Good morning, everyone. Welcome back to the fourth and final day of our 41st annual healthcare conference. Very excited to be back in person. Hope everyone has safe travels home. My name is Lisa Cohen. I'm an associate with our Investment Banking group out here in San Francisco. Very excited to be introducing our next company, Gritstone bio. As a reminder, we will have Q&A following the presentation, so we'll have a mic runner around to the extent any questions. With that, very excited to introduce Andrew Allen, Co-founder, President, and CEO. Great. Thank you, Lisa. Good morning, everybody. Pleasure to see you in person. I'm Andrew Allen, CEO, one of the founders of Gritstone, and it's a pleasure to give you a quick trot through the pipeline and progress we've made at Gritstone. I will be making some forward-looking statements today and refer you to the disclaimer appropriately. We all know that the immune system is fundamental to many diseases, and it's pathologically implicated for an increasing number of important diseases. Unsurprisingly, therefore, immunotherapy and modification of immune responses has become really central to many critical, huge product classes. We've listed three here. Cancer immunotherapy, which speaks for itself, infectious disease vaccines, and autoimmune treatments. Probably could add neuroscience and neurodegenerative disease to this list as well now. Gritstone obviously has been participating in this burgeoning field since its formation in 2015. We focus on vaccines, we focused initially on cancer vaccines because a catalytic event happened in the end of 2014, which was the recognition that neoantigens were likely to be the key targets for T- cells that were killing tumors in patients responding to checkpoint inhibitor therapy. Neoantigens are those mutant proteins derived from DNA mutations that change a protein sequence. The elegant bit of immunology behind that that made all the sense in the world is that your immune system is designed to recognize foreignness on the assumption that something that is foreign is probably a pathogen, typically a virus. Neoantigens are foreign. They were not present in you and in your immune system and in your thymus when tolerance was being formed as a fetus and as a neonate. For the first time, we suddenly had a whole new class of targets that seemed to make total sense to the sort of basic tenets of immunology. It is a foreign protein, absent during tolerance formation, that arises specifically in tumors as a consequence of DNA mutation and can provide really good targets for T- cells such that you can eliminate tumors through the T- cell recognition of such targets. That was the catalytic event that led to the formation of Gritstone. The tools that we'd been using for vaccinology up to that point were unfit for purpose. Namely, how do I find which mutations create neoantigens? Because it's only a small fraction. Secondly, how do I drive really strong CD8 T- cell responses against those neoantigens? Most vaccines are not very good at priming CD8 responses. In a cold tumor, operationally, you have to assume that that patient is naive to their neoantigens. The tumor has successfully hidden from the immune response. You have to prime CD8s, which is a tall order. We built the company to solve those two key challenges, and so new systems were required. That's the innovation that we've been participating in and driving and leading really at Gritstone. Let me start at the top. Which mutations create neoantigens? Really challenging problem. We solved it by collecting large numbers of human tumor samples using a elegant mass spectrometry technique to identify the peptides displayed on the surface of those tumor cells, and then sequencing the tumors, DNA and RNA, and use natural language machine learning mathematics to build a model that correlates sequence data with surface peptide display on HLA molecules on target cells. We can do all of that prediction work just using sequence data. Today, all we need is a Formalin-fixed, paraffin-embedded specimen or a virus sequence, and in silico we can predict where the good T- cell antigens will lie within that sequence. That's called the EDGE Model. We published it in Nature Biotechnology in 2018. We've patented it. Those patents are now issuing in the U.S., and we've partnered it. It was the core of a bluebird bio collaboration we announced a few years ago. Very important piece of the puzzle. Next, having found the targets, how do I prime CD8s? We had no vaccine vector system in-house when we began the company. We had the luxury of looking around and saying, "Which vector system drives the best CD8s?" The answer was clear: adenovirus. We built an adenovirus platform, and that will drive very good CD8 responses, but they'll go up, and then they'll come down again. For cancer, you probably want them to go up and stay up for a while. You need to boost. You can't just give a second dose of adeno quickly because the adenovirus will elicit strong