Hello everyone, my name is Poorna Kannan and I'm part of the Biotech team here at Needham. It's my pleasure today to be hosting Dr. Andrew Allen from Gritstone bio. Gritstone bio is a clinical-stage biotech company that is developing next-generation vaccines for cancers and infectious diseases. So without further delay, I'd like to turn it over to Dr. Andrew Allen. Just as a reminder, please send in any questions that you may have for the team in the dashboard. Thank you. Great, thank you Poorna, thanks for inviting us to speak today. So it's my pleasure to share with you the preliminary data that we recently revealed from our phase II study of GRANITE, a neoantigen-directed personalized cancer vaccine, which we're exploring in a phase II/III study that has registrational intent in patients with newly diagnosed metastatic colorectal cancer. And this is of the microsatellite stable type, the MSS type, which typically does not respond to checkpoint inhibitor therapy. That's been the conclusion for many different trials. So what we're trying to do is hard, but I think I'm going to show data with you today that show the exciting progress that we've made as the field really is advancing into new areas using neoantigens as the key targets for vaccines. We just completed the financing around these data, which closed last week of $32.5 million, which obviously is all helpful. So without further ado, let's get going. So I'll be making some forward-looking statements today, and you should obviously refer to our recent 10-K. Before I move into the data, a quick recap on Gritstone. We are an end-to-end cancer vaccine company that also has an infectious disease business, and we contain all of the elements necessary to develop and deliver these innovative products. It starts with identifying the targets for T cells in particular, and we use our EDGE artificial intelligence platform to figure out which targets we should capture in our vaccines. We deliver those targets or antigens within next-generation vectors, including an adenovirus and self-amplifying mRNA. We have in-house GMP manufacturing. We do a lot of work on the design of the vaccines and particularly how you present multiple different antigens within a vaccine. That's referred to as the immunogen design phase. Importantly, you learn a lot from humans that you then capture and use to your advantage in the next clinical trial and the next product. So there are iterative design manufacture test cycles that one can do with products such as these vaccines, which is rather unique, in the field of drug discovery and development where you are usually locked in on whatever your drug is. It cannot change. It is immutable. That is not true with vaccines, which we can iterate upon. That's a very powerful tool. Our pipeline covers oncology, where we have the personalized product GRANITE, which we'll be talking about today. We also have an off-the-shelf or shared platform centered on KRAS, which I won't be discussing today, but is very interesting and is entering a phase I combination study at Dr. Steven Rosenberg's site in combination with his KRAS mutant-directed cell therapy. And we also have infectious disease businesses, which I won't be discussing today. This is a high-level summary of the data I'm about to share with you. So from this study, which is now 67, 68 patients who've actually received product in a randomized form, we're seeing a clear signal in progression-free survival. The data are currently immature. There's over 60% censoring, but we're seeing an early trend in favor of product, with a hazard ratio of 0.82 and an exploratory analysis that I'll share with you that looks at the high-risk population where the data are much more mature. And there, I'm very happy to tell you the hazard ratio appears to be stronger at 0.52. And so obviously we're waiting with great enthusiasm for the mature PFS data, which we'll have, in the third quarter of this year. So, you know, next quarter we did use ctDNA as an exploratory biomarker, and as a primary endpoint, we set a definition that relied upon an assumption for controls that was not met. So we did not meet that endpoint. But nonetheless, we've learned obviously a lot about the behavior of ctDNA in these patients. Such data did not exist previously. And the long-term trends in ctDNA are highly encouraging and complementary to the PFS data. And we're doing all of this in a very typical metastatic colorectal cancer population. Roughly half the patients have KRAS mutations and 75% have liver metastases. So this is a very unselected standard population, the kind that you'll meet in the community. And the design of our product has always been with community delivery at the center stage, because we're not looking to develop a product for last-line disease in academic centers. This is something for front-line disease, which is primarily managed in the community. So this is the agenda that I'll walk through today, and I'll move reasonably swiftly. The basic premise is that cold tumors, of which colorectal cancer MSS type is a very good example, do not have pre-existing neoantigen-specific T cells. The tumor has effectively