Morning everyone, and thanks for joining on day three of our Biopharma Innovation Summit. I'm very pleased to start the day this morning with a fireside chat with Kaivan Khavandi, who's Head of Respiratory, Immunology and Inflammation R&D at GSK. Quite a lot of disease areas there. So Kaivan, thank you for joining us. Maybe just to kick off, if you could just give us a brief intro on yourself and your background, which just for people who may be less familiar. Sure, yeah. I'm a physician scientist trained in inflammatory mechanisms of cardiopulmonary risk. Had a lab at King's College London and practice at Guy's and St Thomas'. I've worked in the industry across large pharma, Pfizer, GSK across I and I, internal medicine, cardiovascular medicine, and also a brief stint at BenevolentAI as CMO. At GSK, I lead an organization that spans from target concepts through to approval for specialty medicines, as Rajan said, spanning respiratory immunology and inflammation, and also lead an organization cross enterprise at GSK for translational sciences. Got it. As in within those areas, there's obviously a lot of catalysts this year. Maybe we'll start with camlipixant. We've got the phase III readout coming later this year. Before we get into the details, can we just talk through the opportunity in chronic cough? I think it's something that's still debated amongst investors and in the market, so it'd be helpful to get your perspectives there. Yeah, sure. Chronic cough is very prevalent. Refractory chronic cough has a prevalence of about 40 million globally. It's debilitating. You know, this is a cough that's defined as lasting more than eight weeks, but actually, in many of those individuals, it relates to bouts of cough that are frankly incompatible with normal life. 100 coughs an hour. You can imagine what that looks like if, for example, I had that symptom now, it wouldn't be feasible obviously to participate in this panel. You're driving on a motorway. The consequences we see from that, as well as daily activity of living, is things like incontinence, rib fractures. Interestingly, even in our own program we can see the patients that are being recruited into the study have been on some mad things like opioids and neuropathic meds. The patients are desperate. There's a significant demand for a safe and effective medicine. The other point I'd add is that if you go to any academic respiratory center, there'll be a department that's focused on obstructive lung disease, there'll be a department that's focused on interstitial lung disease, and there'll be a department that's focused on cough. The respiratory community is already heavily vested in that space, but they haven't had an effective therapeutic. Whilst it might feel novel to investors and to some degree to large pharma, actually the demand has been there for decades. Okay. Maybe just in terms of the target product profile, could you just run us through that? With Merck's gefapixant, which was obviously the same mechanism, there was a 15% reduction in placebo-adjusted 24-hour cough frequency at 24 weeks. Is that a fair bar for camlipixant? Yeah. If you think about a 15% relative risk reduction, and you're talking about patients that have, in our study, have been enriched for those with over 20 coughs per hour, it relates to a very clinically important difference. Gefapixant, you know, to many, to a large extent actually de-risks the efficacy of the pathway as you described. Obviously in the U.S. at least, we're unable to translate that into a compelling benefit risk proposition. Obviously their program was compromised because of the lack of selectivity for P2X3, and the taste disturbance was problematic. The 15% threshold, we believe, and this has been substantiated by investigators and the external community, is certainly clinically important. Okay. The other thing that's notable here is that if you look at the phase II data, both for gefapixant and for camlipixant, you're at sort of 30+ risk reductions. Why would that be lower in a phase III trial? I mean, if you think about refractory chronic cough in the context of, you know, analogous diseases like pain or maybe IBS, migraine, there is almost universally a discounting of placebo-adjusted efficacy when you move into pivotal programs with longer treatment duration. That is a placebo-adjusted target value. The change from baseline, we would probably expect to be similar to the data you described in phase II. It's really a function of the operating characteristics of the pivotal program. Okay. I guess the other thing, looking at the phase IIs, so the camlipixant data, which was I think run by BELLUS right before the acquisition, that was a reduction at week four versus gefapixant, which used a 12-week endpoint, and again, it was kind of comparable reductions. Is that speed of onset important here? Yeah, I think so. Again, you know, as a sort of base assumption, we might assume similar efficacy to gefapixant, but the benefit risk calculus is different because of the selectivity. I think speed of onset is important. Durability of response across obviously the primary efficacy time points designed into CALM-1 and