antibodies to the surface of the virus. If you give a second dose too soon, it's immediately neutralized by those antibodies. The field, recognizing this, had adopted the phenomenon of heterologous prime boosting, meaning you prime with an adenovirus and your antigens of interest. You boost with a different vector. Same antigens, different vector. We looked at options for boosting vectors, and we settled on this new form of RNA called self-amplifying mRNA. It's similar to mRNA, which of course you're all familiar with, but in the vaccine, we encode a copying enzyme, a polymerase, that makes copies of the RNA. That means the antigen is around for longer in the lymphoid system. That's probably important. You get double-stranded RNA intermediates that are very strong stimulants of the innate sensors, and that seems to drive a qualitatively superior immune response compared to just mRNA. For that reason, we took this into humans. We were the first to put SAM, self-amplifying mRNA, into humans, and it does indeed seem to elicit very nice differentiated immune responses. This is our platform, adenovirus prime, SAM boost, a heterologous prime boost system. Next, making individualized patient vaccines is a tricky thing to do. When we started, we tried doing it through outsourcing. It was impossible. Well, it wasn't impossible. It could be done. Highly fragile system, lots of shipping things around, very expensive. Early on, we recognized we needed to do this ourselves. We built our own biomanufacturing facility, 43,000 sq ft. It's around 70 or 80 people. It's in Pleasanton, which is a town about 30 miles east from here, where real estate's a little cheaper than on the peninsula. This has been critical for our success because we can now move it incredibly quickly and iterate with our products easily because it's all under our control. Partners, most notably Gilead, obviously this has been a key element of why they were interested in partnering with us, because we could make things for them. With Gilead, for example, we signed a collaboration on HIV cure with them in January of 2021. The IND was cleared before the end of the year. Inside 12 months, to go from literally receiving some sequence, making, releasing, filing, and having an IND approval is the kind of pace that you can do when you're in control of your own biomanufacturing. For individualized products, being able to do this and scale this out, automate the process is important, of course, because we need to start delivering or thinking about delivering hundreds and then thousands of personalized products per month. This is easier than cell therapy, right? No live cells or anything. Nonetheless, there is some obviously technical challenges to making personalized products at scale. By owning it, you can really innovate in that manufacturing platform. We develop different kinds of vaccines from the traditional ones. Immunogen design. Immunogen is the bit of the vaccine that contains the antigens. It's really important. I won't spend time on that. Then we put it into humans, and we learn a ton from humans. Translational data that we then bring back in, and we use it to iterate on the design of the vaccines. For those of you who have an appetite for some deep science, we had a very nice manuscript in Nature Medicine in August of 2022 describing our personalized cancer vaccine and the immunological data that we observed and how that correlates with clinical efficacy that we clearly observed with this personalized product, most notably in colorectal cancer patients. Okay. Coming out of all of that capability building, we have a pipeline that looks like this. The individualized, patient-specific personalized cancer vaccine is called GRANITE. Currently in a randomized phase II/III study, we'll have preliminary data from the phase II component in the fourth quarter of this year. Very big event for the field. This is the first time a personalized cancer vaccine has gone into a cold tumor. Moderna, of course, had recent positive data. That was in melanoma in the adjuvant setting. We're doing something a little different. Obviously, if we can unlock cold tumors, that is transformational for patients because patients, so many patients really want the long-term benefit that immunotherapy can offer, but they're currently it's inaccessible or unavailable to them because most checkpoints just don't do anything in cold tumors. Unlocking cold tumors would be a huge advance for the field, so we're excited about that prospect. SLATE is our off-the-shelf product that uses shared neoantigens shared across patients, so you don't have to make a personalized program. We've been focusing on that using KRAS as the key target, we've iterated again, I'll describe this a little bit more. This year, we intend to launch a randomized study with a third form of SLATE focused on KRAS, but not limited to KRAS, and that'll be in a frontline lung cancer setting. A randomized trial, frontline lung cancer, aiming to launch later