hidden from the immune system of the patient, and therefore the patient is immunologically ignorant of their tumor. If you treat them with a PD-1 antibody, therefore there are no pre-existing T cells to activate, and therefore nothing useful happens in these patients. Our therapeutic hypothesis, and it has not changed since we started the company, is that we will sequence the tumor, identify which mutations create high-quality neoantigens. We do that with our artificial intelligence EDGE platform, which we've published. I'm actually here at AACR in San Diego, and we just had a poster outlining the further iterations on the EDGE platform. Class I, we're now predicting with over 80% positive predictive value, and Class II is getting better every all the time. Having identified the top 20 neoantigens, we make two vaccines containing those neoantigens, an adenovirus that we use to prime, and we use it because priming with adenovirus drives very strong T cell responses. Then we boost the same 20 neoantigens in a self-amplifying mRNA construct. We give these roughly every one to two months as intramuscular injections, just like your standard COVID shot, s o nothing complex about the delivery here. And they're given in sequence, a so-called heterologous prime boost. And the goal is to drive strong, sustained neoantigen-specific CD8 T cell responses. We've published a couple of important papers in the Nature series of journals showing that the technology does everything that you need it to do in principle. So we published back in 2018 about our EDGE platform. We've shown, as I say, that we now predict with high accuracy which mutations create neoantigens. We then put those into our vaccines and delivered them to patients with cold tumors. And we've shown that we activate T cells, that we prime a strong neoantigen-specific CD8 T cell response. You can see it in blood. And then you see those same T cells trafficking to tumors. You can see them turning cold tumors hot, and you can see them killing tumor cells. In our phase I study published in Nature Medicine a couple of years ago, we showed that in roughly half of our advanced colorectal cancer patients, we saw clear evidence of tumor cell destruction with reductions in CEA and the other standard protein biomarkers mirrored by reductions in circulating tumor DNA, which I'll refer to as ctDNA for brevity. In those patients who had those tumor destruction responses, we saw an apparent extension of PFS and overall survival. Of course, it's an uncontrolled study. You are using responder versus non-responder analysis, but this set us up very nicely to test the product in a randomized phase II/III study. Today I'm going to share with you the preliminary data from the randomized phase II portion. These are just snapshots from that manuscript showing the induction of strong neoantigen-specific T cells against the neoantigens in a variety of tumor types, particularly colorectal cancer. These are the overall survival data that I alluded to. The blue are the patients who had molecular and, apparent, biochemical responses compared to those who didn't. The standard survival in this disease is six or seven months. Our non-responders, unfortunately, progressed and died at exactly the expected rate, a clear difference in those with molecular responses. The data I'm sharing with you today are not coming out of a clear blue sky. They're building on these data. These were uncontrolled data. Now we're going to talk about some controlled data. First, a word on MSS colorectal cancer. This is the second leading cause of cancer-related death, about 50,000 deaths a year in the United States. This is second only to lung cancer. So this is a big problem. The reason so many people are still dying is because there's been very little innovation in the field for the last 20 years. Oxaliplatin, which remains the mainstay of front-line chemotherapy, was approved in 2002, and the addition of bevacizumab had a very small effect upon the overall survival. It is used, but the effects are very modest, as you'll see shortly. The median progression-free survival for patients treated with front-line combination chemotherapy is 11 months, and the median overall survival for a patient with this disease after diagnosis is just two years. This is a bad tumor. Immunotherapy in the form of checkpoint inhibitors does not help. Here is the largest study that's been run. This was run by Roche a few years ago, the IMblaze370 study using atezolizumab in a maintenance design, which is what I'm going to show you today with our product. And very clearly, there is no impact, not even on a subgroup. There is no benefit on PFS or OS. ctDNA is an increasingly versatile and powerful tool, and you'll see it today being used in a couple of different ways. It is now often being used for patient stratification because it is a very potent prognostic factor, and you'll see us using it in that fashion. It can also be used for therapeutic monitoring, and you'll see that as well. The rules for therapeutic monitoring are not yet well established, and you'll see today that we fell foul of an assumption that we made given the lack of