Good morning, everyone. My name is Daina Graybosch. I'm the analyst covering immuno-oncology here at Leerink Partners, and I'm excited this morning to welcome Anat and Eran from Compugen. So we're going to start with a quick introduction from Anat, and then we'll jump right into the fireside questions.
Thank you, Daina, and thank you for Leerink for inviting us. We're happy to be here in Miami. A few words about Compugen: we're a clinical-stage cancer immunotherapy company, and we are a pioneer in the field of computational discovery. We actually had a great 2023, both in terms of execution and in terms of validation of our computational discovery capabilities and early-stage pipeline. We ended the year by entering a collaboration with Gilead around a leading asset in the field of IL-18 pathway inhibition with an antibody licensed to Gilead. We presented data in 2023 showing that COM701-driven activity in multiple tumor types. We showed some initial biomarker data. We initiated two proof-of-concept studies. There was also progress with the AstraZeneca collaboration that we have with the TIGIT/PD-1 bispecific.
We're off to a good start in 2024, with a strong cash position taking us into 2027, allowing us to put money on our clinical-stage programs and early-stage pipeline.
Awesome. So let's start with the proof-of-concept studies because we have data reading out from both of them this year. I want to start with COM701, COM902, and Pembro, so PVRIG, TIGIT, and Pembro, and microsatellite stable colorectal cancer. Maybe you could talk about the data that's coming and what you need to see in that readout to take on additional clinical risk and additional investment in that indication on your own.
Sure. So, you know, MSS-CRC, the indication that we're pursuing, third line, and more, this is really a very challenging indication. Patient population is not responsive, many IO failures and other failures. And this patient population is, actually consisting of more than 70% of the patient population, including patients that have liver met. And this patient population is really not sensitive to drugs. So the data that we've presented up until today is about 12% overall response rate and some stable diseases in this liver met population. What, the way we're looking at it is we'll look to see, the current study is also adding TIGIT. So this is the data that we showed was in a doublet of COM701, plus PD-1 inhibitor, so PVRIG, PD-1 inhibition. Now we're adding also TIGIT. It's a small study. We will look, obviously, at overall response rate and disease control rate.
We will also look to see if we can get some survival advantage over standard of care. The current standard of care is Bev, TAS, and the overall population survival is 10.8 months. We'll try to see how our data looks against this data, taking into consideration two things. First, this is the overall population, and we, our data indicated that we have some edge in the liver met. There is no data on overall survival in liver met. So we'll take this into consideration, and also the fact that this is a single-arm study, non-randomized, small study, so taking survival with caution. I think that, you know, disease control rate is also very relevant in this indication, and I'll explain.
If you look at the Bev/TAS study, so you could see that the overall response rate was 6.3%, as compared to TAS, which was 0.9%. But the disease control rate was 76%, and this was more in correlation to the overall survival. So we believe that also disease control matters, durable stable disease. So we'll try to look at all the data and see how we're compared to benchmark and make decisions.
I think you really appropriately caveated how to use the overall survival data. I think others have gone forward with maybe even higher responses. The one that comes to mind is the Merck study with favezelimab plus Pembro, which had a higher overall response rate but still failed in their phase 3 to improve on this just actually not great benchmark and standard of care. I mean, is there anything else you can look at in terms of biomarkers, molecular response, to help you make this decision?
So I'll let Eran relate to biomarkers or molecular responses, but I think that in general, when the overall response rate is low, it is very beneficial to look at additional parameters and chime everything in. And that's what I was saying, that to look at the overall survival rate plus the disease control rate and durability of stable disease, that will make sense. But Eran, go ahead with.
Yeah, probably refer to ctDNAs, which are extremely informative in the earlier stage of disease, in which you can follow minimal residual disease and get some more information. In late-stage patients, metastatic settings, it is also could be informative. There are some studies showing correlation with overall survival, so it could in some way support our decision. But while you put a lot of efforts into biomarkers and mainly into predictive biomarkers, can we find ways? And this is relevant also, by the way, to MSS-CRC. We are following to see if we can find ways to enrich for these responding patients we have recently shown. So in terms of biomarker discovery, yes, we are looking at all parameters from blood to tumors, baseline and also on study.
But in terms of making clinical decisions, we personally follow the more traditional ways of responses, as Anat mentioned: durable stable diseases, and eventually overall survival rather than ctDNAs.
Is it fair that we should see higher ORR or DCR in this study, given you're adding on TIGIT? And can you think you'll be able to make a decision following this data on whether to include TIGIT or not? I mean, three is harder in terms of contribution and regulatory path and commercial path than a doublet.
