, which is exactly where we had hoped we would get that entrenchment. Once again, we're not stopping there. This is the last thing I have in my pocket, I swear. This is what the OBI looks like, a simple device that you can attach at home and get your Ocrevus treatment twice a year. This is innovation. This is freedom. This is being developed in conjunction with Enable Injections, and we are hopeful that we will be able to have data here in 2027 with our high-concentration device, which is just incredibly exciting. Just like for the Hemlibra autoinjector, this is out at lunch as well, if you would like to take a look. Of course, we can't end a conversation about our on-market drugs without talking about Vabysmo, which has continued to reset the standard of care in retinal disease. The U.S. market has stabilized, and we continue to see branded market share gain. We continue to see significant volume growth in the EU, which is overcompensating for the pricing effects that began at the beginning of this year. Rest of the world, we are seeing strong growth driven by market share expansion as all of our newly launched countries come on board. Vabysmo is the market leader in retina, and that is exciting because we do know that it is just a superior product. We're not stopping there. We are doing a lot in ophthalmology. It's one of our key focus areas. Chris is going to talk to you quite a bit about this this afternoon, but you can see the entry into UME with vamikibart, the data with Enspryng in TED, additional indications for Vabysmo, but also an incredibly exciting pipeline that will allow us to continue to bring revolutionary treatments to patients with eye diseases. Now let's move on to Expand. Here you'll find some of our upcoming launches, like giredestrant and fenebrutinib. In addition to NMEs, we're also expanding some of our existing medicines into new TAs. I think Enspryng is a great example of that. Let's start where I'm sure a lot of you are interested with giredestrant. Giredestrant, as we know, has the potential to become the new standard of care across a broad spectrum of risk groups in the adjuvant breast cancer space. We see that strong potential for giredestrant to become that new backbone because it helps us overcome the limitations of our existing therapies. If you actually look at how we think about the treatment of hormone-receptor-positive, HER2-negative early breast cancer, you can see that we break out the populations by risk groups. These groups are largely defined by what we see in our clinical trials. The lidERA population covers quite a number of those, more than what we see with the CDK4s. They're primarily targeted in the high-risk and the medium-risk group, where you see lidERA going all the way through into intermediate risk. As we think about what the opportunities are for giredestrant going forward, clearly patients initiating endocrine therapy, monotherapy, would be great patients to actually go on giredestrant, patients who are discontinuing, patients who have discontinued either their AI or their CDK4/6 for tolerability reasons, all really great opportunities for giredestrant off the bat. We are also exploring future combinations with CDKIs, and of course, we will also look into moving into the low-risk patient population. We are hearing a tremendous amount of excitement on giredestrant from our physicians and from patients. What gives us additional confidence in the fact that we will be able to disrupt this market significantly is the footprint that we already have in breast cancer. More than 5 million patients have already been treated by a Genentech or Roche product. 50% of the physicians who we believe will ultimately prescribe giredestrant, we've already been in contact with. We have a robust plan for how we will think about access in the U.S., including patient assistance programs to clear those initial reimbursement hurdles. While it's hard to believe that it's already almost October, Breast Cancer Awareness Month is just about to spark. On that same day, we will launch a new patient awareness program to raise patient awareness that there is a new opportunity for them to seek treatment. What I love about this program is the name. It's about damn time. Because it is about damn time that we recognize what these women go through and the difference in the perception of what their treatment is and where they are in their treatment paradigm versus the reality and how giredestrant can help give them a better experience and ultimately better outcomes because that is what it's all about. We do have some exciting new trials to announce for giredestrant, as I'm sure you saw this morning. We obviously have our intermediate to high-risk trial, which we have already talked, I'm sorry, we have our all-risk AI intolerant trial, which we've already talked to you about, the all-risk upfront switch. We also have two new trials that are just coming. Today, I'm happy to announce the AFT-70 NAVIGATE study, which is in collaboration with the Alliance Foundation Trials, and Natera, which will use MRD-guided diagnostics to personalize treatment based on ctDNA status. Just this morning, we announced a combination study with Pfizer CDK4 inhibitor in intermediate- to high-risk patients. This study expands on our ongoing single-arm lidERA studies with abemaciclib and ribociclib to explore giredestrant in combination with CDK4/6s. You'll hear more about this from Levi in a little bit, but we are confident in the profile that we are launching with lidERA. We know we have the right footprint to make that launch a success, and we're actively putting the things in place to ensure that we'll be able to maximize the success of giredestrant going forward. Now let's talk a little bit about fenebrutinib. Fenebrutinib, our BTK inhibitor in MS, has the potential to be the first high-efficacy oral therapy for both RMS and PPMS. I think there is one key point that's actually really worth emphasizing here, and that is that MS is a disease that is characterized from relapsing to progressing. Fenebrutinib is the first and only BTK that has positive phase III data in both settings. RMS fenebrutinib achieved the lowest ARR rate seen in phase III trials to date, paired with consistently positive trends in reducing disability and progression versus teriflunomide. In PPMS, it's the first and only medicine to achieve a numerical benefit in cCDP12 and a head-to-head trial against the standard, which is Ocrevus. Over the course of the last number of months, we have spent a lot of time talking to our KOLs about what they're looking for in new medications. One thing is very clear. They are especially interested in therapies that can both prevent relapse and progression. Taken together, we believe that fenebrutinib is extremely well-positioned to have a very solid launch and to once again redefine what great treatment means for MS patients. Let's quickly touch on Enspryng with a differentiated risk-benefit profile in both MOGAD and TED. This is a great example of how we leverage pathways across our entire portfolio, bringing assets into different pathways based on their mechanisms of action. Enspryng, as you know, is an IL-6 inhibitor, and it is maximized for, it is designed for maximal signal inhibition. That has turned out to have great efficacy for both MOGAD and thyroid eye disease. We are extremely excited that we are going to be able to bring this to you. Hopefully, by the beginning, we have our PDUFA for TED in October, and then we have MOGAD at the beginning of next year. Some very exciting new indications for Enspryng. Now finally, let's pause with Establish, because this is where we have some of our most exciting NMEs that we are working on that truly have the opportunity to be disruptive across the largest areas of unmet need in the global landscape. We all know that Roche has a long history of entering into new disease areas successfully. The last 10 years, our medicines rewrote the textbooks in MS, hemophilia, SMA, and in retinal diseases. In the next 10 years, we intend to do so in a whole new suite of diseases, from IBD to Alzheimer's to obesity, type 2 diabetes, and hypertension. Of course, we recognize that these are different types of disease areas that we have been in in the past. We recognize that the scope and scale of these is different. That is why we have been investing ahead of our pipeline in our commercialization capabilities to make sure that when these trials read out, we are capable of bringing them to patients around the world who are waiting. Let's start quickly with trontinemab. Trontinemab has the potential to be a best-in-disease treatment for Alzheimer's, where there is a clear unmet need. In combination with our diagnostics organization, we believe we can truly revolutionize care here. For instance, our new Elecsys pTau217 blood test makes diagnosis faster and easier and more reliable and is far less invasive than traditionally used. Let's not forget, and it's kind of surprising to know this, that three out of four Alzheimer's patients are actually undiagnosed. There's a huge diagnosis problem with AD. Largely, that's because you don't have great or easy diagnostic solutions, and treatment opportunities have been quite limited. When you have both of those things together, the opportunity to unlock that disease area really reveals itself. We believe that trontinemab has the opportunity for true differentiation, both from an efficacy and safety as a next-generation amyloid antibody. The phase III trial, TRONTIER, is now fully enrolled, and that is well ahead of our plan. That speaks both to our efforts across R&D excellence as well as the extreme excitement from the Alzheimer's community. We're looking forward to readouts in 2028. Next is prasinezumab. Prasinezumab has the potential to become the first disease-modifying drug in Parkinson's. Parkinson's was one of the few remaining white spaces in pharma, where there really aren't a tremendous amount of treatments. The current treatments that are on the market really only provide symptom relief. That's only temporary symptom relief. We've taken all of the learnings from our phase II studies, and we've optimized our phase III, and we're really excited to see what might happen here. Clearly, PD studies take a little bit longer. We won't see the results of this until 2029. We continue to think of it as a high-risk, high-reward trial. It's undeniable that if we were to see results that were positive with this phase III, that we would change the lives of millions of Parkinson's patients. Next, afimkibart, which is currently in phase III trials, both for Crohn's disease and ulcerative colitis. We will get that first UC data expected early next year. afimkibart has the potential to overcome the efficacy ceiling that we've seen with prior treatments and to have the, and it comes with a very convenient profile. We are excited about afimkibart in UC and Crohn's. However, we know we are also exploring it in other disease areas, which Larry will give you an update on later today. This is only our first entry into the IBD space. We also have a number of earlier assets, including a p40 x TL1A bispecific and a number of other programs in our early research and development. This is an incredibly deep area of unmet need and one where we have really seen an efficacy ceiling that has been hard to crack. We expect afimkibart to be the first one to make a run at that. Hopefully, our pipeline will come in with even more revolutions for patients. Of course, we couldn't end our segment without talking with CVRM and where we are advancing that pipeline. Last week, as Bruno mentioned, you saw the very strong phase II results for enicepatide in type 2 diabetes with an HbA1c reduction of 2.65% at 48 weeks. More of that on Manu in Manu presentation, but it's clearly something that we're really eager to play out. We are very proud of the phase III program that we've built. We believe in the last number of years, we've built an industry-leading CVRM portfolio, which includes petrelintide, CT-996, as well as enicepatide, and of course, the phase II combinations of enicepatide and petrelintide. Beyond this, we're developing pegozafermin for MASH, where it has the potential to be best-in-disease in a new market with a strongly growing growth trajectory. To round us off, zilebesiran has the potential to boost adherence via twice-yearly dosing, offering continuous blood pressure control. This trial, I have to say, is enrolling like gangbusters, which is a great indication of how excited the HCP community is about it. There's lots going on in CVRM, and from 2028 forward, we expect multiple phase III readouts across the portfolio. Now let's quickly end with our future growth opportunities, and we'll talk quickly with a review of our updated consensus slide. This is a slide that you're very familiar with seeing. I want to just quickly level set you here on what you're seeing. The values are based on the post-half-year 2026 consensus numbers from the sell-side analyst models. The last time we showed this at Q1, we used the full year 2025 consensus models. The timeframe is still the same, from 2020 to 2030. You can see that through consensus models, we see a biosimilar gap of CHF 6.2 billion. You can also see where the consensus sees our sales growth coming from in that same time period with a CHF 16.3 billion number. Those things compare very favorably. If you look at where our growth is coming from with our on-market portfolio of Vabysmo, Itovebi sort of lead the pack here. When you look at the pipeline, you see relatively modest numbers in there for quite a number of programs that we're talking about. In terms of potential upside, there are a large number of our programs that aren't even represented in these models yet, which again provide significant additional upside. This is why we feel very confident that we will continue to grow through the end of the decade and beyond. This is a new slide for us. We're sharing with you our 2027 key news outlook. You can see right away from looking at this slide that it is going to be a very busy year from a regulatory perspective. Our launch teams are going to be working on overdrive over the coming year. But we also have a number of important and exciting clinical results coming. As I've said in our previous slides, over the course of the last couple of years, we've shared with you a lot of plans. We've talked to you about what we're going to be doing. But what's important to know is that those weren't a vision on a piece of paper. That was a plan in action. Now all of those actions are paying off. Our pipeline is in a materially different place than it was. Our core capabilities, which were always strong, have been further strengthened and expanded to get ready for our new disease and launch areas. Our commercialization engine is revving up and is ready to go. We are excited to bring these patients the medicines that they need and deserve. We're going to be thrilled to share with you the results of how that happens, the value that it creates, and most importantly, the patient outcomes that it delivers. With that, I'm going to turn it over to Levi. I'm going to take Bruno's glasses in case he needs them. Well, thank you, Teresa. Good morning, everyone. It's always a pleasure to be able to describe our progress in R&D performance at Pharma Day. As Teresa mentioned at the beginning, it's now been over three years since we've introduced this enterprise R&D initiative here at Pharma Day. In addition to updating you on the year-over-year progress, we'll, at various points, comment on progress made since we embarked on this journey back in 2023. The R&D Excellence Initiative, as you know by now, has two overarching objectives. One is to deliver many more transformative medicines by 2030 so that we can easily achieve our pharma ambition of 20 such medicines by the end of the decade. In the process, we want to achieve top-tier, top quartile productivity across the biopharma sector. The first big unlock for us emerged when we defined exactly what we mean when we say transformative medicines. That archetype allowed us to develop an evaluative framework through which we could assess the transformative potential of every asset, regardless of where it sat in our end-to-end R&D pipeline, but also regardless of whether it was an internal program or a business development candidate. Of course, we call this framework the bar. It has five key elements that are shown at the top of this slide. I presented the bar in detail in past years, so I'm not going to walk you through each of these elements again today. However, the bar has arguably become one of the most ubiquitous manifestations of R&D excellence. Over the past three years, we've implemented, we've fully operationalized the bar so that it can recursively identify the programs that are worthy of investment and attrit those that don't. We also reconfigured our governance around the bar in general, but with a particular emphasis on focusing on those gating experiments that interrogate the crux of programs and also on designing clinical trials, especially late-stage clinical trials that are highly likely to be successful if our underlying hypothesis is correct. The ultimate goal here is for our end-to-end pipeline to consist entirely of candidates that either already clear the bar or have strong potential to do so. We've now implemented six solutions designed to improve our R&D productivity and, most importantly, to be on track to achieve that pharma ambition. Three of the solutions are shown on the top, and those are the bar, reconfigured governance, augmented external innovation. Those are really about boosting the effectiveness of our R&D engine. Then two solutions at the bottom, ambitious productivity objectives, evolving our R&D engine. Those were really about making the engine more efficient. Lastly, there was a solution that aligned employee and team incentives around this approach. Let's take a look at some of the results that we've seen from application of the bar. Our clinical pipeline has certainly undergone substantial curation since application of the bar three years ago, and Teresa introduced this, but we're looking at a more detailed look. The numbers in red indicate attrition, and the numbers in green represent additions. You can see that we've had a lot of both. Some of the attrition is the usual ebbs and flows of drug development, but a lot of it resulted directly from the bar assessment. Similarly, the additions have been quite deliberate, and many of those come from business development activity. I'll come back to that later. You can see that whereas we had over 80 clinical development candidates in 2023, we're down to about 65 candidates today. Of course, the goal here isn't actually to shrink the pipeline. R&D effectiveness requires sufficient shots on goal. We certainly plan to grow the total number of development candidates again as we go forward. The idea here was to maximize the quality of the pipeline in addition to the quantity of the pipeline. Now let's take a look at how we're doing in terms of pipeline quality. We'll start with some very basic metrics, which we think support the premise that we are boosting pipeline quality. As you can see on the left, the proportion of programs with best in disease potential, it was already reasonably good, but we're continuing to increase that percentage. On the right, you can see that although the total number of clinical pipeline assets went down by 15%, our portfolio value projections have gone up substantially. They're actually now really at an all-time high since we've been tracking this measure. In keeping with that, you can see that our average peak sales have gone up by almost 80%. Our total portfolio value has gone up by over 70% since the year-e nd 2023. Here's some data we showed you at Pharma Day last year. We're looking at a directional plot of team-assessed PTS of programs going into pivotal trials on the Y-axis. Then we're looking at adjusted net present value of the new molecular entity on the X-axis. The light blue dots show programs that were initiated before implementation of the bar, and the dark blue correspond to programs started after initiation of the bar. What we showed last year was an upward and rightward shift. In other words, we're seeing higher team-assessed probability of technical success and higher adjusted NPV after initiation of the bar. Now we can add some NMEs that have moved into phase III since Pharma Day last year. As you can see, all of them are at or above the threshold that we've set for ourselves that we want to see for team-assessed PTS headed into phase III. Now, you could say, well, some of these are not yet mapping all the way to the right on the adjusted NPV side. As we add additional phase III trials for new indications, line extensions, as long as those studies continue to have the same reasonably high PTS, the adjusted NPV will increase over time as well. Overall, application of the bar and these other governance reconfigurations and other measures, they seem to be having the desired effects of enriching our pipeline with molecules that have transformative potential and with pivotal trials that have a reasonably high likelihood of being successful. Of course, the ultimate goal is to boost our actual phase III trial success rates. As you saw in several of Teresa's slides, there's a lot of green check marks. We're still early days, but it does look like we're starting to make progress on that front as well. In fact, our numbers thus far this year of phase III successes are tracking above 80%. Another guiding principle of R&D excellence from the start was that we would apply the bar to identify and invest in exciting candidate medicines wherever they might be found, whether they're in our internal pipeline or in the broader biotech ecosystem. Indeed, bringing in great candidates through business development has been a substantial component of R&D excellence. The business development efforts have largely, in the past several years, involved clinical stage assets, and there are a range of them. I won't go through all of them, but they start with, on the left side, the deal with Alnylam, which brought in zilebesiran that Teresa mentioned. That's our RNAi molecule against angiotensin for hypertension. The deal with Telavant brought us afimkibart. Of course, our acquisition of Carmot brought us in enicepatide and also CT-996. You've got the Regor deal that brought us a CDK4/2 inhibitor, and then it goes on and on. Zealand brought us an amylin. We acquired 89bio, which now brings us an FGF21 analog for MASH. Then, of course, in the past year, we had MediLink, which brought in the very exciting B7-H3 ADC, and of course, a collaboration with Nurix for BTK degrader. What you can see with these is that not only are they really fortifying our pipeline with medicines against foundational targets across multiple therapeutic areas, they've also really boosted the breadth of modalities that are going to be so important to drug these targets. Teresa mentioned this as well. There have also been some interesting platform deals, including Poseida, which brought in the allogeneic CAR-T modality. More recently, we've had deals with Duality Bio, which is focusing on novel antibody-drug conjugates, Earendil, novel bispecifics. We've always been active in business development. This track record shows the results of an intentional strategy to augment our pipeline. That strategy is anchored in the bar, and then it's further shaped by critical choices within specific therapeutic areas and disease areas, as Teresa outlined earlier. Now let's discuss progress towards our second objective, which is top-tier R&D productivity. It's often helpful to frame this conversation using our so-called R&D productivity equation. There are five key measures in this equation. The three in the numerator, volume, value, success rates. Again, that's about R&D effectiveness. The two in the denominator, cost and cycle time, about R&D efficiency. You can see that over the past three years, we've made significant progress in each of these measures. I've already talked about the sculpting of our clinical pipeline. In addition to that, 47 assets have progressed into GLP toxicology studies across our two early R&D units. That should help enhance our early clinical pipeline further over time. Now, I've already mentioned the increase in projected average peak sales from clinical stage assets, as well as the increases in both team-assessed PTS and actual phase III success rates. Looking at the bottom, we've now allocated, reallocated, I should say, a total of CHF 1.6 billion across the R&D enterprise to more directly support the pharma portfolio. On average, we've reduced our end-to-end cycle times from lead identification to product launch by approximately 19 months. That was progress over three years. You can see here that a meaningful proportion of this progress occurred since Pharma Day of last year, whether we're talking about entries into GLP tox or reallocated savings or cycle time compression. Now, let me double-click on the cycle time topic for a moment, because two years ago, you might remember, we selected three programs that we felt were exemplary of transformative opportunities: afimkibart, enicepatide, and trontinemab. We created a one-time fast-track designation. That designation unlocked additional resources and front-loading of activities. Actually, each of those fast-track programs achieved meaningful acceleration as a result. Those are indicated by dots on this slide. What we're seeing now is that the learnings from the fast-track program and the successes that we're seeing are being applied, along with other R&D excellence solutions, to many other programs that were not part of the original fast-track group. As a result, we're starting to see similarly impressive time savings for those programs as well. There are two examples shown on this slide: zilebesiran, which Teresa mentioned, and prasinezumab. Moreover, really, a significant number of our recent phase III trial starts are enrolling ahead of expectations. This is exactly what we hoped to achieve as a side benefit of the fast-track designation. It illustrates how R&D excellence is allowing productivity gains to impact multiple pipeline assets. Now, there are multiple other efficiency-oriented initiatives stemming from R&D excellence that I don't have time to describe today. I will mention one very important effort that's currently ongoing, which involves the agentification of many workflows across the entire drug development continuum. Now, obviously, this has to be done thoughtfully because we're in a highly regulated environment with drug development, the regulatory requirements. We are making good progress, and we certainly have high hopes for the potential impact of this on R&D performance. This is something we look forward to describing in much more detail at a later date. Our enterprise R&D footprint in pharma is now nicely aligned with the critical existing and emerging global biomedical ecosystems. You can see this in the U.S., in Europe, and of course, China. Recently, I think Bruno mentioned this, we had the grand opening of our Boston Innovation Center. That's going to focus specifically in CVRM, but also on applications of AI and ML in drug development. In a moment, Aviv will describe applications of AI on the drug discovery side. To support the expanding applications of AI, really across the R&D continuum, we're doing a number of things, including building in-house data centers in both the U.S. and Europe that together will house more than 3,500 NVIDIA Blackwell GPUs. In conclusion, three years into R&D excellence, most of our proposed solutions have been fully implemented. Application of the bar is allowing us to assemble a pipeline of assets with transformative potential, and our R&D engine is also becoming more efficient. Our pharma ambition