data. But of course, we've learned, and now we'll move forward with that knowledge. So let's turn to the trial. This is the design of the study, that we'll be talking about. It is a front-line study taking patients with untreated metastatic colorectal cancer, MSS, and this is patients who are electing to undergo standard induction chemotherapy with FOLFOX plus bevacizumab or FOLFOXIRI plus bevacizumab. So that's 5-FU and oxaliplatin plus or minus irinotecan, with bevacizumab. We randomized patients at the beginning, two stratification factors, and then all patients receive the same induction chemotherapy. They're blinded during this phase, and typically at around five months, and the protocol allows up to six months, but most people don't get to six months. Oxaliplatin induces peripheral neuropathy, tingling of fingers and toes, and you drop the oxaliplatin and just move to the maintenance phase with 5-FU/bev, which you've been on since the beginning. The controls just receive standard maintenance chemo. The actives receive that plus the investigational therapy, which I'll outline on the next slide. We were nervous about pseudo-progression with our product, and therefore we elected to put PFS as the first secondary endpoint and chose ctDNA reduction as the primary endpoint. I'll show you the definition shortly. Overall survival, of course, is another important secondary efficacy endpoint. We expect to have those data in the first half of next year. This is the regimen. At the bottom, you can see in the test arm, you receive vaccinations beginning at the start of maintenance. That's the study treatment stage. On your, you hear me using that language. Study treatment stage is when the treatments diverge between the arms, and we start to give the vaccines. We give adenovirus followed by two SAM boosts. Those are the grey circles. We give a second adenovirus at month five, two more SAM boosts. So six doses over a year. We give very low-dose ipilimumab subcutaneously with each of the first two vaccinations. That's given adjacent to the vaccination site. We push drug concentration in the vaccine-draining lymph node that augments the magnitude and quality of the T cell response to the antigens. And we give systemic atezolizumab, which on its own, as I showed you, has no effect in this disease. And frankly, adding ipilimumab doesn't do anything either. That's been studied extensively by AstraZeneca with some single-arm trial work from Bristol Myers, with no real benefit observed. We screen 247 patients. 171 met the overall criteria. Then we do a neoantigen burden screen where we remove roughly the bottom third of patients on the assumption that there are some patients who may not have enough good neoantigens. We do not know that that's true. That is a conservative assumption that we make. At some point, we will test that assumption because right now we remove that bottom third, and perhaps they will benefit. We will see. But we're left with 108, of whom 104 continued, and were randomized. 51 and 53 randomized to vaccine and control. In the first part of the treatment, the induction phase, some patients have early progression. Some patients realize they don't like coming to clinical trial sites for the many visits inherent in a clinical trial. So there's some dropout that's roughly symmetrical between arms. Then they finish induction and move to the maintenance phase, the study treatment phase. Here the patients were unblinded and signed a second treatment consent. Nearly everybody on the vaccine arm continued, apart from one patient who was doing well and had a tumor ablation. We did lose eight patients from the control arm who were just disappointed that they weren't getting vaccine. These patients were biologically the same as those that continued. There's no introduction of bias here, but we do lose a small number from the control arm. 39 and 28 progressed into the study treatment stage and actually received their randomized therapy. That gives us 67 patients in the treated analysis set. Of the vaccine arm, 12 have had end-of-treatment visits, and one of those was for non-compliance. 11 had progressive disease. 69% of patients remain on GRANITE. On the control arm, half of the patients had end-of-treatment visits, all because of progressive disease. So just 50% of patients on the vaccine, control arm remain on study treatment. The demographics of the study and these slides are on our website, so I won't spend long on these. You can see there's good balance in demographics, in age, gender, ethnicity, and baseline performance status. Left and right side was a stratification factor, so it's well balanced. Roughly half the patients have KRAS mutations, slightly more in the vaccine arm, and there's good balance for sites of metastasis. 