CALM-2, which is up to 24 weeks, is important. Okay. Just on that tolerability, so taste disturbance, although it's clear from the phase II that camlipixant has a lower rate of taste disturbances there. What do you think is acceptable from a clinical perspective? Obviously just bearing in mind that there were some issues with the gefapixant and CALM around whether that trial was actually sufficiently blinded. Yeah. Gefapixant's experienced a very problematic level of taste disturbance. You know, it's over 50%, I think over 60%. Imagine running a study where, you know, you're meant to be blinded to the active arm, and six out of 10 of those patients are experiencing taste disturbance. Highly problematic for the conduct of a pivotal well-controlled study. That hasn't been a problem for us. You know, we see the blinded data, it's low. In terms of what's acceptable then for patients, you know, we're expecting based on the order of magnitude greater, in fact hundredfold greater selectivity. We were expecting an order of magnitude lower problematic taste disturbance and almost no taste disturbance that would result in discontinuation. We're confident that's the profile we're gonna see. That would relate to, you know, 7% or less. I expect we'll see even lower level of taste disturbance with camlipixant. Just switching to the phase IIIs which are expected this year. There's CALM-1, there's CALM-2. Can you just go back a little bit and remind us what the rationale is for the two trials here, and then also why CALM-2 is now reading out later than CALM-1? Yeah. BELLUS initiated the studies and, you know, given the environment they were in, they were staggered. That's the only reason why CALM-1 and CALM-2 are, you know, not reading out at the same time. Since we took over the studies, we have increased the sample size for a couple of reasons. One was that there's an appreciation that as part of the gefapixant AdCom, the FDA would want to see data on patient-reported outcomes with an appropriate clinical instrument, which we've introduced into the study with the Chronic Cough Diary, as well as powering for 24-hour cough frequency as a quantitative measure, we wanted to make sure we were able to evaluate measures related to the CCD. That results in an increase in sample size from, I think, 600-something to over 700. The second was really a positive interaction we had with the FDA, which was to agree that we would enrich the study three to one. Three being the proportion of patients with over 20 coughs per hour at baseline, and the remainder being eight to 20. And so again, to translate that ratio forward, we then increased the sample size further. CALM-1's completed. CALM-2 has actually. I saw your comment Raj on it beforehand. CALM-2 has completed enrollments. We're just. It's open on clinicaltrials.gov 'cause of the long-term extension. Okay. You beat me to that question. Just on that enrichment, is that both in CALM-1 and CALM-2? Yep. Okay. The other thing, the other difference in the studies is obviously timing of the endpoint. CALM-1 is cough frequency at week 12. CALM-2 is cough frequency at 24 weeks. Could you just kinda talk to the rationale for those two different endpoints? Is that an FDA requirement? That's what was agreed with FDA at phase II. You know, there's obviously limited precedents for refractory chronic cough, but given the characteristics that I described, I think it was important to have confirmatory efficacy in a second pivotal study. I don't think there's a line in the sand in terms of what primary efficacy you would need to get up to for to demonstrate benefit risk. Obviously in one study we've got 12 weeks. We felt that 24 weeks was an appropriate time point to support durability of response. Obviously the second point is ensuring there's an adequate safety data space for the file. Okay. Just on the two pivotal trial, the trials again, is there the requirement from the FDA that you need two positive trials to file here? Well, again, limited precedents. We'd agreed with them that we would conduct two pivotal studies. We'll evaluate the primary efficacy individually in each study. We will of course then pool data. As you know, the FDA's lens will be one of benefit risk, and they'll look at a range of efficacy measures and safety measures. I suspect we would obviously pool across the 12-week endpoints across both studies and then look to what we can substantiate from 24 weeks with CALM-2. Okay. Good. Just on the trial eligibility. There's refractory chronic cough and there's also unexplained chronic cough. Are you confident that you have enough sort of a dispersion of both of those patient groups? How should we think about potential differences in efficacy, if any? Wouldn't expect to see differences in efficacy. Again, as we've characterized the population, there's a high burden of cough. Everybody in the study has at least eight coughs per hour at baseline. We've been clear in terms of exclusion criteria to make sure it's not confounded by patients with, for example, uncontrolled asthma or fluctuating underlying disease activity related to something like interstitial lung disease. Okay. We've been