for the current decade is to launch 20 transformative medicines by 2030. We're halfway there with the potential for two more approvals in the foreseeable future. In our clinical pipeline, as Teresa mentioned, we have up to 20 NMEs that could potentially launch by 2030. What's nice is these clinical candidates are distributed well across our five therapeutic areas. You'll hear a lot more about that after lunch, of course. As I mentioned, our portfolio value projections are at an all-time high. Yet, as Alan will point out later, our ongoing financial discipline keeps us well positioned to do more deals if and when exciting opportunities emerge. We're not done. There's actually a lot of work ahead to turn these pipeline molecules into medicines, transformative medicines, and then to achieve sustainable top-tier R&D productivity. We do believe that the evidence that we're seeing shows that we're on the right path. With that, I'll turn it over to Aviv, who will tell you about our drug discovery efforts and how AI is boosting those. Thanks, Levi. These transformative medicines, the ones that Teresa described, like giredestrant or fenebrutinib or Ocrevus or divarasib, they're really amazing. We know that because we discovered and we developed them. We also know that discovering and developing these transformative medicines remains really, really hard. It's not just the case in Genentech and Roche. It's not just about one thing. Every step in making medicines is hard. All the steps have to align. It has to be the right target and the right medicine and the dose and the patients. All these things are not just hard in the sense that they're time-consuming. Most of the time, they actually end up not succeeding. That, of course, begs the question, why? There is actually four specific ways in which biology and medicine are hard. The first one is that all the numbers are huge. They're bigger than the number of patients or cells on the planet or the atoms in the universe. The second is that the system spans many scales from atoms to patient populations. The third is that we can measure this in many different ways, but not simultaneously. Finally, everything is super varied. We never do the same thing exactly twice. That's actually why it's so hard. It's hard to predict, and it's hard to generate, and to interpret, and to execute all this work. That's why it's hard to make transformative medicines. Very fortunately, all of these are also things that have the properties that make AI really, really relevant. Yeah, the space of possibilities is nominally huge, but AI can reveal the true dimensionality. Yes, you have many, many scales, and all the transformations are nonlinear, but AI can actually translate between them. Yes, you have these many, many different views, but AI can integrate them. Yes, you can't simply automate it, but agentic AI can operate autonomously. The question is, can AI take something that feels almost intractable and most of the time fails and somehow make it better? Like any difficult problem, people usually argue about what's the most important thing. Some people will say it's all about the AI model. Some people will say it's all about doing the right experiment or clinical trial. Some people will say it's all about how you bring together all the data. Of course, they're all right, but they're all a little bit wrong. Each of these is necessary, but on its own, it's not sufficient. What we need is actually to put these together in what we call the lab or the clinic in the loop. You start with an experiment, you collect data, you train a model, we use the model to define the next set of experiments, and then we iterate our way through drug discovery and development. This is what we've been building in Genentech across the full arc of drug R&D, target discovery, drug discovery through lead identification, or LI, and lead optimization, or LO, preclinical work, and clinical trials. We've been on this journey for about six years now. Roughly speaking, our first two to three years were really about just building. Then, while we continue to always build and iterate, we focused on deep deployment and delivery. We're just about now reaching that point where you can start describing substantial tangible impact on the research portfolio and now starting into the clinical portfolio. That's what I'm going to talk about today. Building and deploying this requires not just AI models, but many other interconnected investments and innovations. We established our science strategy in late 2020. We invested very heavily in internal capabilities, including new top-notch large AI teams. We built a center for large-scale iterative data generation. We consolidated a large computational sciences center of excellence. We established a lot of collaborations with NVIDIA and OpenAI and Recursion and Medra and Anthropic and others. They span many things, large-scale data generation. We have one of pharma's largest AI factories, as Levi described to you. More recently, we started building autonomous AI-driven labs. Over time, impact compounds and grows. Now we can start and look at where our pipeline stands. Over the last three years, our LI entries grew 2.5-fold. Those gradually transitioned to 75% growth in our LOs and a 50% growth in our internal clinical candidates by the end of 2025. We absolutely expect these numbers to look substantially bigger by the end of 2026, so this year. This means that the early increase that we saw in LI entries was not just volume. Volume is easy. It was high-quality volume because it became an increase in LO, and it's becoming an increase in clinical candidates. There's two other important things that you should bear in mind. The first one is that this entire shift happens with nominally flat resources, which in real terms actually means a slight decline in resources. We didn't actually grow these programs by spending more. Also, even though there's much higher volume, especially this 2.5-fold increase in LI, we didn't actually clog our pipeline. In fact, our cycle time in LI roughly halved during the same period of time. You can also look at this in a slightly different way. Ultimately, any work on projects should lead to an impact on a portfolio decision. A critical contribution to a decision means that something from AI, an analysis or a prediction or a scaffold or a molecule or a modification, was material or decisive. A supportive contribution means that AI and computation had a meaningful impact on our confidence in the decision, along with other things. An informative contribution means that it was relevant information, but it was more peripheral to the ultimate decision. In the past full three quarters, across all the phases and categories, AI and computation impacted in one of these three ways 40% of our decisions, and is steadily growing, and about 40% of these contributions were critical. Their nature along the pipeline, though, varies. In target assessment and in phase zero, the critical contributions are more common. In LO, we see the largest impact overall with almost 70% of the decisions. What I'm going to do with the remainder of my time is actually take you through illustrative examples of what's hiding under the hood that delivers these kinds of impacts that I just shared. I'm going to work step by step through the stages. I'm going to start with target discovery and assessment, which leads to LI entry. Our increase in high-quality LI portfolio entries is based on a scientific shift in how we discover targets. We use this thing that we call AI-enhanced pathobiology maps. In this kind of approach, we rely on an input of large-scale, high-resolution data, lab data and human patient data. That includes a lot of things, genetics and disease atlases and perturbation screens and more. Also our whole trove of historical data. We put these into an AI that allows us to derive this AI-based disease model that we call a pathobiology map. We also by now have an agentic system that allows our scientists to wade through the models and ultimately helps them generate pipeline targets at this unprecedented scale. Now, lab experiments follow in both large and complex bespoke ones, and they can feed back, but that's not enough. Actually, we also needed a new organizational process that would allow us to operate in this way. We weren't used to operating at such volume. We established something that we call the TAP teams, which is a cross-functional system that allows us to ensure that target assessment is speedy, that it's guided by experts, and that we direct resources in the right way. Okay. Our first pathobiology map was for IBD, and we queried it originally for targets and target combinations that were predicted to break the efficacy ceiling. The first version of the IBD map was built over a year, essentially, between 2023 and 2024. It relied heavily on human genetics and cell data. By now, we're multiple versions downstream. In the next version, it incorporated a lot of spatial data from the patient's tissues, including from our own clinical trials, AI models that can find multicellular communities and agentic enablement. If this first effort took a whole year, just a year later, we built maps for four neurology diseases in six weeks. By now, we have maps across most of our key diseases, and we continuously update them. That's how we get to 2.5-fold growth in AI in two years. The ethos of lab in the loop is that you always have to iterate and be better. For this, we innovate in several ways. First, a couple of years ago, we leveraged the fact that we have these capabilities now in data generation to launch an initiative that we call Foundational Datasets for Foundation Models, or FDFM. In FDFM, we collect datasets not to answer just one biological question, but actually in order to get a better foundation model of the disease overall. We invest in data collection and partnerships across different pillars, sequence to functions, perturbations with high-resolution readouts, patient tissues from key diseases, clinical genomics. Then we train our foundation models so that they can answer questions beyond the data that was actually measured. For example, in the perturbation realm, we can ask questions like, in this disease, if I agonize protein X, what would happen to the tissues? Or if I want to shift the tissue from this to that, what should I actually be agonizing? Of course, there's a challenge that's hidden here. It's the fact that it's actually really hard, and it's sometimes completely impossible, to really design upfront the most informative set of experiments that would allow us to learn a great model. This is where something called active learning comes into play. I'm using here an illustrative test case where we wanted to learn a model of the impact on the whole cell expression of any of 180,000 pairs of knockouts that exist in a universe of 600 phenotypically impactful genes in macrophages. Naively, this would require doing 180,000 pairs of perturbations and measuring 18 million cells. Instead, we're initiating just with 700 perturbations, and we let the model learn from them. Then we perform the next experiment with 600 pairs that the model told us to run. We compute this against simply choosing 600 random pairs out of the 180,000 in each iteration. We do it for three iterations in this example. It's actually a very hard baseline to beat by computer science theory. And yet, the active learning trained models outperform at every iteration. And it actually identifies cell states and gene programs that the baseline never covers. Third piece is that we need to operate scientifically when there's all these data and tools and models, and there's no one size that fits all. This inability to standardize or automate has actually been a massive historical bottleneck. This is where our agentic system called Target Nexus comes into play. It has access to all our tools and our agents and foundation models and foundational datasets and pathobiology maps and project data and our entire history of decision-making and projects from the past several decades of drug discovery. It assesses hypotheses across all the dimensions that are critical for drug discovery pipeline: causal biology, drug discovery, the context of the portfolio, as well as the competitive landscape. We're on track for Target Nexus to be part of 80% of the research portfolio decisions by the end of this year. Now, before I fully turn to the molecule making that comes next, I want to spend a moment on the transition or the interface between target discovery and LI, which can merge through essentially what we call AI-driven phenotypic screens. An example of this comes from our work with Recursion in the past few years. In this example, we started with a very important, well-known, undruggable pathway in cancer. We defined a cell state, and we set up a lab system. Then we took 250,000 compounds, small molecules, and looked at their effect on the cell image called the cell painting assay. We did the same also with the whole genome CRISPR screen. We also took a whole genome screen and looked at the cell's gene expression. AI turned these into two models: a phenomap and a transcriptomap, and also put those together and helped us identify not just gene targets, but actual compounds. By now, these compounds have passed multiple downstream steps through very high-bar validation and discovery cycles. It discovered both new biology and actually a radically novel mechanism of action that shows on pathway activity on this undruggable pathway. We confirmed it in vivo. This mechanism of action is so unusual that I can absolutely say confidently and with a straight face that we would not have discovered it without AI-powered large-scale phenotypic screens. Okay. It's great. But even a 250,000 compound phenotypic high-content screen is a lot to run. There's a way bigger universe of molecules out there. We developed AI for high-content virtual cell screens. First, we train a foundation model over compound screens with microscopy images or molecular profiles. These are things that were measured in a lab. Then we take a known important target or a compound that we want to optimize, and we generate its phenotype experimentally. Finally, we run a virtual screen over billions of compounds with our model as an oracle to identify compounds that mimic the same effect. Again, an illustrative example. We use here a model that was trained on 100,000 compound public screen. We took one of our own PI3-kinase inhibitors to generate an anchor phenotype. Then we ran a virtual screen for about 4 billion compounds and synthesized 400 compounds selected by the AI out of these virtual hits. Then we tested those back in the lab in three rounds of experiments of increasing specificity, the cell morphology they cause, whether FOXO3A moves to the nucleus, and finally, direct biochemical inhibition of the kinase. The results were really strong. It's not just that these are real hits. These are actually multiple novel scaffolds. Okay. With this, let me turn to molecule making. Here, AI again impacts every step from drugability and target assessment, hit generation and expansion in LI, and the optimization cycles overall. I'm going to focus especially on small molecules today. I'm going to touch really lightly on biologics, even though we do just as much work there. The same principles we also apply to peptides, to gene therapies, and to cell therapies. To start, AI is really crucial for identifying druggable binding pockets. For example, this is a proprietary model that we call Boltz- Jump. We see it on the left. It combines co-folding with AI-based sampling, and it generates a plausible ensemble of protein conformation. Then it supports the design and analysis of experimental data like HDX-NMR. That HDX-NMR guides the refinement of the conformational ensemble. Now we can bring again the pocket finding algorithm on the refined ensemble to find more druggable pockets and repeat lab in the loop. Then on the right, you have molecular hit finding. Here, we start with a defined pocket already. Then we use other models. We have proprietary generative models like VoxBind that can generate molecular design that just bind into this pocket. Or we can go back to virtual screens with proprietary oracles like SAGE to score virtual molecules, not just for potency, not just optimizing binding, but also for their ADME properties. This approach can give us both acceleration and a new scaffold, which is what we see at the bottom. This is actually a first-in-class immunology target. We did it faster, and we got to new scaffolds. Okay. Then finally, this lab in the loop means that we're not just iterating on the molecules. We're actually iterating on the model as well. We can improve the model architecture over time. For example, we can use the biochemical and biophysical assay data and all these protein ligand structure data from all these campaigns in order to enhance the model. We improved VoxBind to leverage synthesizability. That's a difficult-to-say word. And phar macrophore information. We also gave SAGE a better ADME property calculator. Now we turn to LO. Here, we can rely on active learning. You've seen active learning before with biology, now with molecules. That improves the molecules and the model parameters in real time. But by real time, I mean each week. Each week, our models use new experimental results to guide the generation of new small molecule designs. We also use the new experimental results to update our potency and ADME models. When we select a subset of designs for synthesis and experimental testing, we actually balance. We optimize the balance between improving the molecular properties for the specific project and improving the model performance overall for all projects. In this example, which is a first-in-class neurology target, we simultaneously are optimizing for multiple properties. We got both better molecules. These are these green dots that are accumulating on the top right panel. We're also improving the model, which you can see by the improvement in the area under the ROC in the bottom right. Overall, this first-in-class neuro target, same one, AI in hit generation allowed us to look ahead so that when we generate the hits, we don't just think about the binding. We also think about their future ADME properties and other properties that historically we would only get to much later. Then in LO, we used AI to rapidly assess almost 100,000 molecules in less than 35 minutes. Large-scale compute also allowed us to look to about 1,000 compounds by FEP. Overall, we got not only a 40-fold potency improvement, but we also, at the same time, optimized the metabolic stability. Now, as you might imagine, this world of models and tools and data and project context and portfolio context and history is extensive. It requires a lot of skills to operate through. Very similar to Target Nexus, we also have an agentic system for that. Now it's the project context, the SAR, the design goals, all the proprietary models and tools and data and history and legacy and lots of skills. One agentic system harness, reasoning, and orchestration. As I said, I'll be very brief on biologics. I will say the same principles in many methods, analogous methods apply there as well. I'm only going to highlight a couple of areas. As you recall, when we use these pathobiology maps for target discovery and assessment at scale, it also means that increasingly, we have a lot of targets and a lot of ways to combine them. As a result of that, we actually have a growing set of biologics building blocks for the putative targets from our maps. That means we can go to a lot of bi and multispecific formats. We can generate a lot of clinical candidates rapidly. That's all great, but good clinical candidates also have to have low immunogenicity and good pharmacokinetics. In this example, which comes from a series of immunology projects, models that were trained on past data help us predict that an arm would have high immunogenicity and poor pharmacokinetics in patients. We then define a broader sequence space that we can explore with actually scaled experiments, which measure the impact of different mutations on another key property that we don't want compromised, which is affinity. This gives the model sufficient basis to then optimize the sequence again. It retains the affinity, but it cleans up the immunogenicity and the pharmacogenetics risk. Then our antibody engineers and our other scientists work with our internal knowledge marketplace and agent platform, overall called Agency, that by now has many, many tools and skills and agents. In this video, you can see how our antibody engineers set a single goal, design me a better antibody. From there, the agent runs a campaign. It grounds itself in what the lab already knows based on internal data. It processes the candidate designs with internal generative models and scores each of them against three competing properties: binding, affinity, immunogenicity, and developability. It delegates the heavier screens to sub-agents. These designs come back to it. It scores. It optimizes them across many competing antibody properties in one step. It compresses all this sequential modular analysis into single optimization. Even though I'm a really fast speaker, I can't speak as fast as the video. The video is slowed down from the actual action of the agent to the speed at which humans read. This runs and runs. In the end, it can zip up and send you a report. Why stop at a report? In fact, the agent can also send it to an automated lab for synthesis and cloning. In this way, you will close the loop between design and experiments. The agentic world is not limited to the virtual side of the lab in the loop. It's becoming a growing part of the physical part, leveraging physical AI and where we have partnerships with companies like Medra and Anthropic to build those kinds of labs. Okay. Now, the ultimate test, of course, is never in research. It's not in the lab. It's in the clinic. Increasingly, this is where our programs are just starting to arrive. An example is this biologic in oncology, which we believe is a first-in-class opportunity. This novel target was identified for multiple large-scale data sets in 2022. Took about a year and a half to figure it out. Then a molecule had an FPI early this year. LI to dev go was around 18 months. AI also helped us reduce the immunogenicity risk, like I just showed you. This is, of course, not where the effort stops. This is where it really begins when we move to the clinic. But there, we also have to shift from modeling just molecules and cells and tissues to modeling the disease in humans. This is crucial for interpreting trial results. We start with a copious data from our own previous clinical trials to train AI models of disease phenotype and progression, which we then use in order to model the patients for the design and interpretation, essentially, of the next trial. For example, and with this I will close, in geographic atrophy, 3D OCT imaging for more than 2,000 patients from earlier trials allowed us to train AI that predicts lesion growth. Then we use the model to assess what would be the disease progression for patients as they enter the trial. This allowed us to correct for any baseline differences between the study and the control arm. This is illustrated here actually with retrospective analysis on the lampalizumab phase II data. It's painful to even remember this molecule. It was advanced to phase III years ago. Years ago, those phase III failed. This is what you see in the top. That's historical unadjusted analysis. Now, AI-based model adjustment, which we can do today, but of course, couldn't have done back then, would have allowed us to stop the program after phase II. The power that is gained by this AI model is equivalent to a twofold increase in the study size. We now, of course, use this model prospectively. Why stop there? We also use it to close the loop back to discovery. Based on the unique trial data and the model that we train, we define new progression risk phenotypes for GA, such as the ratio here between two features, which we then used in GWAS setting to map new risk loci. Those were not detected in traditional risk studies. We, of course, add this as new input to our algorithms that construct the pathobiology map and close the biggest circle that we can to new discoveries. In sum, what I showed you is that our commitment to is essentially our commitment to AI independence in R&D. If you want to maximize the impact, you have to iterate. You need active learning and reinforcement learning and models and agents and experiments. You need it in the lab, and you need it in the clinic. This iteration means that you have to own the learnings and the improvements that AI can develop. You need to own the loop. We need proprietary AI models, the ability to improve them, not just parameters, but architectures. We need agents and harnesses. We build them based on the best of general purpose AI. They need to be experts in our domain, and they need to continuously learn from our work. We need a massive AI factory for training and inference. Just as much, we need real world. We need the real world of the lab and the clinic. AI makes reasoning a commodity. That actually means that the scarcity becomes lab and clinical reality. This means multiple data generation streams, all at scale, building for iterations in the wet lab, having autonomous lab capabilities, and above all, actually having the same thing: scientists and scientific organizations that are willing to embrace this kind of opportunity and deliver more and better medicines for patients. I'll close with this and turn it to Alan. Yeah, tough act to follow. What can I say? Yeah, I can guarantee you my presentation is simpler. Yeah, perhaps a little bit more intuitive, I guess. No, welcome. Thanks for coming and thanks for being with us. I've seen really a great pleasure for all of us. Before I start, I would like to thank my colleagues. I think when I remember it well, Pharma Day 2025, the share price was on a different level. Yeah, I have in mind the CHF 271. Things have changed, and they have changed really based on tremendous work and effort. You've seen all the opportunities now that arise. You're worried now because we all know, yeah, that share price is driven by two things: pipeline progress. I think we have delivered that, yeah, in a very impressive fashion. There's more to do. We all agree to this. A little bit on my side is now to give you comfort on the financial results because we have to bring that together. We have to have pipeline progress on one hand, and we have to have great financial results to drive the share price. Let's see whether I can help with this. Three points I would like to go through. One is really, on one hand, scalable automation based on AI implementation, but also on the process transformation we're going through. Aviv, I think, has paved the way into this. I think really it takes a whole system to make an improvement. I think we will improve the business process landscape quite significantly. That's what we're going through. I would argue set a platform for AI application. I think that will help us in the foreseeable future. The second point is capital allocation that I will go through. Last but not least, the project. We all went through with the R&D leads, the cost assessments that we have done on the R&D side, which was very helpful, helped us to find additional funds that we can put into new innovation, but also helps us to mitigate potential R&D cost increases. Good. Let me start with the AI strategic framework. I just want to put it into the context of what Roche is doing at the moment on the AI side. We have three levers. One is everyday AI, where we embed AI technologies in our daily work to increase effectiveness. Then we have Reshape. This is what I'm going to talk about. This is where I will focus on today with ASPIRE. It's an ERP program. You might say, oh, that's an S/4HANA implementation. It's not the case. It's really a complete business process transformation that we're going through. Last but not least, the big ideas, and the really big ideas. You've heard from Aviv already, and you got an impression what we work on this. What we are really after, you see on the right-hand side, I think value and impact, key element, people and change, certainly want to improve the tech offering. Then we would like to put that into a context of good policies as we speak. Good. Let me make a quick comment on ASPIRE. This is an enterprise-wide business process transformation that we're going through for multiple years now. Major point is we are now in deployment. I can even say that end of this year, I would argue pharma is basically done, which is certainly very helpful. Why? Because we're building a unified, standardized digital backbone. For the company as a whole, we will harmonize the master data and the processes onto a single core system, as you can see on the slide, really in the middle. How do we do that? You see on the left-hand side, we have 11 master end-to-end processes that we are focused on. To really standardize here, we standardize and harmonize more than 1,000 process activities. These are sub-processes, if you like, in the system. I think really very clearly opens the door very, very well for putting the data into the cloud that we have done already, at least for those affiliates that we have implemented and deployed. I can also say that we are really co-innovating with some of the players that you see on the slide, predominantly with SAP, for cloud solutions. Here I can mention the [clinical] supply management, but I can also mention the batch release tools, where we set the standard in the industry. As said, the status looks good, as you can see on the right-hand side. I think corporate 100% deployed, pharma 80% end of this year. We have, as a remainder, diagnostics, which will take us about two to three years until this is done. Finally, the project is finalized. A little bit of a deep dive on the status. You see that on the left-hand side, 118 company codes deployed out of 273. That's basically the number of companies, if you like, that Roche has as a group. 