75% of these patients have liver mets. Roughly 20% of patients elected to choose the triplet chemotherapy induction. This is a low- TMB disease, as you would expect. MSI patients are excluded from this trial. We were able to make and release vaccine for 100% of the patients who are due to receive it. If we turn to efficacy, let's look at PFS. This is the old study that helped get bevacizumab approved. You can see a hazard ratio of 0.83, small effect. Nonetheless, it did something, and so is now standard of care. The median PFS is now around 11 months with the current treatments, and our data is still immature. Our median PFS is around six months. So when you look at our overall PFS, you'll see a lot of censoring. These are the overall PFS curves. You can see all the censoring happening, particularly between months six and 10. I would not pay any attention to the right-hand side of the curves. You're looking at single-digit numbers here. Those data are highly unreliable. But what you can see is there's a separation of the curve starting at around six months, which is when you'd expect the T cells to be forming, given that vaccination typically starts at month five. It takes around a month for the T cells to emerge. So the timing of this curve divergence is as expected, and we're seeing already a hazard ratio of 0.82. Now, because the induction chemotherapy is symmetrical between the two arms, we've been in discussions with the FDA, and it elected to use a statistical test that ignores the first part of the study because obviously events there are unimportant to telling you about the vaccine efficacy. So although a couple of patients actually progress on the control arm, those events are not captured in the hazard ratio. The efficacy assessment begins at month six. Prior to that, all events are ignored. They're given a statistical weighting of zero. After month six, all events are given an equal weighting of one. This is referred to as piecewise assessment, something we've been discussing and have reached agreement with the FDA. You can see this emerging hazard ratio of 0.82. But of course, the question is, there are events here. Is there a way to identify the population in whom events are occurring faster so that perhaps I can look at a more mature data set? Can I do that in an even and fair-handed way? The answer is you can. As I mentioned, baseline ctDNA is highly prognostic in just about every context where it's been studied. We used baseline ctDNA as a stratification tool to split the population in our study into two roughly equal halves. Here are the raw data. Each dot is a patient. You quantify ctDNA as the variant allelic fraction or VAF, and you can see that on the control side, there's a spread of values, and we took the median, the midpoint that splits the control arm perfectly in half, nine and nine. These are all the patients we have data on currently. So it's not quite all of the 67, but it's everybody we have data on at this point. You can see that when we do that, the vaccine group actually is disbenefitted because their median VAF burden is actually a little bit higher, and we know that that's a negative prognostic factor. But when we use this 2% VAF as a cutoff, it splits the patients. This is an analysis just taking all of the patients, irrespective of therapy, and breaking them into these two groups: above 2% in green, so-called high-risk patients, and below 2% in black, so-called low-risk patients. Remember, they're not really low-risk. They're lower-risk, but all of these patients progress, unfortunately. You can see this is an incredibly efficient tool. This is one variable that captures nearly all of the progression events. Now you can ask the question, if I look at that green curve and assess vaccine versus control, is there an emerging difference in this exploratory analysis? The characteristics of those patients are well matched, and what we see in this population is a much less censored, i.e., more mature set of curves that suggest a stronger treatment effect with a hazard ratio at an impressive 0.52. So this is very encouraging and suggests that this is where perhaps our data will mature to disclose a progression-free survival statistic, in this ballpark, which would be absolutely profound for this disease. The low-risk patients just don't have enough events yet, but sadly, they will occur. We just have to wait, and we expect overall population data in the summer. And again, it's not that we expect a disproportionate effect in high risk. We just think you can see events faster in high risk, but we anticipate to see an even effect across all patients in this study. This is simply a way of trying to look into the future. Moderna did a similar analysis using their now well-known adjuvant study with their personalized cancer vaccine in the high-risk adjuvant context. You can see that just as, as we expect to see, there's a group who have ctDNA positivity at baseline. Those are shown on the bottom left here. They occur, or they accrue events much faster, and the treatment effect emerges sooner. And then the same treatment effect emerges more slowly, but just as consistently in that lower-risk population. The approach we've taken has been performed by others with, with these kinds of, outcomes. Let's turn now to ctDNA. To assess ctDNA response, we took one blood draw at the beginning of that study treatment stage, that maintenance phase, and then we needed at least one more sample. We had 49 subjects where we had at least that pair of blood draws. 