this year. Again, that'll be a randomized study. We're now into the phase where randomized studies are necessary to really deliver definitive proof of concept. We have another SLATE program for a different tumor type that is heading towards the clinic. No specifics have yet been disclosed. On the infectious disease side, we have two therapeutic programs. One is the HIV cure collaboration with Gilead, currently in phase I. If that moves into phase II, there's a $40 million milestone that may come later this year. We also have a collaboration with the Gates Foundation around HPV eradication. Really exciting concept for women who are too old to have received Gardasil, because, of course, Gardasil is a fairly recent product, but nonetheless, are chronically infected with HPV, at risk of developing HPV-associated malignancy. Can we help those women eliminate HPV? Many women, most women, naturally are able to eliminate HPV, but not all. Can we help those who don't, and can we push them over the edge? Really interesting prospect, very exciting program, currently preclinical. On the IG prophylaxis side, we've been doing a lot of work with SARS-CoV-2 to exemplify the performance of self-amplifying mRNA. Our biggest study is in South Africa, and we'll have some data from that in second quarter of this year. In research, we've been working on pan-coronavirus with some very interesting focus on T-cells and those T-cell conserved epitopes in coronaviruses. We have an influenza program, which of course is of interest to a lot of folks. Okay, let's drill down a little deeper in oncology. I talked about this. This is a big year for personalized cancer vaccines. We've just had top-line data from Moderna in adjuvant melanoma. We're expecting data this half from Roche BioNTech in the context of frontline metastatic melanoma. Obviously, the Moderna data are very exciting. We can perhaps talk about that in Q&A. We're trying to do something a little harder, but obviously, the Moderna data is extremely encouraging. We are the three big players currently inside 12 months from randomized trial data. We do have some fundamental differences from the other players. The most important one probably at this point is the vector. mRNA is the vector that they've chosen because they are mRNA companies. We've chosen a vector designed to drive CD8s, and we've chosen, as I mentioned, chimpanzee adenovirus and self-amplifying mRNA. Here's how it actually works. The thesis is, and we articulated this back in 2015, and it has not changed, which is parenthetically, I think, always the sign of a strong thesis, that it really doesn't change over seven years with a lot of data being generated. What we believe is that there are many epithelial solid tumors. Most common solid tumors are cold. They have low neoantigen expression, low PD-L1, and there are no infiltrating T-cells. As far as we can tell, those patients' immune systems are immunologically unaware that the tumor exists. This is not a coincidence. This is because, quote, "Successful tumors are very good at hiding from the immune system. They use many of the same tricks as viruses that are able to enter chronic latent states of infection." Cold tumors are there. If you treat such a patient with a checkpoint inhibitor, nothing much happens because there's no substrate. Our thesis was we need to identify their neoantigens, vaccinate them against their neoantigens in a non-immunosuppressive environment. The tumor is highly immunosuppressive. Of course, if we give an intramuscular injection in the arm, the axillary lymph node is the factory of T-cells. It is not an immunosuppressive environment. That's the logic. Those T-cells can then traffic to tumors, meet antigen, and proliferate, and we've shown that happens. Of course, hopefully, we'll see cell kill. We use the heterologous prime boost, as the schema shows at the bottom, the idea is to drive the T-cell response north and keep it sustained. Specifically, the way this works is that we sequence tumors at baseline. I should have mentioned, most neoantigens arise somewhat randomly from mutations, most mutations are so-called passengers. They're not drivers, they're unique to that patient. If the goal is to deliver as many neoantigens as possible, you have to develop a personalized product, okay? If that's the goal. We sequence tumor, identify all the mutations, plug them into our EDGE prediction model, take the top 20 predicted candidate neoantigens, make the vaccine. That's two vaccines, same set of antigens, but adenovirus and self-amplifying mRNA, deliver that, a simple IM injection just like your COVID shots, together with PD-L1 antibody. We're doing this now in a randomized trial in colorectal cancer, where we saw strong signal of activity in our phase I/II studies. That signal was in third line. Now we're in front line, where, of course, we expect to see stronger immune responses. The phase I/II data, don't have time to go to in detail, but if you do care, as I say, Nature