950 process activities enabled. What is key to me is the template adherence. It's 98%. That means it's not like that we now customize everything for every affiliate. We allow really not many exceptions. Sometimes they're necessary, but very, very clearly what we want to get to a harmonized and standardized environment, I think we get to this. As said, key is that you bring data, processes, and systems together allows you a much better application of AI. Will trigger gains for the foreseeable future. Will they come through right away? No, they don't. Because you deploy, you have to change the processes. There is change management necessary. You still have workarounds. Then you do the next step. You optimize it. You potentially put AI on top, and that delivers the efficiencies moving forward. That's why we say the gains will be realized in two to five years from now, and it will really be an ongoing process. Good. With that, I come to capital allocation, switching gears quite a little bit. Let me start really with the cash flow. I think on everybody's mind, and that's what I'm really in love with. You can argue. Am I really in love with my wife? I'm really in love with cash flow. That's what I can tell you. We have really here, you know that slide, really the net debt level for the last five years. I mean, look at net debts. In 2021, really CHF 18.2 billion. In 2025, year-end, CHF 16.2 billion. Pretty stable. That's the first takeaway. I think what happened in between really matters. What you see is the operating free cash flow. That was really averaged annually CHF 17.4 billion. That contains CHF 12.9 billion in R&D investment, CHF 3.6 billion in PP&E, and CHF 1.4 billion in intangible assets. You see the non-operating free cash flow, with the taxes and the treasury. That's what we have to go through. Then the dividends that you hopefully like with CHF 8 billion per year. I have to mention it, 39 dividend increases in a row. If this year goes well, and that's how it looks like, we will get to the 40th consecutive dividend increase. When you look at the business development, we deployed on average CHF 4.6 billion annually. I think that's really a number you should really have on your mind. Roughly CHF 5 billion per year. You see really that we've outlined the business development stuff, so intangible assets, because that's already implemented into the operating free cash flow. That's why it is really in this dotted box. When we put it together with the M&A activities, CHF 3.2 billion per year, you really get to that roughly CHF 5 billion per year. What we can say here and take away from that slide, and that's the message, I think the financial flexibility for us is roughly CHF 5 billion-CHF 7 billion to do M&A without increasing the debt level. I think that's really the message here. That's what we can really operate from. Even you could say now, oh, Alan, net debt has certainly a liquidity and cash element, which is true. But when you compare then the gross debt level from 2021 to 2025, that is also stable. So my message remains valid here. When you go to the P&L and when you really go to group sales, core OP, and the margin at the very end, and I know you're all big fans of the margin, first I have to say for the sake of completeness, we have here the development in constant rates, but we also have it in Swiss francs. So you have both views here. Very clearly, operating performance is measured in constant rates. Let me emphasize this. Let's start with the sales growth. You see really, as we came out of that period where we lost exclusivity on Avastin, Herceptin, and Rituxan, and then certainly we went out of the COVID period, I think really the increase in sales is 7% in CER in 2024 and 7% in 2025. When you then look at the core operating profit growth, I think it's quite striking that we had in 2025 a growth of 14% and in 2024, 14%, 2025 of 13%. I think that really exemplifies what Levi has said, the cost discipline that we have applied all over, but also in R&D. Very clearly, I think a fantastic achievement. With that, certainly the margin moved. The margin moved in 2024 by 2.1 percentage points and in 2025 by 1.9 percentage points. I know you're not super interested about the past. You're more interested about the future. Let me say very clearly, we will defend the margin. If the margin were moving, it will move upwards. Certainly, this is supported by the fact that other revenue becomes more meaningful for us. As said at half year, from roughly 3% on sales in 2026, it will go to roughly 5% on sales in 2030. Certainly, that gives us more financial flexibility. You see very shy on the bottom, there are blue boxes. I think they support the long-term trends that you have seen in sales + 5% per year. That's the CAGR. Core operating profit + 6% from 2011 to 2025. A margin increase of 6.4 percentage points in that long-term timeframe. Really, you see very sustainable development in Roche when it comes to the P&L. As said, if the margin moves, it moves upwards. With that, let's talk about something which is balance sheet-like, I would say. This is really now R&D plus CapEx. Why did I choose that setup? Because you all know we have committed $50 billion of investment to the U.S., which is in the middle of that slide. That's why we are going to start with, from this $50 billion contains $35 billion of R&D investment, really R&D expenses. So 5 x $7 billion, I think, and it contains $15 billion in CapEx. Is that something which is, let's say, stretching our limits? I wouldn't say so. That is anyway something we had in mind and what we wanted to invest. Levi talked about the R&D center. Teresa talked about Holly Springs. I think there are quite some investments we're going through. When you look at the slide in totality, you see that we are pretty balanced with our investments. You see the U.S. in total CHF 64 billion in the last 10 years. You see Europe with CHF 67 billion in the last 10 years of investment. And you see Asia with CHF 21 billion. As we did a significant cycle of investments in Europe and also in Switzerland, we're now coming through a cycle in the U.S. Really, this is really well-timed, if you like. It was relatively easy for us to commit to the $50 billion investment in the U.S. as we are now going out of the investment cycle in Europe and into the investment cycle in the U.S. Certainly, Asia will play a role as well. Very clearly, I think you have a very balanced footprint and we have further investment into innovation in the U.S. When you look really at manufacturing, just to support what Teresa has said, because I got a lot of questions about, Alan, you're investing now so much money into manufacturing and the peptide plants and all of that. How is that playing out? Also here, really building on what Teresa has outlined already. When you look really on the left-hand side and you see really the biologics network optimization since the loss of exclusivity for Avastin, Herceptin, and Rituxan. You see that we achieved a volume growth of 117% since 2018. In the same time period, really, COGS and period cost just rose 19%. What I would like to emphasize is really the efficiencies that we got out of the system. Don't forget, this is already a period where we brought multiple modalities into the game. This is not the simple setups that we had in the past. This is already quite a complex setup. But it also shows there are significant productivity gains that we can work on. I would argue we will do the same for the peptide manufacturing. I see on the right-hand side, peptide manufacturing strategy very clearly will be built on CDMOs on one hand, in-house production on the other hand, very clearly. I think really we're comfortable with what we're going to. You will see a very efficient production landscape moving forward. Good. My last point is about efficiency in R&D. Let me outline here the pleasure. It has been, I know it's a cumbersome project, when you do a cost analysis and cost assessment. At the very end, I think really we did it for patience and we did it really for the sake of innovation. Because all the funds, yeah, that we freed up, we reinvest in R&D. Do not think it will fall through the bottom line. It won't. Very clearly, I think really what you see on the next slide is really, well, all the numbers that you have seen already, but they're split between the one-offs and then the recurring savings, yeah, that we have achieved. The one-offs are very clearly driven by the portfolio optimization. Yeah, these are the programs that we have stopped. This is where we applied the bar, yeah, I think. You see really these savings are significant, yeah, for the time being. Certainly, they will slow down because, well, as you know, the pipeline has already achieved a much higher quality level than we have had before. On the right-hand side, you see really the recurring savings. What they're driven by, I will explain on the next slide. What is important, though, is that this will last until 2030. These are just the measures that we have identified so far. As you can see, that amounts to roughly CHF 2 billion, CHF 1.9 billion that we would like to realize until 2030 and will certainly help us to keep the R&D cost increases in check. What is that containing? Yeah, what's in there? 184 initiatives identified in this, I would even say, historical activity. Let me go through them. I think best fit operating model, 71 initiatives. This is about right-sizing organizational structures, aligning capacity to what we're really doing in the pipeline. Efficient processes, that's what I've talked about. Automation, standardization plays a major role here. At the same time, it's also going against the overhead. Yeah, it's not like that you just have overhead in headquarter. We also have overhead in R&D. That's something we discussed quite well. Then the best cost footprint, I think there are opportunities, yeah, to put really certain activities into low cost. Then the vendor and demand management. Let me emphasize demand management. I think really here it's about drive next-level sourcing initiatives through strengthened demand management and certainly vendor reduction. This vendor management comes so humbly along with the 29 initiatives. I can say these account for more than 50% of the savings, yeah, that we have extracted out of this. As said, we will reallocate these funds in the portfolio. What we have done so far is explained on my next slide. You see that in that colored bar. Now, certainly, all the savings altogether, roughly CHF 2 billion. What you see is that CHF 1.1 billion went directly into the funding of development of external enemies. I think Levi pointed to this. I think we did quite some transactions. Certainly, they trigger costs in R&D. We cover that. CVRM, yeah, I think is a very prominent candidate here. Then CHF 400 million was used to accelerate fast-track enemies, trontinemab, enicepatide, afimkibart, and giredestrant, just to name a few. Then CHF 500 million were invested in future capabilities like lab in a loop, what we've talked about. Very clearly, I think we were able to put this money into better NPV projects. As you've seen, I think the pipeline has really pivoted into a much better value frame. We won't stop with that. With this, thanks for your attention. It's my pleasure now to call my colleagues on stage to go for the first Q&A session. Thanks. I assume this applause goes for everybody who has presented so far. Surely not. We'll start from the front row to the back. Just to remind you, please stick to the topics we have touched in the morning session. If you have pipeline questions, they will get, you know, in-depth airtime in the afternoon. Also to remind people who are dialing, you know, listening, looking in via the web, you have the opportunity to send me questions via the chat tool on the Q&A function. Let's start over there. Ladies, first. Luisa. Thank you, Bruno. Luisa Hector from Berenberg. I feel like we've entrenched here to make it efficient for you. Thank you very much for the presentations. I have a really, hopefully, the easy question is, do you expect an FDA advisory committee meeting for giredestrant? Then I have the bigger picture, trying to pull everything together. It was really, really fabulous presentations. I think eventually my question is, when do drugs declare themselves? Will AI change this? Here I'm thinking about the value unlock. Is it phase II, phase III data? Is it actually eventually a little bit more on the commercial launch side? I'm bringing into this, you've got the bar charts with the pipeline value increasing over time. But everything we've seen in that is largely kind of the assets pre the bar, perhaps. We've had the data on the cert, the KRAS, where you've raised your peak sales. We're looking at it a little bit now pre-bar with some post-bar in the mix, but also perhaps pre-AI. When do you feel that drugs declare themselves? How might this change? What is your level of confidence in that valuation of the pipeline today and how we should think about it when we meet again in a few years' time? Sorry, big question. Wendy, do you want to answer the giredestrant question? Absolutely. Thank you for the great question. For giredestrant, we're on track with FDA discussions right now. We're definitely expecting our PDUFA by November and December, no AdCom. Yeah, no expectation. I think drug declaring itself sounds like an Aviv question. I will at least start. I think, first of all, you're absolutely right. There's a lag time in this industry. It's just baked in. Whatever you're seeing from one part, you can draw a conclusion on any of the others and vice versa. Giredestrant was made as a molecule in the days before what we call modern AI. Having said that, machine learning methods were used way back then in the 2010s as well. It's not without any computational contribution to it. That has existed for much longer. You can draw a conclusion from that backwards into the things you saw there. I think at first, these molecules will have to prove themselves in phase I and phase II trials before you could say it really worked, right? Because, and you need more than one. The fact that one succeeded is that one succeeded. You need a certain volume of them in order to draw a statistical conclusion. I think it is not unlikely that someday down the road, and that day is not now, you would be able to say, my confidence in things starts earlier than it has before. The better the predictive modeling, the more de-risking you can do early, the higher your confidence in something becomes in this world where AI helps us predict better. Today, we can't really predict that well, which is why we only believe things after two phase IIIs or after one phase III, depending on the indication. If we could predict better, we might be able to pull that earlier. It's not today. That's very important to understand. It's not just about us. The data is not out there in the world yet to make that assessment. Yeah. I mean, I think Levi is always fond of saying at the end of the day, you can say it. I shouldn't say it for you. Oh, he doesn't know what he's referring to. You still have to put it in a patient, right? You still have to put it in a patient. You still have to see if it works. You know, you don't ever quite get past that. Just to be clear, what I said was not you would be able to know that in research. It would be you might be able to have more confidence from less clinical data, but even then, you would still have to run through all the trials. One should have humility with AI as well as with anything. The human body is an incredibly complex system on which we have very little data relative to its complexity. I said the first problem we have is the big numbers problem. We have to be careful at overextending our confidence in AI as well. We can move on to the next set of questions. Sachin. Yeah, Sachin Jain, Bank of America. I'll kick off with another big picture question for you, Aviv. It's a great presentation. Mine's a really simple question. Could you perhaps pull out two or three points where you think what you're doing is different to competition? It's very hard for us to assess how that is different. That's the big picture. The second one is for Teresa on the giredestrant sort of launch dynamic slide. Could you just provide a bit more color on the patient assistance program? It's been a recent topic of debate as to how much free drug you're going to put into the system and how would you think about that from a sales curve perspective? It's unusual to have heard that in oncology. Why are you doing that? We'd be interested in. Then two questions you may or may not want to comment on. Pricing. Astra recently launched at roughly [$]300,000 per year. Is that a rough proxy? On regulatory, are you confident you can avoid the ECG monitoring there? Thank you. I'm sorry, can you repeat the last one? ECG monitoring that we've seen for the competition. I'll start. It gives her time to write. First of all, with the disclaimer, I cannot tell more than you can what happens in companies behind closed doors. In particular, I would emphasize that for the research domain on which we usually know the least. On the development domain, we know more. This is my perspective. I believe it's correct, but it's based on this lack of basic knowledge that all of us share. I would believe that it is in three different ways. The first is actually the piece that I focused a lot in the first part of the examples, which is the biological piece. We do know that PTS critically determines that actual PTS, the one that is about the results in phase II and in phase III, it depends on the target. If we are in the territory of first-in-class targets, we really don't know until much later whether we bet on the right horse. I believe that our investment in that, not just quantitatively, but also the style of approaches that we brought, both the experimental approaches and the computational approaches that go with them and the iteration between them is definite, I believe. It is definitely beyond the competition. In many of these areas, we are more innovative than the academic world is innovative in this domain. That is one. The second one is the algorithmic investment. Many, I believe many of the peers do not really invest in the wrong algorithms. They either use algorithms that are in the public domains or ones that you acquire from another party, either through a partnership or as a piece of software that you acquire. Now, we too partner and acquire software. Everyone does. But we develop also our own algorithms. This is particularly critical because of the two points that I made, that you have to update parameters and you have to modify architectures. Both of these things need to happen. We really do not know where this whole thing is going to plateau, but it hasn't plateaued just yet. That is the second consideration. Then the third one, which is the point that I made at the very end, is that I deeply believe that you have to own the loop and the agentic world that we live in now, which was not a surprise for us. You can even see out in the public domain, we released some of the absolute first autonomous agents for the biological domain. That is a learning system. If you use a system from the outside world through a partner, a piece of software that you acquire and so on, you can absolutely use it, but you can't improve it with your own knowledge. That loop of improvement is absolutely critical. You also have limitations on what you would agree to show it, especially if you don't have zero data retention, which is a problem, as you know, today in the AI world. If you do not have ZDR, would you actually let another party see your history of portfolio decisions? Probably not. If you want to then advance it with every portfolio decision, with every scientist that interacts with it, with every shred of information that comes out of your lab, you have to own that. I think our early investment in that is now bearing fruit for this purpose. These would be my top three. We invited Wendy to join us because her team is actually leading the giredestrant launch. We thought there might be one or two questions about that. I'll let her take that and I'll pile on if needed. Sounds good, Teresa. Getting back to your question in terms of how we think about the patient assistance program, but taking that back from one step further, right, in terms of how we ensure that we are product ready, right? Definitely from a launch perspective, we do ensure that a supply chain, we're already given the demand that we're hearing from many of our clinicians, our clinical community, as well as our patients, we're definitely ready for that. Given that this is a Part D drug, we do anticipate that we're going to have to see some of that payer's block as we work through, and hence patient assistance program becomes key. Yeah, and that's actually pretty typical in a launch like this, actually. Now, on top of that, you know, Teresa has also mentioned in terms of how we also build upon this launch. I can tell you that I can't tell you how excited we are. We're all ready for this really the biggest opportunity for us at Roche. When we see in terms of the clinical community, the excitement that we have, we are building, you know, from a market readiness perspective, the disease awareness campaign that Teresa has talked about, it's about damn time, and that will be coming out live in the next couple of days, right at the time of Breast Cancer Awareness Week. In addition, in terms of getting our clinical community ready, not only are we focusing on academic centers, we're also focusing on where the clinical practices are. This is where the majority of our early breast cancer patients reside. That is exceptionally critical as this is our first and only oral SERD that has been launched within the last 25 years. Exceptional opportunity. I do wanted to mention that, you know, as we think about launches, it's very rare that you launch both an adjuvant and metastatic setting at once. In fact, I don't actually remember in my history overall, when we look at all oncology products, that whether there has been any others, right? With this uniqueness, we are definitely ensuring that the clinical community are heavily working with us to understand the patient segments, where we can truly offer benefits for patients, as well as really building up the next level of evidence, as you will hear from Levi and also throughout the program as well. Finally, you know, our team is definitely ready. We're definitely, you know, investing quite a bit in terms of hiring, training, and getting ready for our launch coming up in November and December. Not only is the giredestrant team being the ambassadors, everyone within here together with the panel here are the ambassadors for giredestrant as we prepare for our launch. We're all ready. Pricing, we won't comment on. And monitoring? Monitoring. We've seen no evidence of QTc prolongation, either as a monothyroid or in combination. We don't expect any monitoring requirements. Thank you. Richard? Just to the first row, I think they can hand on. Thanks. Richard Vosser of JPMorgan. Maybe just one extra question for Aviv. You obviously highlighted the 1%- 2% probability that's there in the past and the complexity of human biology. Just thoughts on where that could go from the early experience that you've got in the research side. Where could you take the probability? Yeah, I'll actually reflect slightly historically. A couple of years ago, I was at JPMorgan and I did a panel in some investors, think Bruno might actually remember what it was. I did it together with a colleague. This was a public event. I did it together with a colleague who's the CEO of Generate. We were all sitting there and mumbling, you know, the usual things. We're actually in violent agreement because at that time, we still both believed very strongly that the impact on this biology piece is actually going to take some time and that the impact first will be molecules. I don't believe that anymore. The reason for that is that I think that we have access now in the biology of the human and that the models are also a lot better than we thought they could become. The areas that we really see a benefit from is where the worlds are coming together. It might be also an add to the answer before. It's not just those three things. It's the fact that we're soup to nuts. We're an end-to-end organization. We're a big company. That is a big advantage in the days of AI. We can do discovery work, but we have data from our clinical trials. This ability to be tied in to the ultimate clinical impact and have the layered information, all the levels of organization from the atoms of the molecule all the way to the full patient and have tissue from that patient and clinical imaging from the patient and peripheral biomarkers from the patient and so on allows us now to build models that are essentially like these massive translators. Even though on any given patient, we might not know everything, we start filling in the blanks and then the models become a lot more performant. It's far from a simple problem and it's far from fully solved, but I do think its trajectory is better and faster than I thought two years ago. Maybe just one more question just on the digital twin on the gross impact on the gross margin. Something that might be easier for AI models to help with just on that side of thing. How could we think about the improvement in the gross margin going forward from that to 2030? Yeah, sure, go ahead. Yeah. Look, I think gross margin always was hovering around 80%. To be honest, I'm not in a position to