17 of them actually had negative ctDNA. In other words, chemo had brought their ctDNA down to below the limit of detection, but 32 had positive ctDNA, and therefore we could assess them for benefit. 20 and 12. And we set a definition of response based on our own single-arm data in advanced disease, where a 30% drop in ctDNA was associated with good long-term outcome. We assumed in the control arm that by the time they entered the maintenance phase, although their ctDNA would drop on chemotherapy, that drop would have been played out and their ctDNA would be at a plateau by the time they entered the maintenance phase. Thus, we expected low responses in the control arm. Well, these are what our data showed, and that assumption we made for the control arm was not met. We saw 30% of responses in the vaccine arm, but we saw 42% in the control arm, roughly the same numbers, statistically speaking. So this was a surprise, and the reason is clear from this slide. You can see here, these are the 12 control patients, and the red is the molecular responder patients. The four on the right, you can see that there is a brief drop in ctDNA for the first month, limited to that first month. We think that what's happening here is that there is a persistent effect of the chemotherapy on ctDNA that continues into that first month of maintenance. Then that effect wears off, and the patient's ctDNA bounces up. These patients do not do well. So this is not a meaningful definition of response that we deployed here. So we got this wrong. There were no precedent data. We made an assumption. The assumption was incorrect. If you look at the vaccine patients, we actually see different patterns of ctDNA outcome, and we don't have definitive answers yet. We're learning as we go as these data continue to accrue. But you can see at the top, there are some patients who have these drops. The second row, you can see patients who basically go sideways with bouncing ctDNA. And remember, they're getting intermittent vaccinations. T cells are coming and going. Tumors may be evolving. There's a lot going on in the patient that may be reflected in these oscillating ctDNA assessments. What's very clear are the patients that just have PD, and you can see their ctDNA just takes off and those patients progress. And the axes are different on this slide, and that's probably important as well. So I'll just draw your attention to the fact that those PDs are generally starting higher and just taking off. At the bottom are some of the long-term survivors in our original phase I study, showing some of the patterns that can be associated with good long-term outcome. There's more to learn about ctDNA as an endpoint for therapeutic monitoring, and we're learning as we go. Obviously for us, the good news is PFS seems to be a very good endpoint, and we have no evidence today to pseudo-progression. We really can now focus on PFS, which obviously is a registrational endpoint potentially. This is actually, I think, a very helpful slide where we're looking now at all of the patients with the ctDNA analyses I just described to you. I'll just walk you through it because there's a lot of useful data here. On the left are the GRANITE patients, on the right are the controls, and they're broken into two groups: the high baseline ctDNA, in other words, over 2% at randomization. That's the high, and then the low, the less than 2%. The color coding is PD in red, no PD in green, and the vertical axis is your mean VAF at the very last time point. So, in other words, here we're taking the very first blood draw and the most recent blood draw that we have. So this is an active, live analysis for all of the patients in green because we're just using the last on-study visit for this analysis. The reds, we will have their last ever study visit because they have PD, but the greens, these patients are still on study. What you can see, there are some patterns emerging. If you look at the control arm in the high baseline ctDNA, just about everybody started high, finished high, and they nearly all have PD. So this is a bad group. In the vaccine arm, however, less than half of patients started high, finished high, and have PD, and the majority, even if they started high, they do not have PD, and many actually now are much lower than they were when they started the study. So obviously this is encouraging for what to expect going forward. There's less information in the low ctDNA at baseline group, but you can see perhaps evidence in the vaccine arm that more patients end up with undetectable ctDNA right at the bottom of the axis there. So we look forward to seeing those data maturing. And one final bit of analysis on these very low ctDNA patients. These are the patients who actually went negative on chemo. So remember, they come into the study with ctDNA that's clearly measurable. They have a very good response at a ctDNA level to chemo. Some of these will have residual responses, but perhaps more importantly, some of them, their ctDNA drops to undetectable. And we had nine patients on the GRANITE arm, eight patients in the control arm. And obviously what you want is to stay negative over time. And on the