Medicine has the details. We observed that the vaccines are very well tolerated. We observed molecular responses, meaning ctDNA responses often match with CEA responses in about half of subjects. In those halves, we saw extended progression-free survival. At this point, we're seeing extended overall survival. If you look at the molecular non-responders, their median survival was 7.8 months, which is basically what you expect in third line colorectal cancer. If you look at the recent pivotal trials of Tegafur, regorafenib in this setting, the median survival for the control and the test arms is usually in the five to seven months zone. Our non-responders unfortunately died at the expected rate. The molecular responders, however, half of them have not yet achieved median survival, and it is gonna be over 22 months. Very strong signal correlated with all those other markers of response. This is a highly coordinated set of efficacy signals in this cold tumor population. Based on this, we've now moved into the randomized phase II/III study in frontline microsatellite stable metastatic colorectal cancer. We have Fast Track designation in that context as well. I'll show you a little more detail in a second. We use ctDNA now as a key biomarker. Radiology is unreliable. We've seen this now with multiple forms of novel immunotherapy, most notably Immunocore, for those of you familiar with those data, where radiology really just completely missed any benefit with their, essentially their bispecific therapeutic. Overall survival was massively impacted by their therapeutic with a hazard ratio of 0.51 in their randomized phase III trial. ctDNA change correlates very nicely with overall survival. We're seeing similar phenomena as are others in the field. ctDNA is an important biomarker that seems to be much more reliable, certainly early on, than radiology. The logic for that, of course, is simple. Unlike cytotoxics and targeted therapeutics, which just aim to kill tumor cells, we are trying to drive T-cells into lesions, and if we're successful, they will proliferate. The notion that lesion shrinkage is the necessary and appropriate biomarker for early response is absurd. We're trying to cause cells to proliferate. Therefore, you cannot rely on radiology. Now, we often do see lesions shrinking slowly over time, but it doesn't happen quickly. Molecular response seems to be a much more useful proximal biomarker. We're not the only ones, of course. Many other companies interested in this space, most notably Friends of Cancer Research are pushing really hard on this because they recognize that radiology is letting us down. The FDA, of course, has now issued draft guidance on the use of ctDNA in the adjuvant setting and is clearly getting interested in its use as a biomarker. No one's yet had a drug approved for Accelerated Approval with this. I firmly believe that day will come. It's just a question of when. What we've observed in our cold tumor patients, and here are the patients from our phase I/II study, is, as I said, at baseline, you find no neoantigen-specific T-cells. This is a standard T-cell assay, very routine, looking for neoantigen-specific T-cells. We can't find them at baseline. After vaccination, we find very good numbers, as you see on the right. As I mentioned, those T-cells infiltrate tumors and expand. That's in the manuscript, and that's associated now with molecular response in half of subjects, and this very significant extended overall survival in those molecular responders. It's not a randomized study, of course. That's the trial we're doing now. This is the kind of signal you're looking for in uncontrolled phase I/II studies. This is a strong signal, even in third line colorectal. On the right is one of the phase III studies, and it makes the point that typical median survival is around six months. You can see what we've observed is dramatically different. These patients were very ordinary at baseline, and you cannot tell the responders and non-responders apart from baseline characteristics. PD-L1, tumor mutational burden, interferon gamma signature, presence of liver metastases. It's the same in the responders and non-responders. There's no cherry-picking going on here. This is biology in action. This is the trial we've designed, taking patients with newly diagnosed metastatic colorectal cancer. It's a phase II/III. It is not seamless. These are discrete trials. The phase II is underway. We anticipate preliminary data at the end of this year in Q4, and this is up to 80 subjects with molecular responses, the primary efficacy endpoint. The way chemotherapy is used in metastatic colorectal cancer is to do induction chemo, typically with FOLFOX Bev. Then when oxaliplatin starts to induce neuropathy, you stop that and flip to maintenance with FOLFIRI until progression. That's the standard. That's what we're using as the control. Some young subjects get irinotecan added, so they get FOLFOXIRI. That's permitted in our trial, and we stratify