clear around comorbidities. We've been clear around the burden of cough that's coming into the study. Beyond that, I think we'd be expecting to see a consistent efficacy across the population. Okay. You alluded to it earlier, but gefapixant, there was an AdCom ahead of the or non-approval as it turned out. Is it reasonable to expect that there will be an FDA AdCom for camlipixant? I don't know. It depends on obviously when we've got the data from CALM-1 and CALM-2. I think if there was, you know, the learnings that we took from gefapixant position as well. You know, we've been able to anticipate some of the things that whether it's the FDA reviews or an external expert panel would be asking. I mentioned one of those earlier, which was, you know, as well as the intrinsic properties of camlipixant being favorable to gefapixant, we were able to enrich the study based on cough frequency that was coming in. Really importantly, this Chronic Cough Diary that we had an opportunity to align with the FDA and introduce into the study as a key secondary endpoint. Okay. I guess the two other concerns with gefapixant at the outcome were firstly, there was a question as to whether the trial was truly blinded because of the AE profile, but it sounds like you're quite confident that that's been solved for through the selectivity. Yeah. There's also sort of these high placebo responses, which was an issue that was raised as well as the variability in baseline cough frequency. Can you just kind of- Yeah. Walk us through why you're confident that you won't have those issues? You will always have a placebo response in a setting like this. It's to be expected, and I think it just comes back to the point we made earlier that we, you know, there might be a target value you'd expect as a change from baseline in your active group. But to account for placebo behavior in a study like this, you know, we've made sure two things. One is we had a placebo run-in. So you can imagine then in your day one onwards, you've already accounted for that acute placebo response, which gefapixant weren't able to. We're also then able to exclude patients that had an exaggerated placebo response as part of the run-in. But also how we think about clinically important target values and comes back to the 15% being a placebo-adjusted threshold. Okay. Then final one on camlipixant. Just I think there's a concern among investors as well, and I think it was actually raised again at the gefapixant AdCom from one of the speakers there. That there's a risk that treating for chronic cough is masking or there's a misdiagnosis of a more serious underlying condition. Again, thinking as a former physician, how do you think that GSK can solve for that? Well, it'd be quite exciting for us. Obviously from a trial point of view, as I say, we were quite disciplined around making sure that we exclude patients with interstitial lung disease, chronic obstructive pulmonary disease, and uncontrolled asthma, CF, chronic bronchitis. Our trial population is clean in that respect that we're not gonna be confounded by fluctuating underlying disease activity. In any setting like this, of course, the sort of clinical perspective will be you need to treat the underlying disease. That's always kind of easier said than done. You know, it's the same thing if we come to talk about the broader pharma opportunity in pulmonary hypertension. Treating the underlying disease is being done passively as part of clinical management, and these patients are then presenting with highly burdensome cough on top of that standard of care. Actually, no, I think from a clinical point of view, these patients are presenting, they might be under a pulmonologist, and they're still presenting with coughs of, you know, over 20 per hour. I think clinical management obviously would then relate to both managing the underlying disease and providing them with an efficacious medicine for the cough. Okay. Maybe we'll move on to bepirovirsen, which is another kinda key data point this year. We're expecting the full data at EASL. Before we go to that, I think your guidance was to file for approval with the FDA in Q1. I think we've got, what? A week left of Q1, technically. Could you confirm if you have filed now? We're on track. We normally disclose when we get acceptance to file. We're on track to disclose acceptance to file within Q1. Okay, perfect. In terms of the top-line data, you've announced the top line is a positive trial. I think you've previously talked to, and at least Tony has, a 15% functional cure rate, as being clinically meaningful in this setting. Why is it 15% that is clinically meaningful and, how should we think about the kind of implementation in the clinic? Yeah. 15% is related to a very high bar of functional cure, which is undetectable surface antigen, importantly, six months off all treatments, including nucleoside analogs. What that relates to is a clear quantitative prediction of improved outcomes for hepatocellular carcinoma. Obviously, a 70% reduction in the likelihood of hard endpoints related to malignancy and in turn mortality. When we think about clinical importance, you know, it's a combination