say it really improved significantly with Holly Springs. I think we build Holly Springs as we move. Then we will optimize it. We will see where it goes. I think, but am I, how should I say it, am I concerned about the 80%? No, I'm not. I think we have around this can be stabilized. I guess that's the case. Yeah, I mean, those efficiencies are real. James Gordon, great. Thank you. James Gordon from Barclays. A couple more on giredestrant and one from fenebrutinib as well, please. On giredestrant, I heard the comments about disrupting the breast cancer market, but then also I've heard some comments about physicians liking longer-term data. Are there any good analogs to think about what the launch might look like next year? I think consensus is about CHF 0.4 billion. Is that something that's plausible in terms of how the launch might go? How do you think about the merits of upfront use versus switching? Because you've got upfront use in lidERA, but there's CAMBRIA-1 and EMBER-4 potentially coming fairly soon from Astra and Lilly. Might it appeal to you something where you switch first before doing upfront? Just the funny one quickly was I think you've taken your target across both types of MS to more than CHF 3 billion. How are you thinking about the product's potential in RMS versus PMS? I know we've got some data coming up from another BTK next month from Novartis. Sounds like maybe they would look a bit better on disability, but maybe not quite so good or bad job adjusted on relapse rate. What really matters? Is it the relapse rate or the disability? How have you thought about that? I mean, both matter and both matter a lot, actually. When you think about what a good drug looks like in MS, it has to deal with progression and disability. I think that's where fenebrutinib, frankly, really shines. I think we are very confident in, I'm just trying to think of a word that was other than excited because I use excited a lot today. We're very confident in the uptake in fenebrutinib. I think the clinical data were outstanding. I mean, that above 50%, you just don't see that. The head-to-head against Ocrevus in PMS, again, or PPMS, I mean, that's practice-changing data. I think the consensus right now is hovering at like CHF 1 billion for fenebrutinib. We could get that in PPMS alone. I think we're quite confident in the uptake for fenebrutinib. Teresa, if I just may, as you showed before on the slide, the RMS data, 51% and 59% reduction ARL, it's clearly the best class that we have right now. That matters a lot, especially for the early part of the patients. There's a huge opportunity to market. There's about a third of the patients who are currently undertreated or not treated. That's a huge opportunity for fenebrutinib. On the PPMS side, so the progressive side, that's the other part of the equation. Patients who continue to progress or are over time progressing even under treatment, there's opportunity there. I'm almost going to say on both sides of the equation, there's a huge opportunity for fenebrutinib. We have to say, what other datas come out, we don't know them today, but we have very confidence on our own data that's a best-in-class BTKi. Absolutely. Turning over to giredestrant, as we're looking at where we believe the consensus is on year one, it's very consistent where we think it is. When you think about the source of business in terms of looking at overall, we definitely believe that both the initial new patients versus those that are switching are both equally available for this. If we're looking at across the segments where there are significant discontinuation rates for CDK4/6, tolerability becomes key, especially for patients that you're looking at a long-term treatment. I think that net, if you look at where the source of business is, we do believe that overall it can be either-or. It doesn't preclude us from those that are newly diagnosed patients equally for the simplicity and the tolerability profile that giredestrant will offer through lidERA. Maybe just to put a finer point on the trajectory of the curve, we expect it to look a little bit like a hockey stick. I think the consensus estimates that we've seen are around CHF 600 million for next year. That sounds we're comfortable with that. Matthew, please. Thank you. It's Matthew Weston from UBS. Two questions, if I can. Teresa, in your comments, you mentioned that the trontinemab study is recruiting faster than you expected, but you never bring that 2028 readout date forward. What's the likelihood that we see data in 2027? It's just not possible. On giredestrant, there's a big debate amongst investors as to whether or not Roche would do better on its own in early breast cancer with the launch that you are planning, or whether or not the entry of competitors into early breast cancer with their data could grow the market in totality. While you'll have a smaller share, there could be a bigger slice. Teresa or the team, which would you prefer? Would you prefer to be there on your own for three years in early breast, or would you rather have a competitor driving that market growth to serve? If you don't mind, let's take that. I would rather be on my own and I would rather drive it myself because I actually think we have a really phenomenal medicine that is clearly more targeted, more potent, combinable. It is a really impressive medicine. I think once physicians get it in their hands, and more importantly, once patients begin to experience what it can do for them, I think we will be able to expand that market ourselves. That's just me. No, on trontinemab, I think we're still in 2028. Nice try. I like it. I like the optimism. You can join the R&D excellence team. Yeah, we're still not able to bend the laws of Einstein and time. It takes a certain amount of 18 months. You know, we can't. I think the margin would definitely go up if we could do that. Maybe we do two more. One question each, please, for you, Graham, and then James. I think we have to go to lunch. Great. Okay, thanks. Graham Parry from Citi. Sorry, another one on giredestrant. In terms of the cost and the margin impact into next year, if you think about the launch costs with awareness programs and potentially free drug, is that going to be an incremental positive to margin next year or negative? Question for Alan there. You also talked about the aromatase inhibitor population being switchable. What percentage of that population do you think just doesn't do well on aromatase inhibitors and could be looking to switch? How does that vary across community versus the academic center populations that you'd see? Last one's on divarasib. Do you include adjuvant in your updated peak sales potential? Thank you. On the margin side, yeah, I think I've said everything about the margin already, I think. Defending the margin as I stabilize it or it goes up. I mentioned the point about other revenue. I think we have a couple of things that are playing for us. That's helpful. Honestly, I think nothing more to say at this stage. Is there a possibility for an increasing margin? Yes, there is, as we all know. As said, we have a lot of investments on the other hand that we want to make and that we have to make. Very clearly, I think the probability for margin improvement has increased. We've accounted for all of those costs in our launch planning. Nothing new there. I'll take the giredestrant and divarasib question then. For giredestrant, I think to your point, the different segments across academic and community might look different, as well as across risk patient groups as well, right? When we're looking across, we definitely believe that majority of the patients that will be switching initially will likely to switch over likely will be in the intermediate and as well as the medium patient populations, right? Whereby, A, there's higher met need in the intermediate patient population whereby CDK4/6 is not available and where also the tolerability profile becomes important and as well as the clinical profile, for sure. I do believe that as we work towards really entrenching into the community, this is where the significant uptake will be because that's where the majority of these early breast cancer patients are lying. When it comes to divarasib, your question in terms of where the expansion is coming from, definitely the data coming from K1, and please come to ESMO to join us. We are on plenary session on the first session there, right? The data itself really showcase in terms of the best-in-class opportunity for us and that strengthens in terms of our confidence as we unlock primarily into the frontline patient population. Adjuvant is currently included as well. I'll just make one more addition on your giredestrant. After the break, we'll talk more about some of the trials that are ongoing to look at this. In particular, the question of even aromatase inhibitor and side effects that cause patients to go off aromatase inhibitors. I've actually surprised at how often that can happen. As you know, in lidERA, our disintegration rates are even lower than the control mark. We'll talk more about that after the break. Yeah, I mean, I think we just cannot underscore how important tolerability is in this patient population. It is why you tend to see such poor outcomes because patients don't stay on their treatment. I think this is why we believe that giredestrant is going to be such a game changer. The final question goes then to James. One question, please. Thanks, Bruno. James Quigley from Goldman Sachs. I've got one for Aviv on AI in target discovery. How confident are you that the novel target discovery platform, or how confident are you in the novel target discovery from the AI platform? Have you done any back testing to see whether for established targets, AI gets to the highly established targets, AI gets to the same place? Also, how important is the diagnostics business in terms of helping with data generation in your models? Okay. I'll start by saying what I also said in the answer to the first question. We will really only know for novel targets when we're in the clinic. We can't tell before then with any confidence, and we shouldn't. That's risky, and we shouldn't. At the same time, our pathobiology maps, these disease models, they're not just about novel targets. Sometimes the target is established, and it's about the combination. Sometimes it's about figuring out the right patient population to go after, the indication selection. There's many, many things beyond novel targets, although they are absolutely part of the game. They're not the only part. In addition, you often have a lot of targets that you would think are independent from each other in the traditional way in which people have worked, which lead to a lot of failure, not because the target itself is not meaningful, but because that unmet need has actually already been met. You just don't realize it was met by something else that is already in the pharmacopia. Being able to put targets into programs that operate together, that can be distributed across multiple different kinds of cells and know how to choose and combine between them, is just as important as completely novel targets, meaning the gene name was never known before. To the second part of your question, which is about, and a critical piece of that is, we would spend a lot of time, not just we Genentech, anyone would have spent a lot of time on doing a lot of work to even get a target like that to enter it alive. All that time is resources and money and opportunity cost that is happening. Now that gets shrunk, and you get to assess things really quickly, run them through discovery really quickly. In fact, you have to be very innovative in the clinic at how to assess them really quickly in the clinic because you will have a lot harder for you. Next to your question about doing a retro, essentially, the challenge with retro is that the base AI models that you use today are all contaminated by your prior knowledge. You can withhold from them Genentech own prior knowledge, so you can play out on targets that we tried out and no one in the world knows about. But the moment that they showed themselves out in papers, presentations, conference abstracts, whatever, forget about it, they're in the other day. In fact, just this morning, there was a story like that on another thing that people thought was a big discovery by AI and, oh, maybe in fact, it got to see some human researchers using it along the way. That's not unusual. What we can do is that we can do a lot of assessments based on the history of our past decision making. We have a lot of programs, like everyone in this profession, have a lot of programs that never made it. They didn't even make it out of research, forget about getting to. That is actually a closed garden that models have never seen, that the knowledge is only ours. When we optimize Target Nexus, that agentic system that I showed, when we optimize it, that was a useful source of information for us, but only up to a point. You can only retro so much. You know, time will tell on novel target discovery. We have to balance between novel targets and existing ones, but even this balancing act is something that AI helps us do better. Excellent. I think with that, we will close the morning session. We'll have 40 minutes for lunch, and then we'll meet again at 12:30 P.M. [Break] All right, good afternoon, everyone. I hope that the conversations over lunch were fruitful. As you heard at the introduction, this year our Oncology/Hematology presentations will be a dual act. We're thrilled that Matt Hellmann has joined us earlier this month as Head of Oncology and Hematology Product Development. Matt will cover the hematology side of the pipeline, and I'll start off by walking us through highlights of our solid tumor pipeline. Let me start with an overview of the progress that we've made in both oncology and hematology over the last year. On the left, you can see you're reminded again our five phase III readouts that we've had since the last Pharma Day, all of which were positive. Those include two NMEs that have best-in-class potential: giredestrant, of course, adjuvant HR+ breast cancer, and divarasib in the second-line KRAS G12C- mutant lung cancer. Now, on the hematology side, we also recently announced a positive trial for Lunsumio in follicular lymphoma, and Matt will cover that in his section. In addition to those phase III trials, we've also advanced two oncology NMEs into phase III in the last year. We'll be starting phase III trials with Tam-Peli, our B7-H3 ADC in small cell lung cancer. Then there's the bexobrutideg phase III trial, that's our BTK degrader. That's being done in collaboration with Nurix, and it started enrolling patients last summer. We, of course, continue to apply the bar, as I discussed this morning before lunch. When you look at the 28 oncology and hematology NMEs in clinical development, over 70% have first-in-class and/or best-in-class potential. This bar application, of course, continues to go for our business development decisions. On the right side of the slide, we just remind you of the oncology deals we've done over the course of the year, and this includes a range of both asset and platform-based collaborations. Overall, good progress in fortifying and in progressing our oncology and hematology pipeline. Here's the solid tumor portion of our pipeline. As has always been the case, we strive to ensure that our pipeline is sufficiently diverse in terms of both foundational targets in oncology and the breadth of modalities needed to drug those targets effectively. Indeed, you can see a broad spectrum of modalities that are represented from small molecules and protein degraders through bispecific antibodies, next-generation ADCs, etc. Now we'll go through a few of our solid tumor pipeline highlights. Let's start with breast cancer. We already talked a lot about it, but in recent years, we've expanded our scope, which historically was in HER2-positive breast cancer, to include hormone receptor-positive breast cancer. Those indications represent 70% of patients with the disease. For a long time, the therapeutic attention in the field has focused on a few specific foundational targets, in particular, that drive tumor progression. Those are the estrogen receptor itself, the PI3-kinase inhibitor, the PI3-kinase pathway in general, and the PI3-kinase alpha subunit in particular, which, of course, is mutated in up to 40% of ER-positive cases. Then there are the cyclin-dependent kinases. Those are, of course, a big focus for us, but we also continue to address unmet needs of people with metastatic HER2-positive breast cancer, especially those who develop brain metastases, and we have a potent and brain-penetrant HER2 tyrosine kinase inhibitor for that purpose. For the past several years at Pharma Day, we've presented the biochemical and clinical case that giredestrant can be a differentiated oral SERD. Giredestrant, as you know, suppresses the estrogen receptor activity, first by immobilizing the receptor protein, which prevents it from accessing chromatin, and then subsequently, the estrogen receptor is degraded as a result. This mechanism can produce a deep and sustained inhibition. Moreover, compared to other SERDs out there, giredestrant comes with the highest potency and the highest target coverage at therapeutic doses. What we mean by target coverage is this is a composite measure of drug concentration, receptor occupancy, and receptor degradation in human cells and tissues. Giredestrant has now shown positive phase III results in both the adjuvant and the metastatic settings. We'll start with the lidERA trial on the left. This was the first time that an oral SERD had shown positive phase III results in the adjuvant setting. We had a 30% reduction in recurrence or death at the interim analysis. Although the overall survival assessment was not yet mature, it was trending in a positive direction with a hazard ratio of 0.79 at 32 months. This magnitude of benefit puts giredestrant firmly in the range that's been observed for the combination of endocrine therapy and CDK inhibitor combinations. Unlike those CDK4/6 inhibitor combinations, giredestrant comes with quite a favorable safety profile, as we discussed this morning. In fact, its discontinuation rates are even lower than those that are observed with standard-of-care endocrine therapy. A lot of patients discontinue the CDK-endocrine therapy combos. 30%- 50% of patients discontinue because of toxicity, and then that discontinuation, of course, can lead to reduced efficacy. Giredestrant was filed with the priority review voucher in the U.S., and we expect a decision from the FDA in a couple of months. The evERA trial, which is shown on the right, this was our positive phase III study in the metastatic post-CDK inhibitor setting. As we showed at ESMO last year, giredestrant conferred PFS improvements in both the ESR1 mutant population and the intention-to-treat setting. Giredestrant is also undergoing regulatory review in this setting with a PDUFA date in December. Now we'll talk a little bit about how we're expanding our evidence generation program for giredestrant across several patient segments, because this has come up a lot already, and I'm sure will come up in the Q&A later this afternoon. On the left of this slide, you can see the studies that we've added in the adjuvant setting. One of them is called the kalitERA trial. This trial looks at how well people adhere to giredestrant monotherapy, and in particular, if they can avoid premature discontinuation due to the common side effects of standard-of-care endocrine therapy. This also will assess the feasibility of switching to giredestrant monotherapy after completing a CDK inhibitor-based treatment. A second phase III-B trial is called the novERA trial. This trial looks at adherence and tolerability of giredestrant in patients who are forced to discontinue aromatase inhibitors because of those side effects. Aromatase inhibitors, there are often musculoskeletal and other adverse effects, so sometimes patients don't stay on therapy. In fact, as many as a third of patients may come off of their aromatase inhibitor at some point before they've completed the course of therapy. This is really important to see if giredestrant can improve those kinds of outcomes and patient experiences, really, across the entire early breast cancer risk spectrum. Those two trials were announced previously, but we're now adding two more studies in the adjuvant setting. One, as was mentioned this morning, is a study in collaboration with Pfizer. In that study, we'll look at giredestrant in combination with atirmociclib. This is not the existing CDK inhibitor. This is a first-in-class CDK4 selective inhibitor. The idea here is that dialing out CDK6 activity could offer the potential to reduce toxicities associated with the current generation of CDK inhibitors. Then, of course, with giredestrant, you could imagine a best-in-class type combination. The phase I portion, which is obviously to confirm that they can be given together safely, that will start in Q4 of this year. We anticipate a quick pivot to phase III if all goes smoothly. Finally, we expect to initiate a study that's sponsored by the Alliance Foundation Trials, the AFT Cooperative Group. This is a U.S.-based cooperative group, and this will also be done in collaboration with Natera. The idea here is to use ctDNA-guided optimization for endocrine therapy. We've already mentioned several times, 30%- 50% of patients who are eligible for the current standard of care, which is a combination of CDK inhibitors, they discontinue because of adverse events. More generally, approximately 90% of adjuvant breast cancer patients are ctDNA-negative at the point that they would start treatment. The idea of this study is to see, if you have a patient who is maybe intermediate or high risk, but they are ctDNA-negative, the question is, can they start on giredestrant monotherapy? You would only add a CDK inhibitor for the subset of patients who become ctDNA-positive. The question is that kind of approach non-inferior to standard-of-care CDK-based combination? As you can see, we've already got a robust evidence generation effort across the entire early breast cancer spectrum. Now, in the metastatic cancer setting, we still have additional opportunities in the front line, despite the unsuccessful persevERA trial. We have the pionERA study, which is in the endocrine therapy-resistant population. That population, by the way, constitutes about 40% of the front-line metastatic ER-positive population. Then we have the heredERA study, which is looking at giredestrant in combination with Phesgo in metastatic triple-positive breast cancer. Based on the learnings from both the persevERA and the evERA trials, we've adjusted the required number of events in both of these trials, both pionERA and heredERA. The idea is to increase the statistical power for both of these studies. Now we expect these trials to read out in 2028. We've moved them to 2028. Then finally, on the early clinical development side, we're exploring giredestrant in combination with our own CDK4/2 inhibitor, and more on that in a moment. Now let's move to our PI3-kinase inhibitor, Itovebi. We've, for a number of years, also been bullish about Itovebi here in Pharma Day. As a reminder, it turned out in preclinical studies that not only is Itovebi a potent and selective inhibitor of PI3-kinase alpha, as shown on the left, but the molecule also catalyzes the degradation of alpha PI3-kinase. That observation gave rise to the hypothesis that selective degradation of mutant PI3-kinase could produce a highly robust therapeutic effect, and we certainly saw that. We saw strong clinical results in the INAVO120 study of aromatase inhibitor-resistant PIK3CA-mutant breast cancer. The hazard ratio there for PFS was 0.43. Then, as shown in the middle, there was a seven-month overall survival benefit observed. Discontinuation rates due to adverse events with the Itovebi regimen were reasonably low, at less than 7%. Now we're awaiting the results of INAVO121. That's a study in the post-CDK4/6 setting, which compares Itovebi + fulvestrant head-to-head against alpelisib + fulvestrant in the post-CDK setting. In 2027, we expect INAVO122 to read out, which looks at Itovebi in combination with Phesgo in the HER2+ PIK3CA- mutant setting. Beyond that, we'll see INAVO123, which is the endocrine therapy-sensitive population, looking at the triple combination of Itovebi, the endocrine therapy, and CDK inhibitor. We expect in the future to go even earlier, if everything works out, that we have a phase II trial, which is called neoTOV, which is the same triplet in the early breast cancer subpopulation that has PIK3CA mutations. It's Itovebi, a lot happening. We also are exploring Itovebi in giredestrant combinations and other indications. We also have an interesting castrate-resistant prostate cancer phase III trial, which we disclosed earlier this year. This is actually in PIK3CA wild-type patients, but it's based on a potential new efficacy biomarker. A few words about our CDK4/2 inhibitor, which is called GDC-4198. We enlicensed this molecule from Regor Therapeutics, and it's currently in early clinical development. There's evidence from multiple sources that suggests that CDK2 cell cycle activity may provide an important mechanism of resistance to the current generation CDK4/6 inhibitors. GDC-4198, on the one hand, it's highly potent against CDK4, but it also exhibits substantial activity against CDK2. On the waterfall plot on the right side of the slide, you can see that we've seen intriguing preliminary activity in the post-CDK setting. We're now looking at GDC-4198 in combination with giredestrant, compared head-to-head with abemaciclib in combination with giredestrant in a post-CDK population. That study is called MoonROSE, and it recently enrolled its first patient in the phase II stage. We'll close out the breast cancer section with work that's ongoing in the HER2-positive setting. Unfortunately, despite a lot of advances in this segment of breast cancer, it's still the case that 30%- 50% of patients with metastatic HER2-positive disease go on to develop brain metastases. ZN-1041 is a highly selective brain-penetrant HER2 TKI. Unlike other HER2 TKIs, ZN-1041 is not a substrate for efflux transporters that's shown on the left side of this slide. This may account for its high retention in the central nervous system. Previously, we have described high response rates for ZN-1041 in combination with Xeloda and Herceptin in a China-only study. At ASCO last year, we also shared results from the global patient population, and some of those data are shown on the right of the slide. These are heavily pretreated patients who were given ZN-1041 in combination with either T-DXd or Phesgo. Both combinations were feasible. They were well tolerated, and there was some encouraging, albeit early, clinical activity. ZN-1041 is now being studied head-to-head versus standard-of-care in a seamless phase II/III design, which is recruiting very quickly. Now we'll switch gears to our RAS portfolio. There's certainly been a lot of exciting innovation in the KRAS space, therapeutics targeting mutant KRAS, RAS-on-based mechanisms. But it's also the case that toxicity, combinability, progression to resistance, these all remain important barriers that we need to overcome. Across really all of our R&D units at Roche, Genentech, and Chugai, we're building a pipeline of RAS inhibitors that encompass several different mechanisms of action and modalities. The goal here is to be able to deploy best-in-class molecules and novel combinations that could produce additional efficacy and tolerability improvements in the future. Divarasib is our most advanced KRAS asset. We continue to see divarasib as a potential best-in-class G12C inhibitor. Certainly, divarasib compares favorably to the marketed KRAS G12C