based on that. Then maintenance, subjects will receive either just standard of care FOLFIRI or FOLFIRI, plus vaccine, plus atezolizumab. Roche did a big study, a phase III study of atezolizumab in this context, definitively showing that it had no impact on PFS or OS, which is why there isn't a third arm here of chemotherapy plus atezolizumab. That study's been done, it's negative, and therefore, any benefit we observe will not be attributable to atezolizumab. This is vaccine. Again, as I mentioned, if we unlock the cold tumors, of which this is one of the paragon examples, the opportunities for this platform are just legion. Obviously, this is the excitement for us, is to really try and help patients who today get no benefit from immunotherapy. The MSI is that lucky 5% who obviously get really good benefit from immunotherapy. MSS, nothing happens. Very exciting prospect. Okay, let's flip to SLATE. This is our off-the-shelf product. We utilize shared neoantigens, most notably, KRAS mutations. It's the same heterologous prime boost platform, and it can work in any tumor where there's a KRAS mutation, which is basically lung, colon, and pancreas. Those are the common ones. What we've shown in our phase I/II study in advanced disease is that, again, in primarily in non-small cell, few colorectal patients, but the majority were non-small cell, we see molecular responses in about 40%, and these are all patients who've progressed on prior IO or IO chemo. Exactly as we saw with GRANITE, if you had a molecular response, overall survival was roughly doubled here compared to the non-responders. It went from 4.5- 9.6 months. Again, not randomized, that signal is just strikingly concordant, exactly as we saw with GRANITE. Remember, this is a different target antigen, different disease, but the biology is the same and the signals, the efficacy signals are identical. This gives me confidence that this is a real signal. Molecular response associated with prolonged PFS and prolonged OS. With SLATE, we've learned a lot. We began with lots of antigens in the vaccine, and we learned that there's actually a dominance cascade, and KRAS is not a very strong antigen, so we iterated and focused in exclusively on KRAS. The challenge there is that you're focusing the immune response just on one neoantigen, and tumors can escape selection pressure. The third version we're going to be putting into the clinic later this year will include additional tumor-specific antigens which are shared, such that each patient will be receiving multiple antigens. We believe that's gonna drive even greater efficacy and start to make SLATE look a little bit more like GRANITE, where you're delivering multiple neoantigens. Remember, when we treat viruses, for example, we treat with combination drugs, because if you just treat HIV with one drug, the virus can mutate around it. We generally give cocktails of three or four. I think the same principle applies here. Targeted therapy on single target is likely to lead to acquired resistance. That's been described even with cell therapy, right? The KRAS cell therapy of Steven Rosenberg had some acquired resistance because of losing the HLA allele that presented that KRAS mutation. The obvious answer is to deliver multiple antigens. You gotta find multiple shared antigens. That's what we've learned to do now. We've got a very active tumor antigen discovery platform that's obviously filling in and allowing us to think about designing multiple assets to deliver multiple antigens to patients with different tumor types. Okay. Let's move now to infectious disease. The Gilead collaboration was announced, the beginning of 2021. That study, as I said, finished the IND enabling studies through by the end of 2021. It entered phase I at the beginning of 2022. That study's ongoing. We received $60 million at the beginning, and there's a $40 million option payment if they elect to move to phase II, which may occur later this year. We don't talk about this. This is a Gilead program, so they speak to the progress. Let's turn to our work. As I mentioned, self-amplifying RNA has some attractive properties that likely differentiate it from mRNA. Number one, because antigen persists in the immune system, we seem to get very durable neutralizing antibodies. That's potentially a game changer because the big challenge with most of the first generation COVID products is that although good antibody levels are achieved initially, they fade quite quickly, necessitating multiple boosts. Nobody seems to want that. If you can drive antibodies that are sustained, that obviously changes the calculus, and that seems to be happening with self-amplifying RNA. We think it's because of antigen persistence in the lymph node. With mRNA, the mRNAs comes and is gone. The antigen comes and is gone. The immune system never enters the phase where it drives durability of response. More durable antigen seems to drive more durable immune responses. It makes sense. There's experimental