GRANITE arm, six of nine stay negative with what, just 1 PD. On the control arm, however, just three of eight stay negative and three have PD. So small numbers, but as ever, strong signals manifest in small data sets. The signal here is obviously that the vaccine appears to be prolonging ctDNA negativity. We hope, of course, that this continues for a long, long time for these patients. So very exciting data in this low volume disease population, that perhaps may represent the lowest overall volume patients treatment volume tumor volume, where perhaps the immunotherapy has the greatest chance to bring disease fully under control. So very exciting early data. In terms of safety, very few grade 3/ 4 adverse events. There are, of course, the standard fevers, rigors, chills, poor night's sleep, flu-like illness, the things we're all familiar with after potent vaccinations. They're all transient and no patient discontinued study treatment due to an adverse event. So obviously for a cancer patient who's gone through chemotherapy, these are very well tolerated therapeutics, I'm happy to say. So, as we gaze into the future, obviously if we're successful here at really driving strong benefit in patients with metastatic cold colorectal cancer, we're opening up enormous opportunities across the spectrum of solid tumors, both in cold tumors and in hot tumors, in metastatic and, of course, excitingly, in the adjuvant space as well. So this is potentially a very important landmark study, and we're looking forward to those mature PFS data, which we'll have in the summer. Then we aim to start the phase III study next year. We'll have overall survival data next year, which will really help us understand the magnitude of benefit, which helps you power the study to ensure that we get the study right and deliver a very positive result, hopefully, which will be an amazing benefit to these patients who so clearly need better therapies than we're currently able to offer them. And with that, I'd like to thank you for your attention and hand it over back, back to Poorna for questions. Thank you. Thank you so much, Dr. Andrew. So I guess just to get started, you mentioned that you were at AACR also. So I guess if you could talk about what's happening in the field today and, how you think you're differentiated from some of the competitors who are also in working in the same space. Yeah. So the cancer vaccine field has got a long history of negative data. And unlike many novel therapies where the field is blank, we have a weight of negative expectation upon us that makes things a bit harder. Nonetheless, the transformation that we've seen in the last 10 years or so was, I think, fueled by the discovery of the importance of neoantigens and the fact that we can identify those now in patients with cancer. So previously we'd been using self-antigens in vaccines, and we'd been using vaccines that often didn't prime good CD8 responses. And that changed when next-gen sequencing arrived and we were able to sequence patients' tumors quickly and cheaply and then use artificial intelligence to predict using so-called immunopeptidomic technologies, which mutations actually create neoantigens, because it's probably a minority that do. The beautiful thing about neoantigens is that they are not present when you're a fetus and a baby. So your immune system is not tolerant to them. They didn't exist when you were young and forming immune tolerance. They arise later in life as a consequence of normal carcinogens that we get exposed to. There are two special properties. Number one, you are not tolerant to them. They look like foreign antigens to your immune system. Number two, they're only found in tumors. So we're able to now, in principle, drive a very strong immune response against these neoantigens and with the goal of eradication of any cell that displays those neoantigens. We also now are using new vaccines, which are very potent. We use the heterologous prime boost, which is a standard approach, but we're doing it using adenovirus and self-amplifying mRNA and driving very strong immune responses, which we've shown. And that's probably why we're now starting to see clear evidence of efficacy in these tumors where we've never seen anything before. Now, our competitors, Moderna and BioNTech, just use their mRNA vaccines, and the ability of those to prime strong CD8 response is not as clear. Multiple vaccinations are needed, and then often we see immune responses, in a fraction of patients, not all patients, against a small number of neoantigens, not the majority of the administered neoantigens. So there's a potency question, and more data obviously will start to resolve that for us. But this is why we felt confident moving into this, space, this cold tumor space, because signal is very clear. And obviously, if we see the PFS magnitude that we saw in this exploratory analysis, that is, you know, an unambiguously strong signal and would be incredibly exciting for the field. That's kind of why we're doing this. And again, if we see that effect there in that cold tumor, no reason to believe you wouldn't see it in pretty much every other tumor that we