of responder proportion and the target value you've set. In this instance, the target value has cure in the title, so it's an ambitious endpoint. I think it's very easy to qualify why 15% is important. To put that into context, nucleoside analogs get to less than 1%. You'll see when we present the data at EASL in a controlled trial setting exactly what nucleoside analogs are able to achieve against that endpoint. It's negligible. Yeah, this is a step change in terms of the efficacy that we can offer patients. Okay. Beyond that functional cure rate, when we do see the data, is there anything else that we should be focused on to kind of put the profile into context? We've designed a study where we're able to demonstrate durability of response as well. Some of these patients are going into a long-term follow-up. I think the combination of achieving functional cure at that target value and maintaining it is the other sort of consideration that I would focus on. Just in terms of B-Well and B-Well 2, which were the two trials. To my eye, they looked pretty similar in terms of design eligibility. Is there any reason we should expect differences in efficacy between the two? Is this a scenario? Peak opportunity in terms of the commercial opportunity here. I think you've guided to $2 billion in peak sales potential. Could you just walk us through how we get there? Is this something that will take a few years of education, or is it something that there is a warehoused population of patients that are waiting for the product? It depends on the geography. While I think there's an acknowledgement that the burden of chronic hepatitis B is enormous in low and middle income countries, and there is obviously an opportunity there around awareness and expanding the treatable markets. Actually, the currently diagnosed population is skewed towards markets like the U.S. I think you've got an attractive proposition where you've got markets like the U.S. that are ready to go. They've diagnosed the patient population, and they're looking for an innovative new therapeutic. And then you've got the low and middle income countries where, as you say, I think there's an expansion of diagnosis that's gonna come with an effective therapeutic. Actually, when you look at the sort of relative contributions that we're thinking about, it's not as atypical as I guess that some of the speculation I've seen externally. Where it's actually a fairly representative contribution from markets like the U.S. versus markets like China. Okay, that makes sense. The other thing I wanted to talk about is some of the deal activity that you've been involved in relatively recently. We'll run through some of the deals and some of the assets that you've bought in a second. But I just wanted to get an understanding in terms of your contribution and kind of the R&D organization more broadly. I think it's been a very clear strategy that Luke's communicated that you're looking for assets which are validated from a biological perspective or the mechanistic perspective, where there's white space where you can improve. I'd just be interested in how the R&D organization sort of is involved in that process. Well, I guess the first thing to say is the business development organization reports into Tony Wood and the R&D organization. Christian and I sit on the same leadership team, so we work very closely with one another. R&D has strategies within each category for vaccines, ID, oncology, and specialty. Those strategies inform the search and evaluation. Then we have a very well-oiled machine actually, in terms of then triaging against criteria that ranges from, as you say, translational confidence through to unmet need, through to market proposition, through to development tractability. I've worked in, as I mentioned, a few organizations, consulted many more organizations. I think it is truly a finely tuned system at this point. Then that's then triaged upwards to the R&D leaders, so myself for specialty. Beyond that, we'll come to Nina, Tony, and Luke. Luke obviously does have ideas of his own, so he will also initiate interest in certain activities. Again, aligns with the overarching R&D strategy, then it plugs into the same sort of triaging process. Of course, we've been working with Luke for several years before he took the helm as well. Okay. Just in terms of that strategy, in terms of finding the differentiation or some white space on validated mechanisms, how do you think about that relative to time to market? You can make the argument that if there's another GLP with another one percentage point of weight loss, is there a reason to be launching that five years later? Similarly, with the checkpoint inhibitors or five or six and beyond the three or four big ones, there is not really much of a commercial impact. How do you assess those opportunities and are there areas where there is like, is there a matrix? Do you think about these are the key things that we need to hit which will offset the later market? Yeah. That's a good question. I guess, you know, that paradigm of best in class and first in class is starting to be, I think, challenged a little bit because you can