data to back this up. Secondly, the breadth of the T-cell response, including T-cell antigens, which are invariant, means you may have variant-proof immunity. It's a complex field. No products do this currently. There's some groundbreaking work to be done here, including with regulators. This is the prospect. Those of you who read, I think it was Scrip yesterday, Peter Marks was speaking about this yet again yesterday, encouraging companies to think about invariant epitopes driving T-cell immunity in their vaccines. This is obviously our bread and butter. Finally, low doses of SAM seem to drive very good immune responses. That matters if you're starting to think about multi-pathogen vaccines. There's a lot of energy right now about the idea of delivering a single vaccine to older folks, protecting them against SARS-CoV-2, influenza, and RSV. One vaccine. How can I do that? If I have to give large doses of RNA for each of those to drive good antibody responses, I've got a problem, because I can't just give 3x the dose, right? The maximum dose is capped by tolerability. With self-amplifying mRNA, we see good efficacy, even at low as three micrograms, and we have good tolerability up to 30 micrograms. We have room to start multiplexing products, and that obviously may become very important as this notion of multi-pathogen vaccines gains currency. This is our SARS-CoV-2 platform. I won't go through this in detail. The big study is the middle one, backed by CEPI, over 300 subjects in South Africa, and we anticipate robust data from that study, including 6-month antibody data, at the ECCMID meeting happening in April this year. It unfortunately conflicts with AACR, you're gonna have to pick your, I would say poison, but that's probably not the right noun. Pick your delight in April as to which you wish to see. The NIH has been running a study that's entirely under their control. I think that's gonna be reporting out this year, but obviously we don't have control over that. These are briefly some of the data we've shown. As I mentioned, low doses are important. Here on the left are our data as a heterologous boost after AstraZeneca prime. You can see the 10 micrograms of self-amplifying mRNA elicits a nice boosted neutralizing antibody response. On the right are the data with both Moderna and BioNTech delivered in exactly the same context by a cooperative group in the U.K. You can see that at 10 micrograms, we're essentially delivering the same antibody titers. It's not the same study, we have the usual caveats. These are standardized assays. 10 micrograms seems to be delivering the same boost as 100 micrograms of first gen mRNA. This is the durability issue. On the left is, you know, you could pick pretty much any mRNA vaccine study, and you'll see a nice peak, but then it fades. On the right is our early data suggesting that at least in this boost setting on the right, we get a nice peak, and then it's flat. That, if sustained in a broader N, is obviously a really important observation. We see these nice broad T-cell responses emerging. On the left, you can see a baseline. We've got no responses to these non-spike epitopes from SARS-CoV-2. After vaccination, we get induction of CD8 responses against a broad set of epitopes. It's that breadth of immunity against invariant epitopes that potentially is important for durable variant-proof immunity. In conclusion, SAM seems to be filling some of these key gaps we're seeing now with the first generation products: durable immune responses, induction of both NaBs, neutralizing antibodies, and CD8 T-cells, and an ability to give low doses and start multiplexing. In conclusion, we're a company that's been building a differentiated set of capabilities that enable us to generate next generation vaccines to solve some of the big problems: cancer and novel infectious diseases. We have our own biomanufacturing, we've got data coming this year, most importantly, perhaps from the individualized cancer vaccine program, GRANITE, but some important SAM RNA data coming in the second quarter, the launch of a SLATE trial, an off-the-shelf randomized cancer vaccine trial later this year. With that, we'll flip to Q&A, I think. Thank you very much for your attention. Thanks very much, Andrew. We'll also have Celia Economides up here with us. She's the CFO. We will take any questions from the audience. Andrew, you touched on it briefly, but would love to just better understand the difference between Gritstone and Moderna. Yeah. This week we've had a lot of investors obviously, excited by the Moderna data and asking us about how we stack up. It's probably worth spending some time on that. If we step back, again, we designed our platform to solve the problem of neoantigen vaccine immunotherapy. Moderna is an mRNA company that set about trying to deal with cancer, neoantigen immunotherapy as well, but it was not designed for that purpose. One of the key differences I think really does relate to CD8 T-cell induction. Why do we