would explore. Okay. That's, that's really helpful. So I guess how should we think about the patient population that would be applicable for GRANITE? You mentioned that there are a certain section of the population who are going to be the molecular responders. And then you also spoke about the high-risk group who respond better to this therapy. So just kind of wanted to understand in terms of quantification, how should we think about the patient population? Yeah. We don't think that the high-risk group responds better. We just think they respond earlier. It's always more, it's visible earlier because events are accruing earlier. So we anticipate seeing just as strong an effect in the low-risk group. It just takes longer to see. In terms of who can benefit, if we are able to show activity in this context, then most of the common solid tumors have the same number of neoantigens as colorectal or more. Therefore, we should see broadly similar effects. Now, we don't see a 100% benefit rate. Of course, no drug does. I mean, think about mutant EGFR lung cancer. It's a very homogeneous disease. We treat it with mutant EGFR selective kinase inhibitors. We don't see a 100% response rate. So you never get a 100% response rate. Let's be clear. But there are clearly patients who benefit and there are patients who don't. And obviously we're very interested in the why don't some benefit? And we'll be looking into that. That of course doesn't mean that the therapy doesn't advance across the population because you're showing a population-wide effect. So we're obviously excited to push this forward. Now, GRANITE is a personalized product and it requires us to sequence, predict, and make product unique for that patient. And that's always going to take a little bit of time and be a little bit more expensive than an off-the-shelf product. There's a lot of interest in off-the-shelf products. The problem is that if you look in the standard mutated part of the genome, the exome, meaning the classical protein encoding bit of the genome, there isn't that much sharing of neoantigens. KRAS is the commonest. There's not much behind KRAS. And so if you try to build off-the-shelf products, with KRAS, you end up with a reasonable population of patients that can benefit. But if you want to add a second antigen and a third antigen, you start to run into some challenges. You have to now go looking into what's often referred to as the dark matter of the genome. And there's been a lot of presentations here at AACR on this field, and we work in this extensively as well, of course. And so here you're talking about the non-exomic, non-classical parts of the genome. It's things like, endogenous retroviruses. It's alternative splicing. It's non-canonical open reading frames. It's long non-coding RNAs. It's things like cancer testis antigens. It's things like fusions, unusual genetic fusions. There are lots of different, other sources of tumor-specific antigenic material that seems to be widely shared that potentially we can put into vaccines and drive broad, multi-targeted, strong anti-tumor immune responses. That's how we'll get to off-the-shelf products. So I think that's something that we'll see a lot more of over the coming years. Right now we're seeing efficacy with the personalized products because we're delivering multiple neoantigens. Everybody believes that multiple is good. If you deliver a single neoantigen, the worry is always the tumor will mutate around it. So multiple, multi-targeted attack is almost certainly the right thing to do. We can do that with personalized products. We can't easily yet do that with off-the-shelf products, but that will change. So I think the field's going to continue to evolve in a beautiful direction. We're very excited about where we're at. Okay. That's super helpful. So one of the questions that we have here is, what is the benchmark for your next data readout, which is supposed to be in third quarter? And, what should we be expecting from this data set? Yeah. So obviously, we'll be looking for this PFS data to mature. We have 0.82 currently, and we're looking for that to strengthen. You know, below 0.7 is a meaningful effect that makes a big difference. You know, that's a 30% or greater reduction, in the chance of progression or death. That's the kind of benchmark we'll be looking for. As you saw, Avastin had a hazard ratio about 0.83 in that study I showed you, which was a big phase III study. That's a very modest effect. I think we'll be looking for something stronger with this kind of therapy. The exploratory analysis I showed you in the high-risk group is certainly very encouraging. That doesn't mean that's the hazard ratio we will see necessarily, of course. But I think we're looking for something that in the 0.7 or better ballpark. And if we get there, then we'd be obviously very excited to move forward into phase III, because patients obviously potentially will have a new therapy becoming available to them. And as you saw, they desperately need that. And then, of course, we'll be thinking about adjuvant trials because, again, if it works in metastatic, there's a widely held belief that things will