be first in class and then you can rest assured that if it's a credible mechanism, someone's gonna be on your tail, particularly with China's involvement. Best in class is again sometimes a bit of a fallacy 'cause you're only best as a function of getting your indication correct, your population correct. When Luke talks about validated targets, that's not mutually exclusive from thinking about how to further differentiate by selecting the right population, thinking about responder phenotypes and ultimately resulting in a differentiated medicine profile. I think when you think about all the deals we've done, none of them are pure fast followers. We bag the validated biology so that we've discharged that risk, and then we're able to experiment, but with assurance around populations that will also result in a more compelling market proposition at launch. One of the recent deals that you have done is RAPT. And then maybe just against that criteria that you set out. Xolair obviously de-risked the IgE mechanism. Could you just outline what the residual unmet need is and how Arzerra is solving for that? RAPT is a particularly, I think, unique setting where you've got a mechanism that was validated for type I hypersensitivity reactions and T2 pathways, which obviously received a lot of attention in the last 10 years with many successful therapeutics like Nucala, like Dupixent, but was relatively overlooked by circumstance because Xolair was almost ahead of the curve in terms of that journey. Obviously, as you know, it was a consortium effort that led to the indication and marketing authorization for food allergy. You've got a validated pathway that had been overlooked. Despite all of that, the success of Xolair in its first year demonstrates the desperate need for patients and prescribers in the food allergy space against the backdrop where the standard of care is effectively, you know, what I would describe as primitive, which is avoid the allergen. Easier said than done with cross-contamination, et cetera, of food. Guidance that if you then have a life-threatening anaphylactic reaction, carry an EpiPen with you. The type of efficacy that we're seeing with IgE mAbs, I think is game changing. The problem is the market data suggests that even with that encouraging uptake, most of that use is through a two-weekly administration. That is problematic for patients, particularly when you think about pediatrics, adolescents, and ultimately a preventative therapeutic. It's also problematic with Xolair because I don't know if you've seen the USPI, but there's a nomogram. Again, you asked me for my clinical opinion earlier. I would not want to be having a, you know, a complex nomogram with an axis of baseline IgE levels and an axis of weights which are, you know, either end of those bookends excludes patients up to 120 kg but down to 50 kg, and then IgE levels as low as 500 kg. It's, you know, frankly a very attractive setting to be able to observe all of that and then come in with a three monthly therapeutic. Okay. Food allergy in itself is obviously quite broad. There's multiple allergens within that. Are there any that you're specifically focused on or is there any way that IgE is probably the most relevant pathway to be targeting? IgE is responsible for, I think approximately 95% of food allergies. What's important is that it's allergen agnostic. Say relative to things like oral immunotherapies, which are allergen specific, IgE takes that complexity out of the routine management of food allergy. The trial's been designed in that way where they mandate patients with two or more food allergens. Okay. I guess one concern that we hear from some physicians and other companies that are developing alternative food allergy products is that IgE antibodies are not disease modifying. There are some concerns about sort of long-term blocking of IgE, particularly in pediatrics. How do you think about those? I'll start off with the second one. Xolair has been marketed for asthma, including in pediatric populations, for I think 23 years. I've seen no data that is cause for concern with chronic treatments in pediatric populations. To your former point, disease-modifying is always an interesting term, right? It seems to mean different things to different people. I think in this space what's important is that you've got a efficacious therapeutic that relieves the burden and therefore results in strong adherence that allows the patients to, you know, live more normal lives, without the concern of a life-threatening anaphylactic reaction. I'm not really sure what disease-modifying would constitute unless you totally rewired your underlying immune system that currently I don't think is a tractable proposition. Okay. I guess it's sort of some of the immunotherapies can increase, sort of, or reduce the risk of like for example, we've seen with some therapies that you can go from having a third of a peanut to a full peanut. It's just sort of reset or rewiring the immune system as you talk to. Okay. I mean that's something that we can obviously consider as well in terms of reintroducing food, studied foods as part of LCI. As you know, the oral