keep talking about CD8s? Well, solid tumors express Class I HLA molecules, and I don't wanna get too deep into immunology, but T-cells do not recognize whole proteins on the surface of targets. It's very different from antibodies. T-cells recognize short peptide fragments displayed by HLA molecules, which are these highly variable sort of platform molecules that are different between all of us. The reason they're different is so that there's no super virus that can escape immune control and wipe us all out. That's the general belief. Polymorphism of HLA is helpful to us as a species to survive. Predicting which epitopes are there is a hard problem. It's how our immune system has evolved to deal with viruses. CD8 killer T- cells come in, recognize short viral fragments displayed on the surface of target cells, typically virally infected cells. Now, the Achilles heel of tumors is, as I mentioned, that they create these mutations that enables the function of the cell to change. That's why it behaves like a cancer cell. The vulnerability is I've now created foreignness, and so our immune system can tackle that. Again, these are not typically cell surface proteins. Occasionally they are, but mostly these are mutations in intracellular proteins that really can only be seen by cytotoxic T- cells. So you need a T- cell that can recognize the peptide on the surface of the abnormal tumor cell. Those tumor cells generally only display Class I HLA molecules. Class I is recognized by CD8. Okay? So that ultimately, if you believe in direct tumor kill of target cells, it has to be a CD8. I'm not saying that CD4s are unimportant. They are critical for generating an immune response. They can release cytokines. There may be some notion of bystander kill, but the direct killing is a CD8 T- cell. So you need CD8s. Secondly, if you look at tumors that you're about to treat with a checkpoint inhibitor, the strongest predictor of efficacy is baseline tumor-infiltrating CD8 T- cells. It's not CD4s, it's not NK, it's CD8s. Again, the light is shining on CD8s as the key effector cell, the final effector. Thirdly, when you're trying to predict which mutations can function as antigens, for technical reasons I won't go into, it's a lot easier to predict for Class I. The whole field has been really focused on Class I prediction. Again, Class I is recognized by CD8s. You see, everything points you to wanting CD8s. That's why when we began the company, we said, "Right, which vaccine vector system drives the best CD8s?" That's why we chose the adenovirus. Moderna did not. They were an mRNA company, of course, they were using their vaccine. The vaccine has many wonderful properties, as we all know. However, it is not as good at inducing CD8s. You can see this in the SARS-CoV-2 literature, that you have to use really sensitive assays to find any CD8 priming with mRNA vaccines. Even when you use very sensitive assays, the cell numbers are very small. That is not the case with adenovirus. As you saw, we were using bulk standard ELISpot assays and seeing nice responses measured in blood. The power of that CD8 induction is materially different between the two platforms by design. You can see that manifest, I think, in clinical data. If you look at MSS colorectal cancer, as I said, we treated, you know, 13 subjects in that data set I showed you. Half of This is all in third line. Half of them had molecular responses, extended overall survival of things I showed you. Moderna did a similar study, presented at SITC in 2020. 17 subjects treated with their personalized cancer vaccine plus pembrolizumab. 16 of the 17 patients progressed at or before the first scan at six weeks. There was just no signal. If you just look at the two platforms, in the basic science, there's differentiation. If you look at the clinical data, we saw a very clear signal in cold tumors that was not visible in that study for Moderna. Now smartly, Moderna said, "Okay, let's ease back on the whole cold tumor thing. Let's focus on hot tumors." Hot tumors are fundamentally different because you've got preexisting infiltrating CD8 T- cells. Immune sensitization has happened. Neoantigen priming has happened. What you're trying to do with the vaccine probably is boost. Of course, you can't clearly differentiate 'cause it's a complicated, you know, multiplex system. I think mostly what the mRNA vaccine is doing is boosting a preexisting CD8 population, and that could have real utility. Moderna asked the question, in a hot tumor, melanoma, can we augment the efficacy of pembrolizumab? Which of course is pretty effective, but can we make it even better with a personalized cancer vaccine? To further make it, you know, to bias in favor of products, let's go to the earliest possible setting, which is the adjuvant setting. They did everything right. It was a smartly designed study. Happily, they saw a positive result, right? Hazard ratio 0.56 is what we've been told. That's important because it tells you for the first