inhibitors, starting in laboratory experiments, particularly when you look at potency and selectivity. That's shown on the left side of the slide. These same preclinical characteristics were borne out clinically in the phase III KRASCENDO 1 study. That's shown schematically in the middle of the slide. There, divarasib showed superior efficacy compared to the approved G12C inhibitors, including an overall survival benefit that already was positive at its interim analysis. The full dataset will actually be presented during the Presidential Symposium at ESMO next month. The next divarasib study to read out will be KRASCENDO 2 in front-line metastatic non-small cell lung cancer. As a reminder, we presented highly encouraging phase II data from the KRASCENDO 170 study, which are shown on the right side of the slide. Those were presented in ASCO back in June, and they reinforce the potential for divarasib to perform in this setting. I'll conclude the solid tumor oncology portion of this talk with a few words about Tam-Peli, which is our next-generation B7-H3 antibody-drug conjugate that we recently enlicensed from MediLink for ex-China development. This ADC was developed on a MediLink platform, which basically leverages a dual-release payload technology. The linker is hydrophilic, and it's plasma-stable. The release of the payload is minimized in the circulation. This is important because this has been a reason postulated for many of the current generation ADCs having more toxicity than we would like. Improving that safety profile is critical, especially when we're thinking about combinations and moving to earlier lines. Tam-Peli also carries a potent TOPO1 payload. The drug-to-antibody ratio, the DAR, is eight in Tam-Peli. Finally, the target, B7-H3, is actually expressed across a spectrum of tumors. On the other hand, the expression of B7-H3 in healthy tissue is much more limited. This pattern of expression may offer opportunities to go after multiple indications outside of small cell lung cancer and lung cancer, etc. We'll soon be starting global phase III studies in small cell lung cancer, and we expect to expand to other indications soon thereafter. These are the phase III results from a Chinese population in the second-line small cell lung cancer context. These were presented at World Lung earlier this month, and you can see Tam-Peli conferred actually significant improvement in both PFS and OS. It's the OS results that we're showing on the left of this slide. In keeping with this dual-release linker payload design, the Tam-Peli safety profile is also encouraging. That's what we're looking at on the right side of this slide. First of all, Tam-Peli had a lower rate of grade 3 or greater adverse events compared to the topotecan control arm. Then when you look specifically at interstitial lung disease rates, which is a very important toxicity concern in this setting, we saw zero grade 4 or 5 events, less than 1% grade 3 events, and a 4.9% overall ILD event rate in the Tam-Peli arm. Broadly speaking, these ILD rates appear lower than what has been seen with other B7-H3 ADCs in small cell lung cancer, despite the fact that patients that were enrolled in this trial seem to be more sick in general than some of the other trials. Overall, Tam-Peli is showing excellent efficacy and at least competitive safety, and this certainly offers promise as we move into other pivotal indications. This wraps up the solid tumor section, and now it's my pleasure to turn it over to Matt Hellmann to take us through the hematology pipeline. Kind introduction here. As others have mentioned, I have just recently joined Roche- Genentech, and it's a real pleasure to be here. I hope you've heard so far why this moment is so important and exciting for our portfolio. These are the sorts of things that really attracted me to join Roche -Genentech in the first place. I think it has a really uncommon pairing. There's this long scientific legacy and an incredible long-term horizon for investment. Those two things together, which don't often come together, offer an opportunity to build on these learning loops. You heard the concept of recursion, but this is happening in our portfolio, where we're constantly creating breakthroughs and then trying to beat that over again. Those are the sorts of things that drive me, and I think why Roche Genentech is so well-suited to succeed in this moment. Hematology is one of those clearest examples across our portfolio. I want to spend a little bit of time telling you about why this is so exciting. It's not just a concept. Also, this is real execution and action. With the track record to a portfolio delivery, you can see where we have 30 years of bringing new molecules, new targets, new modalities into the clinic. That track record is something we're incredibly proud of and eager to build on. You can see on the right some really clear direction in terms of the key priorities within hematology, with very focused efforts in non-Hodgkin lymphoma, CLL, as well as multiple myeloma. I'll return to non-malignant hematology at the end as well. If the first slide gives you a sense of where we're headed and what our key disease focus is, I also want to take a step back and talk a little bit about the how. What are the elements that inform what success is going to look like in hematology? As we reflect on some of the macro trends in hematology, I think there's four big ones that we should focus on. The first is novel modalities. Hematology is a space where new modalities often begin, whether those are antibodies or ADCs, cell therapy. I'll tell you a little bit about bexobrutideg as the next one of a protein degrader, where new modalities begin in hematology. The second key theme, and this is a place that has been absolutely defined in hematology, is combinations. Multi-drug combinations beginning from the earliest stages of chemotherapy use in children with leukemia were absolutely transformative. Building regimens together to be able to achieve curative potential is absolutely the bedrock of hematology. That model has been exported to other diseases, so it's a place in which hematology really shines. Those two things together, new modalities with combinations, also bring us to the third thing, which is curative potential. Hematology is a place where words like cure, normalization, they're not just aspirations. They're real ambitions. It's what's in our reach. There's some examples throughout our portfolio of where we're really pushing the bar towards achieving something remarkable for patients. The fourth element is the patient experience. This is a disease where things like outpatient experience, the auto-injector, low toxicity, getting drugs to where they actually need to be for patients can absolutely be life-changing. With that, I will turn a little bit to our portfolio here. I hope you'll see, just like in the solid tumor portfolio, a real breadth of modality and targets, but also a longitudinal investment over time that represents our sustained commitment to hematology. I want to then deep dive a little bit into some of the places where we think we're most poised to change practice in the near term. Speaking of practice changing, let me double-click in non-Hodgkin lymphoma. This is a setting where that pairing I mentioned, that sort of legacy with ambition, really comes to life with CD20 targeting. Starting with DLBCL on the top, just a few years ago, the POLARIX study established a new front-line standard of care. It was the first time in 20 years that there had been a new standard of care. We're aiming to push that bar even higher with SKYGLO, which I'll talk a little bit more about in the next slide. We're also adding a new slide, EVERGLO, to this armamentarium to target on the elderly and unfit patients with a chemo-free study. In the middle, you'll see mantle cell lymphoma, where we have pivotal data planned for 2027. At the bottom is follicular lymphoma. We're moving Lunsumio-based regimens earlier. Levi had already mentioned the CELESTIMO study that read out just a few weeks ago, and we're excited to present later this year. I hope to give you an overall flavor that this is a true franchise strategy for NHL, where we're building on anchors, we're putting complementary pieces together, and aiming to move to the earlier lines to be able to improve outcomes for patients. We've talked a little bit about the STARGLO study already, SKYGLO study, excuse me. DLBCL is a place of key leadership for us. I wanted to spend this extra moment here. On the left, you can see that progressive, recursive increase in the standard of care. We've established new regimens by adding Polivy to the current standards and establishing a new front-line standard of care with Polivy, R- CHP, in 2023. Now with SKYGLO, we're aiming to raise that bar again by adding Columvi on top of that backbone. On the right, you can see why we're so excited to do that, why this phase I-B data gives us so much confidence. 100% response rate, 96% complete metabolic response rates, and meanwhile having a differentiated safety profile. These are the sorts of things that allow us to be able to achieve outpatient administration and be able to continue to build the next standard of care in the years to come. I think there's three key proof points for SKYGLO. We're building on the single best standard of care with excellent efficacy and a differentiated safety profile that's enabling us to give a fully outpatient treatment. We're excited as this is one of the key readouts for next year that you saw on Teresa's slide for the news flow to be excited about next year. I'm turning now to Lunsumio, which is an intentionally advanced differentiated CD20 bispecific. It has the safety and the administration profile that allows us to get TCE therapy to match the right patient in the right treatment settings. There's two key diseases where we think that that can really matter. On the right, you see second-line DLBCL, where our SUNMO study showed a substantial improvement in progression-free survival. The hazard ratio is 0.41. This is offering a chemo-free, fixed-duration, short-length regimen with also a very low rate of grade 2 CRS, which allows us to give the optimal profile for this disease setting. This is one where we filed and we're expecting a PDUFA date in February. On the other side, you'll see the CELESTIMO study, which I mentioned just read out two weeks ago, and we're excited to share that data later this year, hopefully. This is another place where the indolence of follicular lymphoma is so important to be able to match the right drug with the right patients. CELESTIMO is distinctly more convenient, offering a fixed duration, far fewer steroid premedication, and a simpler strategy for treatment with fewer injections and infusions. I hope across both studies, you'll feel this common thread: efficacy plus making TCE therapy practical for patients. Levi had addressed the bexobrutideg that we're so excited to have collaborated with Nurix earlier this year. I want to tell you a little bit why we're so excited about it. CLL obviously is a place where we're proud to have leadership here, but the disease has been really transformed by BTK inhibitors of the last decade. There's a real unmet need here, the clinical resistance problem. About 70% of patients develop mutations in the BTK protein that limits the efficacy of the current BTK inhibitors. Bexobrutideg is trying to do something fundamentally different. It's a protein degrader. The ability to not just be sort of a cork in a bottle, but actually fundamentally degrade the protein allows us to circumvent those resistance mechanisms, both in wild type and mutants, and to be able to provide differentiated efficacy. Just to illustrate that, I want to bring your attention to that heat map on the left there, because across the top, you'll see bexobrutideg with all green. That's showing high potency against all those forms of resistance, both wild type as well as resistance mutations and CLL. You'll see, as you look down the rows, you see much more varied efficacy. That's the covalent and non-covalent inhibitors in the marketplace, and you can see where those resistance mutations really matter. It's not just the resistance and the efficacy that differentiate us. I think even within the protein degrader class, we have differentiated selectivity, which you can see in the middle plot, which we hope will give us a differentiated safety profile. Similar to the other slides, this isn't just a concept. This is where real data and action give us confidence. What you can see on the right is data from our earlier phase studies that highlights just how potent this mechanism can be. We've seen 83% response rates, 22 months of progression-free survival, and all of this is across both wild type and mutant disease. You'll see the mutations along the sort of yellow bar on the bottom, as well as CNS disease, which is in the sort of the green-black asterisks at the top. These three key features we think can drive our enthusiasm is broad, deep, durable efficacy, differentiated selectivity that can give us differentiated safety, as well as CNS activity. These are the sorts of things that we think are going to define the next big leaps in CLL. To deliver that, we've already begun the phase III Daybreak 306 in second-line CLL, and we have additional combinations coming as we move towards first-line. I'm going to turn now to cevostamab, which I'll go through some data that we presented just last week at IMS. In multiple myeloma, this is a disease that has been really transformed by targets like BCMA. What we hear from clinicians is that in the field, what they really need is differentiated targets, new targets to be able to raise the bar further. That's exactly the story with cevostamab. It's a first-in-class bispecific targeting FcRH5. This target is different in that it's expressed across nearly all myeloma cells, but it's independent of GPRC5D as well as BCMA. That differentiated expression profile appears to yield a differentiated safety profile, such that in comparison with BCMAs, we are not seeing that same degree of severe infection rates. We are also seeing, importantly, the return of humoral immunity without continuous IVIG. In comparison with the GPRC5Ds, we're not seeing that characteristic on-target, off-tumor toxicity that affects the skin, hair, taste, etc. Again, this is not just another drug in this otherwise crowded market, but a really purposefully orthogonal target that's new and is ideally suited for combinations to come. You can see, excuse me, I went too far, you can see that data that we presented just last week at IMS with, I think, a lot of clinical enthusiasm. This CAMMA 1 data showed a response rate with cevostamab plus PomDex of 86% deep responses, durability with a median progression-free survival that was not met at 20 months of follow-up. Importantly, a difference in that safety profile, as I mentioned, 17% grade 3 infections compared to much more than that with some of the targets that are already out in the market and the return of humoral immunity. Consistent with the enthusiasm that we've heard from clinicians as we presented this data last week, we're excited to move this forward. The phase III CEVOLUTION study is already ongoing, and we have other combinations to come as we move into earlier lines of therapy. I want to end here and talk about hemophilia. This is a place where it's one of the clearest examples of that Roche innovation, where we've already changed patients' lives, and we're looking to do it yet again to push the bar higher. On the left, you'll see a little bit of the timeline of where innovation has come in hemophilia that has been nothing short of life-changing. For decades, treatment with patients with hemophilia meant recombinant factor VIII, and that required intravenous infusions multiple times a week. You can imagine how burdensome that is, especially for children who are suffering with hemophilia. Especially with that, the short half-life of the recombinant factor, it left them naturally vulnerable to bleeds in between those trough cycles. Hemlibra came along and fundamentally changed that paradigm, replacing frequent IV therapy with a subQ. Many of you saw the auto-injector to come. This has enabled a very different life for patients, 80% of them experiencing zero treated bleeds. As Teresa mentioned, this impact was awarded the Lasker Award just this year, which we're really proud of, and the auto-injectors coming. Like so many of the stories you've heard so far, we're not stopping there. Our ambition is to move even more patients into that bleed-free life and achieve something that qualitatively is different, something that hasn't been achieved before, which is hemostatic normalization. I'll go through that a little bit in the next slide. That's our goal with NXT007. As you can see there, we've already moved this into phase IIIs, three phase IIIs actually, in fact. This is another one of the key news flows that we'll hope for in 2027. To tell you a little bit about why we're so excited with this molecule, zemocimig, it is that we recognize that Hemlibra's efficacy has been transformative. What it does often is to move patients with severe hemophilia into a milder phenotype. Zemocimig is designed to close the gap towards normalization. On the left there, you can see the molecule, and some very specific antibody engineering went into engineering a molecule that's 30x more potent than Hemlibra, while also being able to deliver a molecule that can be given as an injection subQ infrequently, about once a month. This is, again, towards achieving that bar of normalization. Much like many of the other molecules, the data backs us up. I want to spend just a moment explaining this middle plot because it's precisely what we're after in terms of normalization. What you'll see in the middle are two horizontal lines that are dotted. That represents the lower and upper limit of normal for patients that do not have hemophilia, the thrombin generation that they experience as a "normal" person. What you can see in the purple and green is the pharmacodynamic effects of zemocimig in patients with hemophilia smack in the middle of the normal range. These are patients with and without hemophilia sitting right next to each other in pharmacodynamic terms. That's translating clinically. As you can see in the bar chart all the way to the right, we're pushing the boundary even higher for bleed-free life now into the 90s. Hemostatic normalization gives us the opportunity to achieve something very different than what's been experienced thus far, while simultaneously maintaining an excellent safety profile where we've seen no thrombotic events, no TMA thus far. As I mentioned, we have three phase III studies that are already underway. ZEBRHA 2 in particular, I want to point out, is the first and only phase III study that's going head-to-head against Hemlibra and giving us that opportunity to demonstrate how we're continuing to push the bar higher. We'll look forward to data later next year. As I close, I hope what I've given you a sense for across these stories in hematology is that spirit and track record of continuing to push the bar higher. These are precisely the things that drew me to Roche in the first place. I'm so excited to tell you more as these years go of the progress towards that. Of course, it's not just in hematology-oncology. That spirit is characteristic to Roche. I know you'll hear more of that across the TAs to come. With that, I'm going to pass the baton to Hideki, who's leading neurology, to tell you about exciting things there next. Thank you. Thanks, Matt. Okay. Thanks, Matt. I'm really excited to present to you our progress, our rapid progress in neurology at Roche. At Roche, we are the industry leaders in neurology, both in our on-market portfolio and our R&D pipeline. How do we get there, and how are we going to sustain that number one position? We're going to focus on the critical capabilities, as shown on the left-hand side of this slide. We have some unique therapeutic modalities, including Brainshuttle, which we're using in both Alzheimer's disease as well as in multiple sclerosis. We, in neurology, have a unique and productive collaboration with our diagnostic colleagues. There, the diagnostic colleagues have developed a blood-based biomarker for Alzheimer's disease, which allows rapid identification of Alzheimer's patients, as well as now, for the first time, we're talking about preventing neurologic disease before it even happens in Alzheimer's disease. We're able to identify patients with this blood-based biomarker who are at risk of developing Alzheimer's disease, but otherwise completely normal. We've begun a phase III study in this population of patients called PrevenTRON. Finally, we continue to focus on our end-to-end disease areas, Alzheimer's disease and multiple sclerosis, as well as in select breakthrough indications like Parkinson's disease, where today there are no disease-modifying treatments available. With these capabilities, we have built this pipeline, which has advanced from what you saw last year. I will highlight a few of these. First of all, as you heard this morning, we have filed for registration fenebrutinib for PPMS and relapse EMS. We have a number of phase III readouts toward the end of this decade. Next year, in 2027, we expect a readout for Enspryng and autoimmune encephalitis. The phase II part of the pipeline has really advanced with multiple molecules for multiple sclerosis, Alzheimer's disease, and Parkinson's disease. Compared to last year, the pipeline has become much more robust and diverse. Let's dive a little bit deeper into fenebrutinib, which is our unique oral, non-covalent, reversible, and very deeply penetrating BTK inhibitor, now with impressive phase III results. As announced earlier this year and as you heard this morning, fenebrutinib is the only BTK inhibitor with positive results across relapse EMS and primary progressive MS. Fenebrutinib is designed to address both the relapsing and progressive part of the biology of multiple sclerosis by targeting both B cells and macrophage and microglia in the brain. In fact, if you look at the graph on the right, we've shown that fenebrutinib can decrease the activation of microglia by the concentration of IC90 greater than 90%. In other words, we have near total inhibition of B cells and microglia in the CNS. Now, I want to show you the relapse EMS data that we've been talking about so much this morning, with our phase III trials, FENhance I, FENhance II, and relapse EMS. It's important to remember that in this study, the active comparator was teriflunomide, which by itself has effect on relapses as well as disability progression. As you heard, we have a 51% and 58.5% reduction in relapses in this study. In fact, the relapse rate that's shown here is the lowest ever reported for any phase III study, and it translates to about one relapse every 17 years for the patient. Really profound activity. In the middle panel, you see a profound reduction in new lesions within the brain of 76% and 82%. Finally, in the right panel, we see a reduction in disability progression in the composite measure CCDP. In fact, when we pool the data from the two studies, we see nominal significance in reduction in disability progression. Again, this is over and above the active molecule, the active comparator, which is teriflunomide, which already shows a reduction in disability progression. There's a recent publication in neurology that classified all the available treatments for multiple sclerosis into high efficacy, moderate efficacy, and low efficacy based on relapse rate reduction. Based on this classification, fenebrutinib would fit into the high efficacy category. In fact, this would be the first and only molecule, oral molecule, to go into that category. Turning to the PPMS study, fenebrutinib is the only BTK inhibitor with positive results in PPMS. Again, we use an active comparator, Ocrevus, which by itself is a high-efficacy treatment for MS and by itself showed a 24% reduction in disability progression in our phase III study. Over and above that, we have fenebrutinib, which first of all met its primary endpoint of non-inferiority to Ocrevus, and in fact, showed a 12% reduction in disability progression in this study. With these results of effect on disability progression in PPMS, as well as what I just showed you with RMS, we now have a molecule that has improvements in disease progression throughout the spectrum of multiple sclerosis. Therefore, we conclude that fenebrutinib effectively targets the underlying common disease biology, disability progression, throughout the entire MS spectrum. The end of our MS strategy does not stop with fenebrutinib. We have a leading early-stage pipeline with multiple next-generation approaches, including on the left, a Brainshuttle anti-CD20, which is moving toward phase II. On the right, we have an oral MAGL inhibitor, which is designed to enhance neuroprotection as well as to reduce neuroinflammation. This now is in the phase II study in combination with Ocrevus. As you heard this morning, with Enspryng, we had fantastic results on a disease called MOGAD, and I have now the pleasure to show you the actual results. MOGAD is a relapsing autoimmune disease where patients experience relapsing visual loss, cognitive decline, as well as loss of ambulation. Today, there are no treatments available for this disease. On the right panel is our phase III results in a study called METEOROID, where we see reduction in time to relapse versus placebo of 68%. This is highly significant. Not only in the primary endpoint, but in multiple secondary endpoints, including analyzed relapse rates, MRI lesions, and rescue therapy use, we hit significance in all these secondary endpoints. This was filed in FDA as well as EMA. FDA, we achieved priority review with a PDUFA date expected early next year. Now, let's move to our Alzheimer's portfolio. Despite two treatments, as you know, approved for Alzheimer's disease, two anti-inhibitors approved for Alzheimer's disease, there's a tremendous unmet medical need. We need better diagnosis of Alzheimer's. We need greater clinical benefit and efficacy. We need improved safety, as well as we need improved dosing and administration. With trontinemab, our Brainshuttle-coupled anti-inhibitor, we believe we can achieve all these. I'll show you some data that suggests that. As you heard this morning, our phase III studies in trontinemab, called TRONTIER 1 and TRONTIER 2 in early Alzheimer's disease, has fully enrolled, and now we expect data in the first half of 2028. Just to be clear, answering the question from this morning, the study was designed such that there's a fixed 18-month dosing period for the last patient in, and that's why that 2028 date cannot come any earlier. As I mentioned, we have started a PrevenTRON study, a prevention of Alzheimer's disease study, in phase III as well. Trontinemab is connected to the Brainshuttle, which allows not only deep blood-brain barrier penetration, but also broad penetration throughout the brain. That's shown in this mouse graph on the far right, comparing a non-Brainshuttle version to a Brainshuttle-coupled version, which you see the green fluorescent highlighting throughout the brain. In our phase II study, we've already shown best-in-disease plaque removal. Now, on the left-hand side, you see that 92% were below the amyloid positivity level at 28 weeks. Not only that, all patients, 100% of patients, had amyloid reduction. In other words, there were no non-responders to trontinemab in this phase II study. On the right panel, open-label extension patients continue to show persistent reduction in amyloid for at least one year. Further, we see profound effect on fluid biomarkers, including pTau217 and GFAP. In exploratory efficacy outcome, clinical efficacy outcome, despite the small sample size and the