immunotherapies haven't been very effective and are starting to be withdrawn. Okay. Just in the interest of time, maybe we can move to the FGF21 space. You obviously did a deal there, I think it was last year. That obviously clearly plays through the strategy that you outlined, but there were multiple FGF21 deals done last year. I think Luke said on the full year results call, that he thinks that you've got the best deal and the best in class profile. Could you just discuss that product profile and why you think it is best in class? Yeah. Well, I think we can qualify this because we had pick of the litter obviously, because we were the first to make the deal. Again, given our broader portfolio, ultra-long-acting COPD, we've got a lot of experience in settings where there's significant burden of comorbidities in specialized internal medicine indications and where there's polypharmacy. We appreciate that you're not protected if you don't take the medicine because it has a high burden. Therefore the FGF21 proposition. I'll start off with the class because I think it's important to talk about the class. The class has demonstrated histopathological reversal of cirrhosis. That's something that was considered to be unachievable even five years ago. I think the FGF21 class is demonstrating a true step change in terms of what you can expect from efficacy in F2/ F3 MASH, but also cirrhotic MASH potentially in alcohol-related liver disease as well. Then we thought about the attributes of efimosfermin which were around immunogenicity, anti-drug antibodies. I think objectively it appears to have more favorable properties to prevent ADAs. Second was around scalability. Obviously, this is a prevalent indication and we're hoping that it's gonna be widely used once launched. We wanted to make sure that the ability to scale and the COGS proposition was attractive. We felt efimosfermin had better characteristics than the others. The former points around ADAs and also potentially what we've seen in terms of time of onset might indicate support there for efimosfermin may have a more rapid onset and a more durable response that might then also be best tolerated. Finally, the sort of the clear proposition is that a weekly therapeutic versus a monthly is a material differentiator in an area like MASH. MASH is the liver's manifestation of the metabolic syndrome. These patients often have dysglycemia, if not diabetes, hypertension, heart failure, chronic kidney disease. They're gonna be on multiple therapeutics. It's not trivial to have to inject yourself every two weeks for the rest of your life. I think that's a material differentiator as well. Okay. How do you think about the opportunity in the context of the GLP-1s? We spoke to a KOL at this event last year. He's actually coming in later as well. His view that the majority of patients will be on a GLP-1, and they had MASH, and both are of course weekly for now. How do you reconcile the fact that there is an advantage for the- Yeah. Less frequent administration if patients are maybe already used to being on a weekly injection? Yeah. Given my background that I started off with, you'll appreciate that I've been closely involved with insulin-based therapeutics for, you know, 15 years. They are effective at targeting MASH resolution. They're less effective at fibrosis improvements in F2/F3, and they are ineffective in cirrhotic MASH to the extent that the GLP-1s have shown an effect that was unfavorable versus placebo in cirrhotic MASH. When we think about that, and we did think about this very carefully when we did the deal, even if you assume the most optimistic scenario, which is unrealistic because of the tolerability and persistence on GLP-1s, which is less than 50%, as you know, in a year. For a chronic therapeutic, that's a problem. Even in the most extreme scenario where everyone's on a GLP-1, the efficacy of FGF21 is preserved, and we are permitting use of GLP-1 as background therapy in the trials. Very binary clear differentiation for cirrhotic MASH, differentiated based on best in disease fibrosis improvements in F2/F3 MASH, and efficacy is preserved on top of a GLP-1 as, and, you know, if you think about the sequence of treatment here, they have been prescribed a GLP-1 in primary care. By the time they get to specialized care and a hepatologist, they're gonna be looking at, you know, products like FGF21. Okay. I realize we've only got five minutes left, but I wanted to very quickly ask about respiratory as well. The pipeline's actually, and the portfolio that you have is very broad in terms of modalities, different frequencies of administration. Just thinking about COPD alone, you have two IL-5s. You have an IL-33 and also a TSLP as well as the, or the TSLP's been investigated by other companies, and then you have the PDE4. Could you just unpack all of that and help us think about how those all coexist in COPD. It ties in with your earlier comments around Luke's strategy for validated pathways. Superficially, you might, as you kind of framed it, you might think, well, you've an IL-5, an IL-33, you've got a TSLP, so has many others. The first thing is obviously the ultra-long-acting