then work better in adjuvant where the disease volume is lower and the disease complexity is lower. There's less antigenic variation, less genomic variation. Okay. That makes a lot of sense. And when you think about the adjuvant setting, would it also be the same where you will have to test for the molecular response of the patient? Yeah. So the trials that folks are thinking about now and a couple of people are running in the adjuvant space do use ctDNA as that baseline prognostic factor or stratification factor, as I, as we used it here. And so, there's a company that's been doing this, Natera, and they've published a lot of data and actually have an approved and widely used product now where if you have localized colorectal cancer, many of those patients are cured by surgery, happily. Unfortunately, not everybody is. And so the question has been, should we give everybody adjuvant chemotherapy? And it's this classic conundrum that you know that some people will benefit from adjuvant chemo. You know that some people don't need it. And how can I tell them apart? And that's been a very big challenge for many solid tumors, including colorectal cancer. But the data with ctDNA suggests that if you post-surgery have one blood draw taken somewhere six to eight weeks after surgery, then, if you're positive for ctDNA at that point, you will recur and you do need adjuvant therapy. You need something. And it's a very efficient way of identifying that high-risk group that is going to recur basically without doubt. That group definitely needs help. Some patients start negative and then go positive. So then there's a monitoring population where, although they're negative immediately post-op, they subsequently become ctDNA positive. In other words, they did have micrometastases, not enough to manifest at that first post-op blood draw, but it emerges over time, you know, a year later, let's say. Those patients, of course, also then need therapy most likely, and there are big trials being run to attest that question. And then there are people that start negative and stay negative and are cured. And if you follow this testing strategy, potentially we can never treat them with chemotherapy that they don't need. So this is the emerging paradigm. There are big prospective trials being run to assess this. Japan, actually, in particular, has taken the lead here, running the CIRCULATE series of trials. So that's the context in which we'd be thinking about adjuvant studies of enriching by selecting for people who are ctDNA positive after their surgery and then randomizing those patients to receive vaccine or standard of care. And actually, it's an interesting question. What is standard of care? Is it now chemotherapy for those patients, or should you wait until they have overt, visible metastatic disease? That's the question that's being answered right now. Okay. That's really helpful to know. I guess how, like thinking about the phase III trial design, like how are you thinking about it and what kind of endpoints do you think are most suitable? Are you looking to get some sort of accelerated approval or like, yeah. Yeah. So obviously we didn't know what our efficacy, how it would manifest. That's why we, you know, chose these different endpoints for phase II. And remember, phase II is a signal-seeking trial. It's not a registrational endpoint. And so, obviously we missed on this primary endpoint the way we defined it, but that's not that important. What matters is the totality of data. Really the question you're asking in a phase II study is, do I see signal that merits going to phase III? What endpoint should I use and which patients should I treat? So are we seeing signal? Yes. Is it strong enough to merit phase III? We'll find out in the summer, but everything we see suggests yes. What endpoint should I use? Looks like PFS is a good endpoint. We'll need to talk to the agency, obviously. So we'll have an end of phase II meeting later this year where, of course, we'll take these data and ask that question and hopefully lock in on PFS as the primary efficacy endpoint for phase III. And the only question is whether you allow crossover or whether you keep the two arms separate to enable the overall survival endpoint to emerge secondarily. That'll be a key topic of conversation. And then thirdly, which patients to treat based on what we know, we would just treat the same patients we've treated in this phase II. Obviously, the high risk, we see the signal a bit earlier, but we anticipate seeing the same signal in the low risk. Again, we'll know that in the summer. So in principle, we should be able to move to a relatively straightforward phase III study, essentially replicating what we've done here in phase II, and using PFS, assuming the FDA is supportive as the primary efficacy endpoint. Okay. That, that makes a lot of sense. And I see that we're out of time now. So, and we also don't have any more questions. So I'd really like to thank Dr. Andrew Allen for taking the time to speak with us today and for presenting all of this and, looking forward to the data. Indeed. Thank you, Poorna. Appreciate it. Thank you so much. Bye. Bye.