time, A, a personalized cancer vaccine is having therapeutic benefit in a solid tumor. That's groundbreaking. Kudos to them. Secondly, it is augmenting the impact of a checkpoint inhibitor. Statistically, that's not a trivial thing to do 'cause checkpoints are pretty good there. That's what they've shown. Very exciting. We're doing something a little different. Of course, what we're doing is related. It's a personalized cancer vaccine. We're combining with checkpoints. There's a lot of commonality, but we're in a cold tumor. You have to believe that we can prime CD8s in order for us to actually see efficacy in this context. I hopefully have shown you the data, and again, it's in the Nature Medicine manuscript in great detail. We do this. We prime CD8s, they enter tumors, they proliferate, and they lead to cell kill and apparent clinical benefit in half of subjects. We've checked that box that we can do something in a cold tumor. Now we're testing that prospectively in a randomized phase II study, data coming fairly soon. That's how I think we compare ourselves with Moderna. Then, of course, the other key player is BioNTech-Roche, and they've been doing something similar to Moderna, I think for similar reasons. They've been doing this in the frontline metastatic melanoma setting. For them it's I think very similar to the adjuvant, but the amount of tumor is different. In the adjuvant setting, obviously if you're having benefit, there is clearly microscopic tumor present, but it's just not visible on imaging or on the skin. Essentially that defines the patient who is going to recur. They have tumor, but it's micrometastatic, as opposed to a macrometastatic patient who is by definition a frontline metastatic patient. I think, you know, qualitatively, they're quite similar, but quantitatively, there's just more tumor in that visibly metastatic patient versus the micrometastatic. I think the biology is going to be broadly similar for BioNTech, Roche. It's a different endpoint, of course. If you're in metastatic disease, you're using progression-free survival as an endpoint. If you're in the adjuvant, you're using event-free or recurrence-free survival as the endpoint. Those are slightly different. That difference may matter because, as I've said before, radiology can be a little bit confusing with immunotherapy. In melanoma, there's this phenomenon of pseudoprogression that is described with checkpoints alone, and maybe you make that worse if you add a vaccine. Obviously, Roche BioNTech are smart people, they know this, and I'm sure they're thinking about that and there are various ways to handle that. But I think their data are going to be super interesting, but it's still all in melanoma. If you show activity there, you're still limited to hot tumors. Of course, that's where the checkpoints can work and, you know, combos of TIGIT and LAG-3 and so on. It's a pretty crowded space. Cold tumors, it's wide open. There's limited competition, and that's why it's such a exciting opportunity. It's not trivial, obviously, but for the reasons I've explained, we think we have a really good chance of showing something very meaningful in that context. We're similar, but we're importantly different, I think is the short way of paraphrasing on the last 10 minutes. Okay, thank you for the question. Thanks, Andrew. Any follow-ups? Yeah. Andrew, for your cold tumor cancer vaccine, you're using SAM or you're not using SAM? Yeah, we use both. It's a heterologous prime boost. What's the difference? Is SAM better or? We've never done a head-to-head, right? Using the adenovirus to prime will give you a stronger CD8 response. Using the self-amplifying mRNA to boost will sustain that CD8 response. We've shown that. We actually then give a second dose of ChAd, of the adenovirus, at month five. That second dose seems to be really powerful. When we measure just T- cells, and no one really knows if CD8s in the blood are a really good proxy, but the numbers take off after that second dose. You end up with something like 5%-10% of the peripheral blood CD8 T- cells are specific for neoantigens. That is huge. When you fight off a flu infection, you don't have that many flu-specific CD8s in your blood. This is a really big number. In terms of absolute cell numbers, it's sort of comparable to CAR T, right? If you look at numbers of CAR-positive T- cells in a lymphodepleted patient being treated with, yes, CAR T, for example, and just try and figure out absolute numbers of cells, it's kinda similar to the numbers that we're generating. That comparison may be completely pointless and meaningless, so I don't wanna overplay that. You know, CAR and NHL is definitely not the same thing as vaccine T- cells and solid tumors. The numbers are impressive, right? Maybe that matters. I don't know. We'll find out, of course. Thank you, Andrew. We are up on time, but I'd like to say, we really appreciate you sharing your story and taking the time today. Great. Thanks so much. Thank you, everybody.