properties across that portfolio and we're gonna be looking forward to seeing the success of that with Exdensur in asthma in the coming months. For COPD, I think that ultra-long-acting characteristic is even more important for all the reasons I described for MASH, and obviously twice yearly in that instance. But what I'd emphasize is that GSK probably has more data in respiratory medicine and translational data for respiratory medicine than any other company or single academic organization in the world. What we have done is COPD is a heterogeneous disease. The bookends of COPD are wider than asthma and COPD, so that in other words, there's more overlap between cohorts of patients with asthma, COPD than there is within COPD. We recognize that. We've got data spanning inhaled therapeutics, first wave of monoclonal advanced therapies, now PDE4, IL-33, TSLP, novel undisclosed mechanisms like the Empirico oligonucleotide deal. We have a causal map of what's driving disease in this prevalent disease classification of COPD at the level of the underlying biology. We've mapped all of those mechanisms accordingly to where they're going to be most relevant to the underlying biology. Those pathways are not stacked up on top of each other. A really good example is IL-33. You know, we've been frankly, quietly happy to be in a little bit of a stealth stage with what is a best in class IL-33 whilst the competition have taken forward that mechanism into precedents of trial designs and populations that really don't speak to the underlying biology of IL-33. I'm gonna be careful around disclosing too much around IL-33 and what we consider to be the target traits pairing. It's a good example of where, you know, we're mapping that to slightly different treatable traits in COPD. Importantly, COPD affects 300-400 million people globally. It's the third leading cause of death. You're able to segment the disease classification, but each one of those mechanisms still map to a scalable area of the market. Okay. Just in the last couple of minutes, also wanted to touch on AI in drug discovery. It's something that we get more and more questions on from investors. Given that your former role was a CMO at an AI drug development company, I just wanted to get your perspectives on the advancements that we've seen in the industry and then how those are being utilized at GSK. Perfect segue from what the conversation we just had actually. You know, AI, I think, has achieved maturity, certainly at GSK. I expect in many other companies around medicine design. Structure-based antibody design. We use AI heavily for oligonucleotide sequencing to be able to exploit the potential timeframe to go from target committed to candidate ready within a timeframe of under a year. But the next frontier for AI, in my view, is uncontroversially biology. Biology as it relates to drug development is the complex proposition of the intersection between underlying disease biology and what your mechanism is able to do. You then have to obviously reason over that data and decide and triage between an infinite number of targets. The targets that are relatively commoditized. As you know, there's a huge clustering of many companies around a relatively narrow target space at the moment. We've seen multiple instances where old targets have had a renaissance because someone has cracked the target trait pairing. When you take that into clinical development, you're then taking forward what is undoubtedly a multivariate proposition or frankly a mega variant proposition, but you're using univariate decision criteria as to whether to advance. Against that backdrop, AI allows us to reason over multimodal data. At GSK we've been very intentional in building field-leading human datasets with proteomics, genetics, transcriptomics, now spatial transcriptomics as well. We're able to pair that underlying cell phenotype to tissue and structural changes based on explanted tissue, but also complex cross-sectional imaging. We then link that to the types of endpoints that are important to how patients feel, function and survive. We're then able to reason over those datasets in a way that is not possible manually, but even with sort of conventional Bayesian biostatistical approaches. We've now got the early proof points of what happens when you're able to reason across those latent edges of different modes of data that historically you would have never been able to. For example, consider a structural change of small airway remodeling with patients that have been treated with IL-5 and how that relates to parenchymal remodeling as relevant to idiopathic pulmonary fibrosis. One of the outputs of that type of AI reasoning over multimodal datasets resulted in our conviction around IL-33 TSLP combination therapeutic approaches. That was an instance where we were able to use genetic instrumentation using combinatorial variance approaches and a novel methodology, but then linking that to all of those modes of data from cell to tissue to structural architecture through to clinical consequence. Perfect. Thank you very much for your time, Kaivan. No problem. It was a great chat. Thank you. Thank you.