short duration of this phase II study, up to six months, we see encouraging clinical efficacy signals, with less decline in two measures of cognition, CDR Sum of Boxes, and mini-mental status examination. Further, in this phase II study, there are indications that we have potentially a best-in-class safety profile, with really low ARIA-E. Why might this be? We think it's related to the Brain shuttle. On the middle panel, which is a mouse brain with a Brainshuttle-coupled monoclonal antibody that targets the amyloid plaques, as you see in yellow, whereas if you don't have a Brainshuttle-coupled antibody, it targets the amyloid within the blood vessels. We believe that because of the Brainshuttle, it allows the molecule to bypass the vasculature and therefore reduce the amount of blood-brain barrier leakage, and therefore reduce the amount of brain swelling, or ARIA-E. In fact, that's what you see in the right table, which is from the phase II study. During the double-blind period, if you look at the far right column, 3.6 mg/ kg dose, which is the dose we're taking into phase III, there's only one case of ARIA-E, and it was an asymptomatic case within the phase II study. In summary, trontinemab has potential for best-in-class amyloid reduction, efficacy on cognition markers, as well as safety. Switching to the early pipeline in Alzheimer's disease, there's an exciting molecule called GSM, gamma-secretase modulator. This is the highly selective oral modulator, gamma-secretase, which has first-in-class potential in Alzheimer's disease. It alters APP, or amyloid precursor protein, and inhibits Aβ aggregation. In a phase I dose escalation study on the right, we're seeing encouraging biomarker results, with dose-dependent reduction of the pathogenic Aβ42 species for up to 70%, and increase in the non-pathogenic, shorter Aβ 37. We've started a phase II-A study in early Alzheimer's disease, with data expected next year in 2027. Another exciting molecule in the early-stage pipeline is an NLRP3 inhibitor. NLRP3 forms part of the inflammasome complex. In Alzheimer's disease, this complex is the link between Aβ aggregation and t au pathology. On the right-hand side, with another NLRP3 inhibitor, we show in a phase I study in Parkinson's disease a reduction in an important biomarker, CSF-IL18, which is associated with inflammation as well as downstream neurodegeneration. You see a reduction in this phase I study. Therefore, we've started phase II studies in both Alzheimer's disease and Parkinson's disease with this molecule. I'll end with one slide update on prasinezumab, which is our anti-α-synuclein monoclonal antibody for Parkinson's disease, where again, there are no disease-modifying treatments available today. As you'll recall, in the right-hand panel, we went to phase III based on the phase II-B PADOVA results, where in the L-DOPA subgroup, we saw a 16-week reduction in the progression of motor symptoms, as you can see in the right-hand graph. I'm pleased to report that the phase III study called PARAISO has been rapidly enrolling, and so we expect data toward the end of this decade. With that, let me turn over to Larry Tsai to talk about our exciting immunology portfolio. Thanks so much, Hideki. Hi, everybody. How's everybody's jet lag doing? It's just about time for me to be waking up in San Francisco, so actually, this is pretty good timing for me. I'm Larry Tsai. I'm the Global Head of Product Development for Immunology, and I'm super happy to be here to give you a quick update on what's going on in Roche immunology from the standpoint of our strategy and our pipeline. Hopefully you're aware that Roche has a long legacy of innovation in immunology, including the first anti-CD20, the first anti-IgE, and the first anti-IL6, all of which continue to benefit patients to this very day. We're very, very proud of this legacy, and so our immunology strategy aims to build on this by focusing on four strategic components. First of all, optimizing validated and novel pathways that are relevant to immune-mediated diseases. Second, developing rational combinations, including bispecifics, to lift the efficacy ceiling for complex immune-mediated diseases. Third, using biomarkers to break down heterogeneity and identify endotypes that are more likely to benefit from treatment. Fourth, for selected autoimmune diseases, seeking functional cures through deeper or more targeted B-cell depletion to reset the immune system. Our strategy centers around two end-to-end diseases, chronic obstructive pulmonary disease and inflammatory bowel disease, while also recognizing opportunities for breakthrough innovation in other immunologically mediated diseases. Because, as I like to say, the immune system doesn't recognize anatomic boundaries, and so sometimes you really just need to follow the science to wherever the most potential patient impact is. As you can see, our immunology clinical stage pipeline reflects our focus both on end-to-end and breakthrough innovation diseases. I'm not going to have time today to cover all of the programs that are on this slide, obviously, but before I take a deeper dive into just a few of them, I want to briefly call your attention to our COPD pipeline, which targets both inflammation and mucoobstruction, which are two distinct pathobiological drivers of the disease. We are particularly excited about some of our next-generation bispecific molecules, including our IL33-TSLP program in phase I, which combines two foundational, validated pathways in order to raise the efficacy ceiling across the entire range of COPD subtypes. Now, one breakthrough innovation area that we are highlighting again this year at Pharma Day is immune-mediated kidney diseases, which are the third most common cause of end-stage kidney disease. Chronic kidney disease, or CKD, including end-stage kidney disease, affects about 1 in 10 people worldwide. It's associated with high morbidity and mortality, and it's even predicted to become the fifth leading cause of death globally by 2040. About 25% of lupus patients, about 50% of IgA nephropathy patients, and about 30% of membranous nephropathy patients, all immune-mediated kidney diseases, will develop end-stage kidney disease. Our pipeline in immune-mediated kidney diseases continues to expand now with multiple programs in lupus nephritis, membranous nephropathy, idiopathic nephrotic syndrome, hemolytic uremic syndrome, and IgA nephropathy. It really has been a banner year for Gazyva, which received FDA approval for lupus nephritis shortly after Pharma Day last year, and since then has notched three additional registrational study wins in extrarenal lupus, membranous nephropathy, and idiopathic nephrotic syndrome. Gazyva is, of course, our higher-potency next-generation type 2 anti-CD20 that's engineered to be less dependent on complement-mediated cytotoxicity. That makes it better suited to diseases which are characterized by depletion of complement, like severe lupus. Building on our experience with earlier-generation anti-CD20s, Gazyva tested the hypothesis that deeper B-cell depletion, particularly at the tissue level, could raise the efficacy ceiling in diseases like lupus and immune-mediated kidney diseases. Of course, the whole emerging field of immune reset is actually based on the hypothesis that even deeper B-cell depletion can restore the immune system to a naive state, resulting in a functional cure. Our own pipeline, including Lunsumio and cevostamab, continues to test this hypothesis. Sorry. Turning back to SLE, sorry, this is not what I was expecting here on these slides. Okay. Well, turning back to SLE. SLE affects 3.4 million people worldwide. Many of them are women of childbearing age. Many of them are actually women of color. Lupus can actually affect almost every organ system in the body, including the kidneys. In its most severe form, lupus nephritis actually affects about 50% of lupus patients. About a third of them will actually go on to develop end-stage kidney disease. Lupus nephritis itself is also associated with high mortality, about six times above the normal patient population. In the ALLEGORY phase III study, Gazyva met the primary endpoint with showing a 4 percentage point improvement in the SRI4 index, I'm sorry, showing a statistically significant improvement in the SRI4 at 52 weeks, with a clinically meaningful treatment difference of 23.1%. Additionally, Gazyva demonstrated statistically significant and robust efficacy over placebo across all secondary endpoints, including a 20% treatment difference on the DORIS remission score. You can see that 33.8% of patients actually had complete improvement of their symptoms with Gazyva, compared to only 13.8% with placebo. This was submitted then to the FDA and to the EMA, and we are expecting approval before the end of the year. Okay, but the Gazyva hit-parade continues with membranous nephropathy. Membranous nephropathy is the most common cause of nephrotic syndrome in adults without diabetes, and up to 30% of patients with membranous nephropathy will progress to kidney failure. In the MAJESTY phase III study, Gazyva met the primary endpoint, demonstrating a 31% increase in complete remission compared with standard of care tacrolimus at week 104. Priority review was granted for Gazyva membranous nephropathy, with approval expected also before the end of the year. Hot off the presses, FDA approval was granted last week for Gazyva in idiopathic nephrotic syndrome, making it the first approved targeted therapy and the first drug of any kind approved for INS since prednisone was approved 70 years ago in 1956. It's been a long time. The approval was based on the INShore phase III study, which showed a 23% increase in sustained complete remission at week 52 for Gazyva versus standard of care mycophenolate. INS, if you don't know much about it, is the most common primary kidney disease in children. 70% of patients cannot taper off of steroids, which are the current standard of care, and of course are associated with major side effects. Gazyva can potentially reduce the need for steroids in these patients, which is highly impactful, particularly for developing children, and makes us super excited about this approval. With these four indications, Gazyva will be well positioned to become the treatment of choice in immune-mediated kidney disease. The good news goes beyond Gazyva. As we announced last week, sefaxersen is the first antisense oligonucleotide with a positive phase III study reported now in IgA nephropathy. Sefaxersen targets factor B production in the liver for selective complement suppression, and because high factor B levels are associated with IgA nephropathy and are part of the pathobiology, sefaxersen treats IgA nephropathy by inhibiting factor B mRNA translation. In the phase III IMAGINATION study, sefaxersen met the primary endpoint, demonstrating a statistically significant and clinically meaningful improvement in urine protein/creatinine ratio compared with placebo at week 37 in a pre-specified interim analysis. UPCR is, of course, an indicator of early response, that's an accepted surrogate for long-term kidney function, and the data will be discussed with global health authorities and presented at an upcoming medical conference. A little bit more about IgA nephropathy. IgA nephropathy is a major cause of chronic kidney disease worldwide. It's actually the most common primary glomerular nephropathy, and about 50%- 60% of patients with the disease remain uncontrolled even on advanced therapies. That's about half of those that then will progress to kidney failure. As a result, there's increasing emphasis on early diagnosis and more aggressive treatment. The current treatment algorithm stratifies patients based upon UPCR, with higher risk patients receiving advanced therapies. Sefaxersen has best-in-class potential amongst the complement inhibitors, which are targeted at patients with more active inflammation at baseline or who are inadequately controlled on other therapies, which again represents 50%-6 0% of patients. Sefaxersen actually also uniquely offers convenient subcu at-home monthly dosing, which may improve patient adherence. We'll switch gears now for an update on afimkibart. Afimkibart is our anti-TL1A antibody being tested in inflammatory bowel disease as well as MASH and rheumatoid arthritis. TL1A, of course, sits upstream of and amplifies a range of inflammatory and tissue remodeling pathways, and thus we believe in the potential of afimkibart to be a pan-immunology pipeline and a molecule. I'll give you an update on our IBD studies in just a moment, but first a few updates on other indications. First, atopic dermatitis, where unfortunately the VELORITE phase II study did not meet the primary endpoint, and thus development activities in atopic dermatitis have been discontinued for afimkibart. Afimkibart was well tolerated, with no new safety signals, and we, of course, know that TL1A biology is disease-specific with different underlying pathobiological drivers. Therefore, we do not believe that the atopic dermatitis results can be extrapolated to other diseases, and we remain confident in the potential of afimkibart to address a wide range of immune-mediated and fibrotic diseases. Turning to rheumatoid arthritis, enrollment in the dRAvite phase II study is on track. A competitor phase II study recently demonstrated activity in rheumatoid arthritis, but insufficient efficacy to support continued development. Now, that was obviously a different molecule, so the read-through to afimkibart is unclear, but it's also worth noting that dRAvite, the afimkibart phase II study, is studying a more refractory patient population where there may be more potential benefit. Top-line data are expected in 2027. Quickly turning to MASH, the FIBROLITE phase I study completed enrollment in 2025 and is expected to read out later this year. A few words about IBD. IBD remains an end-to-end disease within our immunology strategy. 8 million people worldwide suffer from IBD, and although there have been significant advances in treatment over the past decade, the efficacy ceiling remains low. Only about 40% of patients respond to therapy upfront, only about 20% achieve longer-term remission, and even of those 20%, about 40% ultimately lose their response. Our AMETRINE- 1 and AMETRINE- 2 phase III studies in ulcerative colitis are now fully enrolled, with top-line results expected in 2027. In addition, we are preparing to initiate a third phase III study with an all-subcutaneous induction and maintenance regimen. We believe that IBD patients and physicians want more choices when it comes to route of administration and care setting for delivery, and so offering both IV and subcu induction regimens offers maximum flexibility to personalize the approach and to minimize the overall treatment burden. Turning to Crohn's disease for a moment, enrollment in the SIBERITE phase III studies is also on track. Because IBD is an end-to-end disease, we're continuing to progress our early-stage pipeline of bispecifics built on an anti-TL1A backbone, combined with other validated targets, including IL-23. Our SUNCREST phase II-B study in ulcerative colitis for our p40 x TL1A bispecific that was mentioned earlier this morning continues to enroll, with data expected in 2027. This molecule combines two validated foundational and orthogonal pathways, and thus we are excited by the potential of this molecule to break through that efficacy ceiling that I talked about. This is only one example of several bi- and multispecific antibodies and oral therapies that we have in development as part of our longer-term IBD strategy. That was a bit of a whirlwind, but hopefully I've given you a sense of all the things there are to be excited about in 2026 and 2027 and beyond in Roche Immunology. With that, I'm going to hand the stage over to Chris to tell us what's going on in ophthalmology. Okay. Good afternoon, everybody. Ophthalmology is generally considered within the five TAs, probably one of the smallest. I understand that. However, as you saw from Teresa this morning, we have the highest growth rate over the next 10 years, well over 30%. There's really a big opportunity. As we've seen in the last few years, you know, we tend to punch a little bit above our weight with the [VISE] mode, not only in impact on healthcare systems, but really in improving vision for individual patients. I'm really excited about the next 10 minutes just to talk you through a little bit about our pipeline and the progress that we're making towards our vision of preventing vision loss and restoring vision in those for whom prevention is too late. How are we going to achieve this? Again, fairly similar slide that you've seen with our other TAs. I'm going to talk a little bit in more detail about these first two approaches in the next couple of slides. Just briefly, we have approaches to address novel mechanisms of action and new indications addressing multiple disease pathways using multispecifics. We have an extended durability approach, so we strongly believe that there is going to continue to be an important opportunity to treat patients with extended durability therapies. There remains an unmet need there because of the amount of injections in clinics that are required for the current therapies. Susvimo is a good example, and I'll talk a little bit about the upcoming VELODROME study, which looks at nine-monthly dosing intervals. Then, as Aviv mentioned this morning, the importance of digital large volumes of data and our end-to-end collaborations there across the company. Specifically, super exciting. You know, we've talked over the years about the millions of images that we capture in-house, the tens of thousands of patients' worth of clinical data, genetics data. We talk about the thousands of biological samples that we can take. We've already been able to reverse translate some of those to identify new targets in our retinal diseases. As Aviv also highlighted, we've used those to develop algorithms, and specifically the geographic atrophy algorithm, which has enabled us to identify up to, or to predict up to 50% of the future progression over the next year of patients who develop geographic atrophy. We came into a little bit of a paradox as Aviv was talking through that, describing, if we'd had this algorithm, we wouldn't have done the phase III study, but then if we hadn't done the phase III study, we wouldn't have got the algorithm. I was a little bit confused. I think it's important that we have the algorithm because it really enables us to identify these patients who are going to progress, which enables us to actually either halve the sample size of our trials because of the additional power, or we can just accept the additional power and have a higher PTS for our studies. Going into how we take our approach to looking at novel mechanisms of action within retinal diseases, we actually take a very thoughtful, stepwise, methodical approach looking at these, what we've described as hallmarks of diseases. Angiogenesis, inflammation, ischemia, and fibrosis. Starting with angiogenesis, over the last few years, obviously, we've taken VEGF, we've added the Ang-2, Vabysmo has been extremely successful. You heard from Teresa, you know, the number one prescribed therapy in many countries around the world for both diabetic macular edema and neovascular AMD. We're continuing on the angiogenesis pathway. We now have a WNT asset in clinic, which is really exciting. The second hallmark of disease which we're looking at is inflammation, and I'll be talking a little bit more about this shortly. We've taken, again, a very stepwise approach. We have vamikibart in clinic, has been filed in multiple countries for uveitic macular edema. You've heard Teresa talking already this morning, and I'll mention it again shortly, that the VEGF x IL6 bispecific is entering phase III. We also have our trispecific antibody, which is in clinic. Further down the pipeline, we're addressing ischemia, which many of you will recall we have a partnership through SemaThera with a semaphorin target. Even earlier in the pipeline, we continue to look at fibrosis. Now, looking at durability, the port delivery platform, again, I was secretly hoping that Teresa, for those of you watching this morning, she pulled two devices out of her pocket. I was hoping she was going to pull a third, which was the port delivery system, but it's the size of a grain of rice, so it wouldn't have made much of a visual impact. It certainly makes a great vision impact. The reason I say that is because we actually now have seven years of clinical data from this particular molecule, from this particular approach, which is called Susvimo. Also, we have, you know, nearly over 95% of patients can go on a fixed six-monthly treatment interval with Susvimo. It's really a great asset, and the port delivery platform can be used in many ways. We've recently gotten European approval also for Susvimo. In addition, we have Australian approval for Susvimo. Now, this is a unique platform. It's the first and only approved surgically implanted long-acting treatment for retinal vascular disorders. Importantly, it can be used for multiple additional assets. Specifically, we'll talk a little bit about the VEGF x IL6, which can also go in the port in the future. Early next year, we'll be releasing data from VELODROME, which is looking at fixed nine-monthly treatment intervals. Really continuing to shift the treatment durability paradigm. This is where we are with our ophthalmology pipeline. Three things I wanted to kind of share with you. Number one, we continue to replenish our phase I assets. Number two, you'll see here in phase II, and I'll talk about this on a subsequent slide, we've now entered phase II with a complement factor B, small interfering RNA in geographic atrophy. Number three, just to reiterate, we've entered phase III with our VEGF x IL6 bispecific in diabetic macular edema. Looking at IL6 specifically, again, this is one of our key hallmarks of disease targeting inflammation. As a reminder, IL6 targets not just inflammation, but it targets leukocyte adhesion and chemotaxis, blood-retinal breakdown, blood-retinal barrier breakdown, angiogenesis, and it upregulates VEGF, which means it has a role in multiple diseases, including uveitis, retinal diseases, and thyroid eye disease. As you can see on the right side here, we've filed and are eagerly awaiting within the next two to three weeks the result of the Enspryng filing for thyroid eye disease in the U.S., with a PDUFA date on 15th of October. As you recall, that received priority review. We're anticipating that. Vamikibart has been filed for uveitic macular edema in the U.S., EU, and EU. We've mentioned the VEGF x IL6 bispecific, but also we now have the trispecific antibody in clinic. Looking at vamikibart, just as a reminder, really, this is one of the most frequent causes of vision. Uveitic macular edema is one of the most frequent causes of vision loss in patients with uveitis, which is affecting patients, majority of whom are of working age. Despite immunomodulatory therapies, UME persists in about 40% of these patients. The current standards of care are steroids, both systemic and intraocular, which have severe side effects, including cataract, raised intraocular pressure, and if that continues, then patients develop glaucoma and sometimes require glaucoma filtration surgery. On the right-hand side, you can see the pooled data from SANDCAT and MEERKAT, which are two phase III studies from last year, which demonstrate clear, rapid improvements in vision of about nearly 11 letters. Importantly, the therapy was well tolerated with a low incidence of these important adverse events, including low rates of intraocular inflammation, low rates of IOP, and low rates of cataract. We'll be presenting 52-week data in the next two to three weeks at American Academy of Ophthalmology. The VEGF x IL6 bispecific, this, as a reminder, is, as I said, just entering phase III now, is compatible with a port delivery platform. The decision was based on a number of elements here. Number one, on the right-hand side, you'll see the BARDENAS study, which showed that vamikibart, so IL6 plus ranibizumab, was superior to ranibizumab alone. A 3.6-letter difference. If you look at the bottom, in terms of one of the regulatory endpoints, which is 15-letter gainers, you'll see that 44.7% of patients had 15-letter gainers on the combination versus 28% when given anti-VEGF alone. We took the visual acuity superiority in addition to the phase I-B, where we looked at some safety data, and that gave us the confidence to move the VEGF x IL6 into phase III. Now, I just want to add an additional point here, which is how we're looking at our phase III. You'll have heard, for those of you who follow ophthalmology, which I hope is almost all of you, we see many companies talking about how exciting their products are, how much better they're going to be than Vabysmo, and then they go and compare their products with monthly ranibizumab anti-VEGF monotherapy or eight-weekly aflibercept, so eight-weekly anti-VEGF monotherapy. Which suggests to me that they don't have the confidence that their products are actually better than Vabysmo. We will be comparing our VEGF x IL6 bispecific with Vabysmo because we believe it's the best comparator and we have the confidence. We're very much looking forward to these studies starting, and I think this gives patients the best option because we're really into transformational therapies. Moving on to Enspryng in thyroid eye disease. Hideki and others have spoken a little bit about the mechanism of action. This is just to highlight the disease as a reminder. It's a complex inflammatory autoimmune disease that affects 80% of patients, actually, of females with a median age of 43, so working age. Current treatments, which are approved globally, are therefore just to highlight kind of the adverse events which are particularly impactful. Therefore, hyperglycemia, menstrual disorders, hearing loss, and teratogenicity. Therefore, our data, which suggests in SatraGO-1 and SatraGO-2, which had the primary endpoint of improvement in proptosis, which is how far the eyeball sticks out, was a relation to activity of disease. There are about 50%, just a little bit over 50% of patients improved their proptosis rate in the active arm. When you look to combined active and inactive patients, this was also consistent, and we see consistency of treatment effect in many other secondary endpoints, including the clinical activity score. Then on the right-hand side, if you look at the adverse events, really important to note, we saw no events of hearing loss and no events of hyperglycemia, and just a low rate of neutropenia. Really well differentiated product in terms of safety. Then my final slide, I mentioned that we're moving into a phase II in patients with geographic atrophy. We did actually have a complement factor B inhibitor about a couple of years ago in the GOLDEN study, where we saw with a 70% complement factor B knockdown and a 35% PD effect of alternative pathway activity inhibition, we saw a small but non-statistically significant effect on GA lesion growth. We therefore decided to acquire this asset, which is a small interfering RNA. As you can see on the right-hand side, it really knocks down CFB and the alternative pathway by about 100%, at least out to three months. The unmet need in geographic atrophy, just as a reminder, it really remains very robust. The only two approved therapies are approved only in a very, very small number of countries. The treatment burden is great, so it's injected every one to two months. The visual function benefits have not been demonstrated by those approved products. We really see this as a great opportunity to make a difference, and this is entering phase II. With that, I'm delighted to hand over to my colleague, Manu Chakravarthy, to talk us through our CVRM portfolio. Thanks, Manu. It's great. Great. Good afternoon. I just wanted to thank my colleagues for a wonderful presentation of all the other therapeutic areas. I'm happy to close this out, this session with CVRM. I think for this audience, I think everybody understands the burden and the challenge and the opportunities that we're facing, right? The reason why we are in CVRM is because this is going to be probably one of the largest burdens of humanity over the next 25 years or so. Two out of three people in the world will experience at least one what is called a CVRM disease. CVRM disease is indicated on the left there that you can see ranges from MASH to diabetes to obesity to heart failure, stroke, even Alzheimer's to some extent. It really is a multi-organ, interconnected web of diseases that we're really trying to understand how to actually grapple. No doubt that current modern therapies have made a significant impact, right? If you actually look at the state-of-the-art incretin therapies, we see very promising results in terms of end-organ benefits, like improvements in heart failure, improvements in CVD, etc. But still, even with all of that, one out of five people that are treated with the state-of-the-art incretin will experience a heart attack. 40% of people will still have residual inflammation markers, and 6%-8% of residual cardiovascular or cardiorenal events still persist even after therapy. What all this means is that our current evolution, at least the way that we approach it at Roche, is that we're moving from a purely adipocentric view to an adiposity-independent view of the interconnected web of CVRM. This is the way that we've been trying to approach how we think about designing our pipeline. This is the way that we've been thinking about how we want to actually think about the modalities that we want to bring forward. In terms of our capabilities that we want to actually bring to bear to address this incredible complexity, the heterogeneity, the market segmentation that will no doubt be there, is to really position ourselves by building out a pipeline that is really focused on the next generation of molecules, but the broad swath of mechanisms, because we know that this is heterogeneous, we know this is interconnected. Also, as you heard from Aviv today, we actually have to go deep in research to really build out the foundational pathobiology of how these diseases actually work. We are making investments there to understand both the adipose-independent and adipose-dependent mechanisms using a diverse set of modalities, and that's going to be really key to address this heterogeneity. Like you've heard from my other colleagues, combinations are going to be still relevant here too. It would be really facile to think that you could address one very complicated disease with one target. Bringing in combinations, both within our own portfolio as well as within our wonderfully rich cross-therapeutic area portfolio, becomes very important. I would submit that Roche is very well equipped and positioned for doing that. Then finally, I think you've heard several times mentioned today this very unique partnership with Roche Diagnostics. Dia and Pharma reside under the same roof. Just like neuroscience, CVRM is another poster child where diagnostics can really make a big impact. I can give you one example with MASH, for example, but there are many, including in diabetes, cardiovascular disease, and other areas where Pharma-Dia interface is going to make a big difference. When we bring all of these capabilities to bear with the complexity of the biology that I just very briefly outlined, what we have then is a pipeline that we're really excited to showcase the progress over the last 12 months or so. When I was here last year, I showed you all of the light blue lines where we were just entering into phase II. We're starting to read some data in phase II for enicepatide. I'll show you some data on how those results actually turned out. Happy to report that based off of those results, now we're able to make advances of that program into phase III. As you heard from Teresa this morning, we've also announced or made the decision, I should say, to advance our oral GLP-1 agonist into phase III. We continue to make good progress on petrelintide and that has been advanced to phase III. Of course, zilebesiran that you've heard a little bit about, and I'll show you some more emerging data, very exciting, continues to build the case for why this is so important, despite the fact that there are other hypertensive agents, but there isn't one that gives you this level of control. We're in the midst of our ZYNERGY trial now and it's going really well. Of course, pegozafermin. Where we continue to make good progress. Then finally, we're also able to bring in two additional assets into the portfolio, NLRP3 inhibitor for ASCVD, and then a very exciting new target called urocortin-2 that we're doing in partnership with Hanmi for obesity and type 2 diabetes, with a particular focus on body composition. Let's kind of dive into a couple of key areas just to give you the actual data and the feel for why we're so excited about this portfolio. We'll kick start with enicepatide. Again, you'll recall that one of the unique aspects of this molecule that, to the best of our knowledge, is not being replicated by other molecules is a dual-bias signaling property where both receptors, GLP-1 and GIP, are fully biased. When we talk about bias, we're referring to, of course, the signal transduction mechanism. On the left of this graph, on the left there, is an example of an unbiased ligand. One example of such a ligand is semaglutide, for example. It activates both the cAMP pathway and the β-arrestin pathway. β -arrestin is the off switch. When β-arrestin gets recruited, then the receptor gets internalized and turns off the signal. What we had hypothesized was that by removing the off signal, could you keep the receptor in the on state longer, prevent this desensitization, and therefore enhance signaling, and therefore enhance efficacy. Our preclinical data certainly supports that. Our early phase I data supported that. In the couple of slides I'm going to show you, I think we're starting to see this translate into large phase II-B studies as well. Now, of course, phase III will be the final arbiter, of course, but it's very gratifying to see the translation from cells to animals to early studies to now potentially, hopefully, late stage studies as well. Let me start with the data from the 103 study. Some of you have already seen this since we had presented this at the ADA. Just to, again, recap, this is about 450 or so subjects randomized to five different arms, ranging from 4 mg all the way to 24 mg with the placebo. What you can see there on the left is the weight loss, percentage of weight loss over 48 weeks. The thing that strikes, at least us anyway, is the fact that the 24 mg dose arm continues its steady downward decline on a linear trajectory without a plateau. The number that you see at 48 weeks, 22.7%, is a number typically seen by other molecules at 72 weeks or 80 weeks. Now, of course, we recognize this is a phase II study and we have to do the phase III, which we're very excited that we've already initiated. It's called ENITH-1 and ENITH-2, and both have been recruiting extremely well. This trajectory, of course, we anticipate will continue furthermore. Potentially, it has the chance to be best in class. The other number there that's very important to focus on is the proportion of people achieving the critical target weight losses. Typically for organ benefits, 20% or more is considered to be the number. If you look at the proportion of people achieving 20% or more, that's nearly half the population of the 24 mg dose. What was more remarkable is that more than a quarter of people achieved 30% or more in less than a year. That's, again, another reason why we believe this has the potential to be best in class. Now, of course, again, we await the results of the phase III studies and those studies have been going really well. Some of this data and additional subcuts of this data, secondary analyses, exploratory analyses will all be presented actually this week in Milan, the EASD, and where we will talk about the incredible impact of enicepatide on blood pressure, on inflammation markers, on blood volume, and several other critical patient-related outcomes as well. Now, Teresa indicated a really fascinating result this morning, and I'm happy to give you the actual data behind that number, which is 2.65%. That number, of course, refers to the hemoglobin A1C reduction that we saw in what we call the 104 study. The 104 study was almost an identical study to the 103, which is about, again, 450 or so subjects, five arms, and randomized to enicepatide versus placebo. Here, what I'm showing you is just the top line results, largely because we want to actually preserve some of the data for a medical conference that's coming up. But the bottom line is this. What you see is that compared to placebo, there's a very steep and downward decline in hemoglobin A1C, which is reaching a value of around 2.65% in 48 weeks. What was even more remarkable, and again, as an endocrinologist having treated many of these patients, what we see typically is uncontrolled patients. Patients who actually have a hard time getting their blood sugar controlled despite being on two, three, sometimes even four anti-hyperglycemic medicines. If you ask the question, what proportion of people achieved good glycemic control, that's about 90% of people reach 6.5% or less, which is the threshold that the ADA has set to be non-diabetic. If you ask the question, what proportion of people are actually normal glycemic, even a higher bar, that's 62% of people. In those people that are poorly controlled, that is A1C greater than 8.5%, we saw a number of value of 4.13% decrease in A1C. Now, of course, I'll be the first one to say that cross-trial comparisons are fraught with problems, but I know that you guys will do the comparisons anyway, so we've done it for you here. With all the caveats of cross-study comparisons, what we just clearly see here is a set of numbers that gives us the confidence to say that this might be potentially best in disease for glycemic control. It's been a long time since insulin that you could actually reduce blood glucose to these levels without hypoglycemia. That was a really important metric for us. When you actually looked at the proportion of people that actually discontinued because of AEs or hypoglycemia or any other key issues, it was a staggeringly low number, 2%. In placebo, it was 0%. There were no level 2 hypoglycemia. Again, level 2 is the ADA definition of an objective readout of hypoglycemia. This is an actual blood measurement. No level 2 hypoglycemia. The safety and the tolerability was pretty much in line with what you would expect of an incretin class. When you put it all together, what we have now is an agent that we believe warrants a full test head-to-head against tirzepatide, at least in glycemic control. For that reason, we have embarked on a very large program. It ranges from looking at, of course, obesity, type 2 diabetes, glycemic control by itself, which we call the Eni-SPIRE program. We're also initiating in parallel in the first half of 2027, two large outcome studies, Enith-MACE, which includes a subpopulation of type 1 diabetes, which we believe is actually quite unique. Again, to the best of our knowledge, nobody's ever gone into a cardiovascular outcome study with both type 2 diabetes and type 1 diabetes. This is quite unique. Another unique differentiating factor where we believe that enicepatide could really differentiate is in heart failure with reduced ejection fraction. Again, many of you cover this field well, so you know that most people go into heart failure with preserved ejection fraction. We've taken a fairly bold step based on all of our data that enicepatide could really have the potential to be first in class for this indication. Again, of course, the phase III trials will dictate that. In summary, a large program, a broad program, a program that gives us the opportunity to be not only best in class, but best in disease, with the relevant outcome studies that's necessary to be able to claim something like that. Now I'm going to transition very quickly to an update on our oral GLP-1 program. The 996, CT-996, is an oral small molecule biased GLP-1 agonist. We've shown you some of the phase I data. We're actually happy to report that the EASD accepted some more of our phase I results, which we'll actually be presenting this week, but here you can see the preview of that. Again, in a small cohort of people with obesity and type 2 diabetes, it's only four weeks, but you can see here the very robust reductions in A1C and body weight loss in a relatively short period of time. When we did a dynamic test, like the mixed meal tolerance test, the glucose excursion was largely obliterated, meaning that there was no excursion of glucose. Very, very strong glycemic control effect. The PK supports once daily dosing, no food effect, which is very important. Ultimately, when we took all of that data into our phase II program, which was a very robust study, again, with roughly around 350- 400 people, tried multiple different doses, different regimens, different titrations, and the data that we've seen so far gives us confidence to advance this program now to phase III. That has allowed us to then develop what we call the MetORA program, which is going to be for chronic weight management to start with. We'll start with that. Plan is to initiate in the first half of 2027. We have an ongoing phase II-B study in obesity, overweight, and type 2 diabetes, which we're running right now. It's on track, and we will report that data when that comes out. Moving on to petrelintide. You'll recognize this from the last time that we were here. Long-acting amylin partnered with Zealand. The key thing to focus here on is the biological synergy that we expect with an amylin and an incretin. We know this biology, and it's been there for the last 20+ years, and that gives us the confidence that we can actually see that additivity when we actually combine a GLP-1 GIP with petrelintide. This would be our triple agonist in some ways when we do this in an FDC. But in the meantime, where we are focused on is to do this in a loose combination study, which we call the synergy program or the synergy study, where we will dose enicepatide and petrelintide in people with overweight and obesity and look for chronic or weight as the endpoint. Just as a quick recollection, you know this data from the ADA, 10.7% weight loss with petrelintide monotherapy, with a really excellent tolerability profile. The tolerability profile in this ZUPREME-1 results was essentially unremarkable. No vomiting, very low cases of nausea, generally considered to be quote-unquote placebo-like. The main news here is that now, based on all of that data, we've advanced this program. In fact, actually, you saw the release from Zealand, where we've taken this program into phase III as well. Moving quickly to zilebesiran. This is, again, just as a quick reminder, siRNA against angiotensinogen. Angiotensinogen is way upstream of the RAS pathway. This is a different kind of RAS, not the KRAS thing that you heard about, renin-angiotensin-aldosterone system. By blocking it very upstream, what the advantage it confers is that there is very little feedback that happens from aldosterone. This allows for very durable and sustained blood pressure control. Plus, with an siRNA, you can essentially knock down angiotensinogen also to very low levels. We've shown all of this data before, but what you haven't seen is the data that we presented at the ESC last month. What that shows you is this very sustained, and especially the reduction of blood pressure at nighttime. The reason why we want to focus on the nighttime is because amongst all the different parameters that influence cardiovascular mortality, it's how good your blood pressure is during the night that dictates that. If your blood pressure is very poorly controlled at night, there's a higher risk for cardiovascular mortality. This, you can see, whether it's in daytime or nighttime, there's a very clear sustained reduction. This gives us even further confidence to the ZENITH trial that's ongoing right now. ZENITH, again, as a quick reminder, 11,000-person cardiovascular outcome study, classical four-point MACE. Very happy to report that the enrollment is going phenomenally well. As Teresa said, there's tremendous interest from patients, from investigators, all across the board. We are well on track and again, looking forward to actually seeing how this all translates to cardiovascular outcomes. The good news is the meta-analysis indicates to us that for every 5 mm- 7 mm of mercury drop, you increase your relative risk reduction by 10%- 15%. A couple of quick things before I wrap up. Pegozafermin, this is simply to just remind you all that this is our FGF21 analog. There's no new data since 2025. Just to remind us that based on the data that we have currently seen, particularly this middle graph with the population of people that have an F4 at baseline, and to see roughly 45% of people actually achieve one-stage improvement in fibrosis, that was really encouraging. Again, small numbers, totally get it. It's important to see that signal and the fact that it can actually modulate, even at that late stage, gives us the confidence that FGF21, especially pegozafermin, can in fact be best in disease. Two studies are ongoing, which we call ENLIGHTEN fibrosis and ENLIGHTEN cirrhosis. They're going relatively well. Top-line data is expected now in the first half of 2028. Last but not least, just a quick word on urocortin. As I mentioned to you earlier, one of the key unmet needs that we're also trying to solve for is the improvement of body composition, especially in people who are losing a lot of weight. You'll note that most incretins will cause lean mass loss of around 20%- 25%. Urocortin-2 is a novel mechanism that has the first-in-class potential to really improve this body composition while also improving insulin sensitivity and other end-organ benefits, particularly on the heart. It works on a specific receptor called the CRF2. That selectivity is very important. Hanmi's molecule achieves that selectivity really well. That was one of the key reasons why we were excited about this specific molecule and this mechanism. The data on the right just gives you the initial hints. Now, of course, this is in preclinical species, but if we were to replicate this data in humans, this would be really exciting. What this data tells you is that when you compare urocortin by itself or urocortin to semaglutide and then urocortin plus semaglutide, which is the clear bar, you can see there's more weight loss with the combo. That weight loss is almost entirely driven by fat mass loss. That clear bar there on the lower left, that's just fat mass. On the right, you can see that the lean mass is not just unchanged, but it's actually increased a little bit. If that were to actually translate, that would be paradigm shifting. That is the reason why we focused on Hanmi and another reason why, based on the bar requirements that you heard from Teresa, we had to stop emugrobart. In conclusion, before I hand it to Teresa to bring us home, what I want to leave you with is that it is imperative that when we think about CVRM diseases, that we think about it as an interconnected web. The goal is here to address both the adiposity drivers as well as the adiposity independent drivers. The goal is also to not just treat, but to actually prevent. We can do that because of our diagnostics capabilities. We can do that because of our general capabilities that we have built together now with trying to bring in combinations of different kinds. Then ultimately, of course, the goal is to reduce the global burden of health, which CVRM diseases constitute nearly 60% of it. Overall, our focus, again, is to bring the best of internal science, the best of external science, and ultimately to develop these transformative therapies. With that, let me hand it to Teresa to bring us home, and we'll be ready for questions afterwards. Thank you. All right. Fortunately, we only have one more slide. I hope what you're taking away from today is momentum. Over the course of the last couple of years, we have put a lot of things into motion, whether it's R&D excellence, which looks at completely changing our entire R&D engine, applying the bar to all of our assets, dramatically reshaping our pipeline to our operational excellence, the investments that we're making in order to fuel our future, or our commercial excellence. We have a tremendous amount of tailwinds behind us as we head into the next period. For our on-market portfolio, we're making smart lifecycle management investment decisions that will allow us to extend those products and continue to serve additional patients. At the same time, we prepare for the launch of exciting new molecules like giredestrant, fenebrutinib, vamikibart, afimkibart, and all of the things that we have coming before the end of 2027. Then we just spent the afternoon really getting a sense of the tremendous pipeline that follows. As we think about the next couple of years ahead of us, we have a lot to deliver on. But the foundations are strong. The momentum is there. As I said previously, these are no longer just words on a page. They're not a strategy on a PowerPoint. They're not an ambition to be achieved. These are plans that are in action that we'll fulfill for patients. With that, I'm going to invite all of my partners in crime back up to the stage to go through our last Q&A. Nope, fine. Very good. I think then we will immediately start with the second Q&A session. Let's first go now a bit on the back, and then we come back to the front line. Emmanuel, you go next. No. Oh, okay. Now you can hear me. Emmanuel from Deutsche Bank. Thanks for taking the question. A couple of follow-ups on your impressive head-to-head clinical trial plans you announced across various therapeutic categories. Maybe bexobrutideg, you announced you're going phase III in the second line. It looks like that's head-to-head pirtobrutinib. Is the ultimate plan to go in front line, head-to-head with one of the current standard-of-care BTKs, like acalabrutinib or ibrutinib, for example? You announced plans with a VEGF x IL-6 DutaFab to go head-to-head with Vabysmo DME. Would you plan to do the same in AMD? Because that's obviously an important setting for that molecule. Enicepatide, you talked about now being confident enough to go head-to-head in diabetes. Will you also plan to do the same in obesity? I can start just with a head-to-head question. As we talked about, the clinical profile of the molecule, I think, gives us a lot of confidence that it can deliver the next generation of what CLL patients need. We're beginning in the refractory population based on the resistance data we showed at 86% response rate, but we think that there's much more to come. In terms of the specific design of a phase III to come, we're not commenting on that right now, but we're building combinations that allow us to get to first line. I can add to the enicepatide. Yeah, as I indicated, the data certainly warrants us to have that experiment against tirzepatide in glycemic control. We'll await additional data and then decide how the program evolves for other beyond in diabetes, whether we want to go head-to-head or not. At this point in time, we're going to look at this from a data-driven perspective. I'll obviously cover the VEGF x IL6. We've not made a decision yet to move into nAMD, but if we do, yes, Vabysmo would be the priority competitor. Thank you. Let's go on the back to your choose. Simon, please. Thank you. Simon Baker from Rothschild & Co Redburn. Two from me, please. Firstly, on the obesity market, a sort of big-picture question. Can you just give us your perspectives on how this market's evolving compared to your expectations when you started on this journey in terms of the mix between oral and injectable, the impact of price and the level of price and the effect that has on volume? Just really see how things have changed versus original expectations. Moving on to sefaxersen, could you just give us some more thoughts on how you're going to position that in IgAN? That is becoming an increasingly crowded area. Some more thoughts on how that fits in there would be much appreciated. Thank you. Wait, Teresa, you want to start with the pricing question? Yeah, sure. I mean, honestly, it's all pretty much in line with the bookends that we had when we made the decision to go into obesity. Some things have moved faster than maybe we anticipated, but I think we sort of always expected that we would get to a more consumer-driven, price-sensitive market and one where the shift to orals would eventually occur. Again, I don't think it's anything different than we assumed. Maybe a little bit faster, but all within the bookends. We always expected the market to be fragmented. Absolutely. In different parts because there was different mode of actions, different models, etc. We always expected it's going to be over time, not just one big market, but subsets of markets. Yeah, absolutely. We evolved in the portfolio towards that. Then maybe, sorry. No, go ahead. With regard to sefaxersen, yeah, obviously a lot of progress made in the IgA nephropathy field over the last few years. The APRIL/BAFF inhibitors have gotten, I think, most of the attention. Even if you look at the clinical trials of the APRIL/BAFF inhibitors, 50%- 60% of patients are not achieving adequate control based upon current clinical guidelines. That being the case, we think that there will be an important role for complement inhibitors in the treatment algorithm, including sefaxersen, which, again, we think has best-in-class potential as a complement inhibitor. Really, I think the focus will be in terms of first-line therapy on patients who have more active inflammation or at higher risk of progression. Then in second-line therapy, again, those 50%- 60% of patients who are actually not achieving adequate control. Again, just to point out, sefaxersen has a unique value proposition of being every four week subcu at-home administration. Okay. Can we go there, Matt? Thank you. It's Matthew Weston from UBS. Two questions for me, please. One in neuroscience and Brainshuttle CD20. I think it's been talked about, or certainly alluded about at Roche for a number of years. I'm surprised it's still in phase I with the anticipation of phase II. What's been the challenge? If we had a known mode of action, a known technology, why is it taking so long to move forward? The second question about the impressive phase III trial package in obesity and diabetes. We all got caught out when one of your competitors had an unexpected dosing schedule in its phase III trial designs with cagrisema. Can you comment whether or not we should just be expecting plain vanilla dose acceleration in the Roche phase III program? Maybe I can start with that because I'm sure that's on people's minds as well. Yeah, we're well aware of that. It's a fine balance, right? You want the flexibility. You want to allow for patients to have some opportunity to titrate themselves if you can. But all of our studies, all of our programs have had a requirement to get to a top dose. Now, how you get to that top dose, there's flexibility in that and adjustments that could be made, but the expectation is that they need to get to a top dose. Plus, we're also going to allow for enough time after they get to a top dose to have enough sort of follow-up period on the maintenance dose. Finally, I mean, I think one way, I think we had already alluded to this last year, but now we can confirm that we're actually incorporating the open label extension as part of our designs in our, certainly in our obesity programs. We're well aware, but we also recognize that it's a balance between what patient preferences are and how you actually recruit patients and keep them in these studies, which are challenging to do because they're all still on placebo. I can speak to the Brain shuttle CD20. Over time, there's been a careful design of the CDP, and it's definitely moving to phase II. Phase II is starting with Brain shuttle CD20, and we have early phase I data that shows benefit in reduction of B cells in CSF in patients. Can you give the microphone? Pass it on here in the second line. Graham Parry from Citi, a couple of follow-ups on giredestrant and one on fenebrutinib. On giredestrant, you talked about aromatase inhibitor discontinuation patients as an opportunity, but what about the in-situ patient population who just aren't doing well? Do you see them as switchable? If so, how many patients do you think that actually represents? Can you talk to the differences in the community versus the KOL settings and where you see the best place to target for this drug on launch? Also the time you think it's going to take to get patients onto reimbursement coverage if they're in the patient access program. On fenebrutinib, in your FDA pre-filing discussions, has the FDA asked for any additional suicidality risk or suicidal ideation data to be submitted during the review cycle? Thank you. Maybe I'll start on part of the giredestrant question. The short answer is, certainly, we see an opportunity on the switch front. Actually, one of the phase III-B studies that I described will specifically be looking at the feasibility of switching for patients who are, in fact, not tolerating, not doing well on the CDK-based combination. Also, when we think about the discussions that we've had thus far, of course, we don't comment on detail in FDA discussions, but we don't anticipate any label restrictions around switching. Definitely a live question that will require additional studies, but we don't see any intrinsic barriers to that. Maybe Teresa or Wendy, we want to comment on the other side of the question. Sounds good. I can maybe pick it up in terms of two, I think, two aspects, and tell me if I'm what you're thinking about. First is, A, just thinking about in situ, we definitely are looking at understanding more in terms of the market size. I think the market size needs to be balanced also in terms of what the, I would say, willingness to treat, right, relative to the current therapies as well. This is something that we're currently looking at. In terms of community versus KOL or from an academic business, I do think that given our legacy and our leadership in HER2, this is an area that we are intimately familiar with. Unlocking the community is going to be a key challenge for us, I mean, a key opportunity for us when we're thinking about the side effect profile and tolerability that we currently are seeing with AIs and as well as CDK4/6. Last question, in terms of patient access, we definitely recognize that it is a, as we talked about, being part D, patient access will be key in terms of how we ensure that we're bridging, right, from the, not only from the existing patients, but also from clinical trials and onwards, but we're gained and we're definitely ready to serve. I can address your suicidality question for fenebrutinib. First of all, patient safety is Roche's highest priority. We have carefully examined the data. We have submitted a file with FDA, and they've not asked for any additional data. We've had great collaboration with FDA, and we expect acceptance of the file anytime soon. Okay. James? Right, James Quigley from Goldman Sachs. I've got two questions, one for Manu on enicepatide and one for Hideki on fenebrutinib. On enicepatide, would you be able to let us know what the placebo response was in the weight loss, on weight loss, but then also the placebo reduction HbA1c for those with a baseline of greater than 8.5%? Also, what in the data gave you more confidence about moving forward in heart failure with reduced ejection fraction versus competitors looking at preserved ejection fraction? On fenebrutinib, Hideki, would you be able to give us an idea of what the 24-week EDSS progression was as part of the composite primary endpoints? When you look at the remibrutinib press release, and clearly we need to see the detailed data first, but how do you think the MS community will think about inconsistent results between the stat sig benefit at 12-week versus 24-week on EDSS? For the CD20s, it looked pretty consistent for Ocrevus and KESIMPTA. What could this suggest about the quality of the data, or are there any factors that we need to consider about BTK inhibitors that may mean the progression benefit takes longer to come through on EDSS? Starting with enicepatide. As I said before, we've tried to deliberately not provide too much data, largely because we are, not because we don't want to share it's just that there are embargo rules, as you know, right? All I can tell you is that the responses are pretty much what you've seen in the total population. Nothing surprising. They're all pretty much in line with what you would expect of a placebo. There was no strange behavior of the placebo patients or any of those things. It was really reassuring with a very clean data set. In regards to the data that gave us confidence in the ref, there was multiple pieces of data. The first is, of course, our own data, which will be presented at the EASD, actually, is the very significant benefits on plasma volume. That actually is a very important parameter that correlates to how well that people with rEF do. Blood pressure changes, of course. Inflammatory markers. I think the main thing is really looking at the real-world datasets. When we look at the collective datasets that's available and put all of that together with our data, it gives us confidence that we would potentially have a benefit in the rEF as well. The final piece I will tell you, which actually worked out in retrospect in our favor, is that AHA has now actually reclassified heart failure as simply rEF and pEF. There is no such thing as minimally rEF. That classification actually helps us a lot because the population that we're actually studying is exactly what the AHA definition of reduced ejection fraction is. That gives us even further credence to pursue this indication. With respect to your 24-week disability progression data for fenebrutinib, that basically did a lot of what Manu just said. That data will be upcoming at a forthcoming scientific conference and largely consistent with what I've showed today. Can we pass the microphone to Justin? Maybe you go first. Just a quick one for Levi. Justin Smith, Bernstein. On the B7-H3 ADC and the DAR, you emphasize the eight. Sorry if this is a bit simplistic, but are you encouraging us to think about sort of Kadcyla and HER2, or is it a completely different way to think about it? Obviously, it's differentiating what you have, but we'd just love to get your thoughts on how important that ratio is. Probably both. There are two aspects that I mentioned, probably the one I mentioned the most because we think it's particularly relevant, is the linker, the hydrophilic linker, and the ability to prevent leakage of the linker in the circulation. From a side effect profile, that's probably perhaps the most important. In fact, if you look at the, obviously, we don't have head-to-head comparison, but it does appear that perhaps that is creating a bit of it. Now, the DAR could also help in as much that, of course, you're that much better able to deliver the payload into the tumor with that favorable linker situation. I would say maybe on balance, we might even, maybe the linker, the integrity of the linker in the circulation that might be even more important from a side effect profile. James? Maybe if I can build on the cross-this, you can see in all the B7-H3, by far, Tam-Peli actually has the highest DAR ratio, right, in comparison to others. I think in combination with thinking about the hydrophilic linker, we're seeing a really good favorable in terms of tolerability profile, which allows us to combine as well as an efficacy potential as well. Great. Thanks, James Gordon from Barclays, and thanks for taking the follow-on questions. One was also on B7-H3s, but also about increasingly fast followers. I think there's about 10 different B7-H3s in the clinic across companies. Most of them are from Chinese biotech companies. I heard comments about payload, DAR, and linker being potentially differentiated. But do you think there is a bit of a risk that you're going to get lots more fast followers, particularly in ADCs going for the same targets, and the returns are going to be eroded even before the patents go? Does that change how you think about what to take forward or what to license in? That there seem to be certain areas that look like there's been loads of people going for them. The other question was in CVRMs. For obesity orals, at least one of your competitors thinks that the oral market, what's going to be really important for the next wave is going to be able to combine multiple oral ingredients. Do you think that's the case? It's not just going to be about oral monotherapies. It's going to be oral combos. If so, do you need some more oral co-ingredients to go with, to go for other cardiovascular comorbidities? Maybe I'll briefly take the first question. We often have the discussion about the competitive landscape, but we usually end up with, if we were to use the competitive landscape as a reason not to do something, we would do very little because there are so many examples that you gave where there are many shots on goal, many companies in China that are coming with fast followers. That by itself can't be a deterrent. What we have to trust is our ability to evaluate the opportunities and look for components that may signal differentiation. We think thus far with Tam-Peli, we're seeing that. But of course, at the end, it'll be important to have the data speed, but also execution, to be able to execute these trials rapidly, but also high quality. Yeah. Thank you for asking that question on the orals. In one of the slides I showed, that is one of our critical parts of our strategy, right? We don't think it's an either-or injectable oral. The market is highly segmented. The needs are varied, and patients have different needs along their long journey. For that reason, we think that orals are a critical pillar of our strategy, both as monotherapy, and certainly CT-996 can represent that. We're also, as part of the other pillar of our strategy that I showed you, is combinations. Combinations can come in the form of the usual suspects, like a GIP agonist, for example, could make an oral enicepatide, but also given the interconnected CVRM web, if you will, having agents that can address other comorbidities while you have an anchor or the foundation of a GLP-1 agonist makes it very attractive. We believe that both are equally possible. The final piece I would just say is that a lot of people worry that, well, does the injectable sort of take away the oral market or vice versa? What we're actually seeing is the opposite. There's a market expansion, largely because the orals have actually started to come in. We anticipate that that market size will further increase even more so as more people come into the fold. You also might imagine a situation where someone who maybe was a little bit intimidated by a shot is willing to start with a pill, realizes they're getting what they want, but maybe not as much, and then is happy to move on to an injection. I think you'll also, that market expansion will also help the injectable category later on. Thank you. Just pass it on, please. Thank you, Kerry Holford at Berenberg. Question for you, Manu, following up on what we've just discussed on obesity. Limited time, I think, today spent on petrelintide monotherapy. Just keen to hear your level of excitement around that approach. On that product in particular, do you need all three ZUPREME phase IIIs to file, or will it just be there's a traditional obesity and obesity plus diabetes that you feel you need to advance to the regulator? Finally, just an update on your plans for phase III-B for that molecule. Sure. The first is, obviously, we're very excited about petrelintide. I mean, it's a critical part of what we call a foundational pillar. Just like GLP-1 is a foundational pillar and FGF21 is a pillar, amylins are a pillar too. The data that we've seen indicates to us that the experience that people want to have can be varied. Again, it comes back to the heterogeneity and the different needs of patient segments. For those that are looking to have what we would call a gentler weight loss experience, they're wanting to lose maybe a double-digit weight loss, but with very good tolerability, or try something else as a different mechanism, amylin offers that. Having that in our portfolio is absolutely vital. That is the reason why the program has moved into phase III. Now, do you need those three ZUPREMEs to file? Because chronic weight management requires you to show both the efficacy in people with and without type 2 diabetes. Even though there's three studies, actually the package is for chronic weight management. The third one, because we want to file globally, the EMA still has a requirement to show cardiovascular safety. The third ZUPREME study is to support that. Again, all of this underscores our excitement for the program that we're moving forward with the monotherapy. It's important that we have that optionality in our portfolio. Now, regarding the other question about, sorry, can you remind me of the— Phase III-B. The phase III, yeah. That's under discussion right now. We'll be able to update you once we have that finalized. Again, the thought here is if we're going to go forward with the mono, we want to ensure that we have the best supportive evidence that we can take forward to develop the mono. There will be more details to come on the phase III-B. Hi, thanks. Richard Vosser of JPMorgan. Two, please. One on the TL1A class and your product. Just wondering how differentiated this really is, the class to the IL-23s. Where does it fit within the treatment paradigm? Secondly, on SKYGLO, could you talk about whether you've had an interim analysis there for SKYGLO and how you see the market subdividing between the Polivy combination and the one that your competitor is going for, which excludes Polivy? Thanks very much. Happy to speak about afimkibart first. Obviously, great progress in the IBD field over the past decade or so, but again, still that significant efficacy ceiling that we haven't been able to crack. We do believe that TL1A actually will become standard of care for inflammatory bowel disease. But even as standard of care monotherapy, it may be limited in terms of its ability to crack the overall efficacy ceiling. That's why we're also focused on rational combinations, including that TL1A backbone. In terms of where TL1A as monotherapy will actually sit in the treatment algorithm, our clinical development plan looks both at patients who are advanced therapy naive and who are advanced therapy experienced. Only time and data, I think, will tell exactly where the TL1As fit into the treatment algorithm. Again, we believe that to really smash that efficacy ceiling, we're looking ahead to next-generation molecules, including bispecifics based on the TL1A backbone. In terms of SKYGLO, I can't comment on the interim analysis in particular, but I'd mention how excited we are for the readout next year at 2027. In terms of the competitive positioning, I think you highlighted a key element, which is that the control and the underlying backbone that we're building on is fundamentally different in our study versus competitors. Our belief is that the Polivy R-CHP is the standard and has the best efficacy, as I highlighted in that sort of legacy of progress through DLBCL. It's prescribed as the most common regimen across multiple regions of the world. It's endorsed by 20 guidelines. Our belief is that we're building on the best standard against the best standard. There was a question earlier today about sort of how we think about head-to-head comparisons. I think you see in this study, as well as many of the others, routinely going up against what is the best standard. That is maybe relevant to the differentiation question that was brought up before about how do we establish ourselves in a marketplace that's rapidly competitive is by beating the best standard of care and not being vulnerable to questions about whether we've taken an easier route to filing. With the clarity of the data we get from SKYGLO, we'll be able to have one more standard of care change for DLBCL. Sachin? Sachin Jain, Bank of America. A few questions, please. Firstly on fenebrutinib, that's a prior question. You said you hadn't had any further questions from the FDA on CNS. Would you have expected those by this stage, given the file's not even accepted yet? Any color you can give on the level of comfort FDA has with the different trial baselines on sort of CNS risk that differs your program from the others as an explanation for that suicidality? I guess related, we're obviously awaiting the competitor data in a few weeks and we'll compare and contrast efficacy. How do you think the absence or not of monitoring will impact commercial positioning? Obviously, they're talking to being approved for that monitoring. Second one for Larry on the TL1A combination strategy. Very clear on your follow-on molecules, but could you talk to how you might explore free combinations that accelerate a combination strategy to build beyond your monotherapy launch? Obviously, a variety of companies have got the bispecifics a little bit further ahead than you. Is there anything you can do commercially to accelerate a combination strategy? Last quick one for Chris. You talked about effective comparators. I don't know whether that was a comment to the Kodiak data just out, but I wonder if you could just comment to that. I know you've had a lot of time, but any quick comments? Thanks. I could take the first question and then hand it over to you for the commercial question on fenebrutinib. We don't comment about specific interactions with FDA, but we have had a highly collaborative and really great dialogue with FDA. Like I said, they have the entire file and we expect acceptance anytime soon. In terms of the liver monitoring, I mean, in general, it is pretty standard that you get in your label what you studied in your trial. If you had liver monitoring in your trial, chances are you're going to have liver monitoring in your label. I think it's important for us to remember that neurologists are very familiar with liver monitoring. They do it all the time. It's something that they have incorporated into their practice. I think it gives them a high degree of confidence that they're doing the best thing for their patients to keep them safe. I don't think we see any particular barrier to prescribing by having that liver monitoring. In fact, I do think it gives physicians comfort. I can speak just quickly to the TL1A combinations. Obviously, I think in these situations, it's just as important or maybe even more important to get it right as it is to get it fast. We believe that we have a great molecule that's in the lead in our pipeline. We have other molecules as well that are taking similar approaches. We'll see the data in 2027 and look for opportunities to accelerate from there. Quick comment on the Kodiak data. Very fresh off the press. I think it was coming in as I was speaking. I'd say my overall impressions, when we compare Vabysmo with aflibercept, Vabysmo consistently dries better than aflibercept. That data did not exist that I saw in the Kodiak data. That was in nAMD. They have no data in DME, which makes it very challenging to get any product in mass usage in the key markets. More to come, I think. Maybe just to add a tiny little bit of color to that. I was with a group of physicians two weeks ago and a retinal specialist. What they said overwhelmingly was we need to start studying against the standard of care and that standard of care has changed. I think in particular in that specialty, there's a lot of drive to actually make sure we're running the right trials. Urban? Yes, thanks a lot. Urban Fritsche from ZKB. First on Alzheimer's, you didn't talk a lot about the impact of the availability of the pTau217 test on clinical development, but a bit more concrete. Did the availability of the test help to accelerate the phase III patient recruitment for trontinemab? Also on Alzheimer's, what's your thinking on the combinability of nivegacetor plus trontinemab? I think there is a trial going on combining that with KISUNLA, if I'm right. On cardiovascular, did I get it right? The emugrobart discontinuation was quite an opportunistic decision for the urocortin-2 analog. How much of that actually was driven by in-licensing that? Let me take that second one first. It wasn't. I mean, we evaluate opportunities based off of the science, based off of the need, what it is that we're trying to solve for. Ultimately, the data from GYMINDA didn't meet the bar for us. That study was already well in play even before we even started to even contemplate about urocortin-2. The way we approach it is to look at it from a disease pathway or a pathobiology perspective, which is what are we really trying to solve for? The thing that we're trying to solve for is the preservation of lean mass while you're preferentially decreasing fat mass. What are all the different mechanisms? What are all the different targets that can claim that space is a critical factor for us? Then it has to certainly meet the data standards that we want to have to achieve that. Unfortunately, in the case of emugrobart, it didn't meet that bar. It was not one or the other. It was entirely driven by the data. Your question about pTau217, the blood-based biomarker. Yes, absolutely. That has helped in terms of recruiting patients. I didn't highlight in detail, but we have a pre-screener study called TRAVELLER. The pTau217 is a key component of that TRAVELLER pre-screener study. That's helped us identify patients to go into early AD, so counter studies and/or into the prevention study, the PrevenTRON study. It really has helped in terms of acceleration. Your question about combination. We feel that trontinemab is a foundational treatment and we are looking at potential combinations in the future with trontinemab. GSM, nivegacetor could be, but we haven't come to that decision yet. We're certainly exploring that. Okay. Maybe Luisa should get the final one, the first one on the final? So patient. Thank you, Bruno. Just a little bit more on giredestrant. Is there any better understanding for the failure in the first line? I know at ASCO there was not really a clear explanation for that. I'm asking because I see you're now combining with CDK4, with a competitor. Anything to give us confidence around that and also the decision-making? Because I think you had a CDK4 that you've now stopped in phase I. The choice to go with a competitor. The final question would be, perhaps Teresa, what are the key risks to investing in the new disease areas and how is Roche mitigating those risks and defending Alan's margin? Maybe I'll start on the short answer is there's nothing new to report on the sort of mechanistic explanation for the failure of persevERA. One thing, as you know, there was the numeric, there's a numeric difference, but it wasn't wide enough to hit with the power that we had. There was also the late separation. That has led to hypotheses about why a late separation as opposed to an early separation. We don't know. Of course, one hypothesis would be maybe ESR1 mutations are occurring in the aromatase inhibitor arm, but not in the giredestrant arm. More to come on that. Obviously, we're exploring that. I would say if you step back 10,000 ft that there are important differences between the kind of metastatic setting and the early breast cancer setting, namely the tumor burden, the microenvironment. For example, again, we don't know, but I could say when you have a high tumor burden, you already have a lot of resistance mechanisms present, including ESR1 mutations that are kind of latent. Whereas in the early setting, you don't actually have that. So you may have an opportunity to reveal selective vulnerabilities of both CDK and hitting the ER in ways that just don't play out in the late setting. We think there's still equi poise to actually ask the question of the combination in the early stage. This was one of the reasons to do the study with Pfizer. Absolutely. In terms of risks, I mean, the primary risk is always the most theoretical and generic one. That's making sure that you have a clinically differentiated asset to bring into a new space. If you're going to disrupt, you want to do that with a medicine that's truly going to change the standard of care. I hope what you've taken away from today is a sense that we really do have a strong belief in those new assets that we're bringing forward, that they do have the potential to disrupt. When you think about scale in terms of manufacturing, commercialization, we've been steadily making those investments over the course of the last couple of years, sort of investing ahead of the pipeline to ensure that the foundations are there when those medicines hit, that we're not hit with a giant cost all at once, very carefully managing our capital investments and our margin. Honestly, we don't need all of them to be successful. I mean, you look at the sheer breadth of what we have in the portfolio that's already on the market, de-risked with meaningful phase III data, we have a clear path to growth. When we look at what we have ahead of us with these amazing new assets, I think what we have are a lot of things that give us a lot of confidence that we really will be able to come to market with something that's materially different. I have a lot of confidence in the manufacturing and commercial engine that we've built to make sure that they ultimately make it to patients. Excellent. I think we have the perfect closing remark. With that, I think we will close the broadcasted part of the event. I would like to thank all our speakers for all their dedication and hard work every day to make this company go in the interest of patients. I also want to thank a couple of people from my team who have been working for the last weeks to make this event possible. From the IR team, it's Jan-Philipp Schwarzhans, who worked on the overall strategy and the R&D excellence stacks. Jon Bayard, who took care of the group finance stack. Anna Hupalowska, who worked with on Aviv stack, on AI and early drug development. Rafal Pawlowski, who was responsible for the oncology deck and managed the overall slide deck. Loren Kalm, who was responsible for hematology, neurology, ophthalmology slide decks, and also preparing Q&A decks, which needed to be updated constantly as we had a daily news flow, which was very intense for the last couple of weeks. Natalie Chen, who worked on neurology and ophthalmology slide decks and also updated our epidemiology appendix slides. If you have not spotted them, they are again uploaded on the homepage, the latest and greatest. Sabine Borngräber and [Ada Logiri], they were responsible for the immunology and CVRM decks and also updated the pipeline appendix, also uploaded on the homepage. Last but not least, from the IR back office, it's to call out Melanie Wolf, Beatrice Hau, Eva Losert, [Celeste Lloyd] for event organization. I think the people on site here will have another half an hour for the buffet reception. All the management will be around, so they will take additional questions. Thank you.