It's a very good question. Look, the data outside, at least with the multiple phase 2 studies, is showing that TIGIT is adding a benefit over, over PD-1. Obviously, everyone is waiting for phase 3 data, to confirm this. I think that from our data, if I may look at it at the different indications, at least for the ovarian cancer, for the platinum-resistant ovarian cancer, we had 20 patients, with doublet and 20 patients with a triplet without, TIGIT. We could see some difference in terms of overall response rate, durability. It's hard to say that it is attributed to this, combination. We believe that with the, tumor microenvironment data that we generated, the translational data, there is some, COM701-driven effect. And Eran maybe will relate to it later, how PVRIG is sensitizing the tumor to respond to PD-1 and hopefully TIGIT. So that's on the ovarian cancer.
We believe that the data that we have was showing some support of adding TIGIT to the triplet to make it a whole and to maximize the response. In CRC, it still remains to be seen. The data that we shared was doublet data, and we're now adding TIGIT, and we'll see. Still saying that it's more studies in order to make a very, very conclusive statement, but I think that we may be able to see some hints.
Yeah, I think in general, the totality of our data, from pre-clinical to, as Anat mentioned, whenever we did test doublet and triplet in the same indication, is showing that if you combine all three agents, it is synergistic. So we think that triplet should work better, but eventually, specifically in MSS-CRC in this trial, with the caveat of patient numbers and all of that, we'll have to wait and see.
Moving to ovarian, let me ask the same question. What's a reasonable efficacy benchmark in 20 patients with this triplet in ovarian that would make you particularly excited to invest more?
So again, another challenging indication, PD-1s are not really responsive in this, in this indication. And the data that we shared up until now from 20 patients with a triplet, we're seeing 20% overall response rate with really durable responses of those patients that responded more than 16 months. I did not mention it for the CRC, but again, also here, good safety profile, which also allowed for these patients to have a better quality of life, which matters. The investigators in our study kept telling this to us. So that's one. We're looking with this study to see if we can further strengthen the data, to see if we can repeat this overall response rate. I did not mention, but we also showed some initial correlations for a biomarker for the ligand of PVRIG. PVRL2 was correlated with clinical benefit.
We're trying to see if we can strengthen this data as well and see if we have a biomarker that we can pursue forward in this indication. Really repeating the data, overall response rate, durability, safety. Again, also here, at the end of the day, to be able to show that there is advantage in survival, that will make sense. Again, small study, single-arm, but the standard of care at this point in time is a single-agent chemo, which gives survival of about 1 year. There is also the ADC treatment that is showing survival of 16 months. I think that also here in this respect, we're looking to see which patient populations would be relevant for us, those that are ineligible for ADCs, and/or those that progressed on ADCs.
So we'll look at the totality of the data and see how to proceed.
Assuming you confirm the PVRL2 biomarker signal, what's the path forward to actually use that for enrichment? And is that what you would do, or would you do larger randomized studies all-comers looking retrospectively? How should we think about that?
So I will let Eran relate to the path to how to get to an assay for a clinical study, etc., but assuming that we will have data supporting for an enrichment strategy and assuming that the assay will be ready to be tested for patient selection, we believe that moving forward in a biomarker-driven study can give us an edge in this patient population on both fronts. First, increasing the probability of success with patients that may likely benefit from the treatment, but also in discussions with the FDA, which is very favorable for patient selection studies. So we believe that if the data is there, we'll go with this, but it doesn't mean that if we don't have a biomarker and the data will look good, we'll not go ahead without an enrichment strategy.
Yeah, so, so where we are. So, I mean, first of all, PVRL2 is the ligand of PVRIG. So this is the interaction we are blocking. So we're following in all our trials on PVRL2 levels. So we initially developed an IHC assay to be able to really follow that and to correlate to response. And then gladly, we have seen that in this ovarian cancer trial, we see we saw that patients who drove clinical benefit from doublet or triplet combination tended to have higher PVRL2 expression. It was also interesting to see we saw similar observation in a small number of patients in our breast cancer trial, further supporting that maybe we have here a biomarker that really could inform.
And the fact that some patients on the trial really drove long-term benefit with long-term survival, and these patients, many of them had higher PVRL2 level, this drove us to now go to the next stage, in which we're optimizing now the assay while we continue to follow in our ongoing trial. We continue to follow do continue to see this correlation. We are now optimizing the assay to be able to select patients based on that and to find the threshold, what is the level of PVRL2 that eventually we're gonna we're gonna focus on. And hopefully, if all goes well in the next trial, we could do a randomized study or other type of study in which we preselect patients based on this, on this assay that we are currently optimizing. And then, of course, all the way to CDx is if all goes well.
I remember early on talking about PVRL2, is that it has expression in normal cells as well? Is that challenging, per this assay development, or do you see, like, a big difference?
Well, PVRL2 has expression also in some normal cells, but overall, it is induced in cancer. Eventually, yeah, we do see this initial correlation to clinical response. In terms of the assay itself, it's not an issue. We're focusing on the tumor cells, and we have an assay ongoing, no specific issues there.
Business-wise, what's on the table business development-wise for both of these assets, the PVRIG and the TIGIT?
So I'll say that first, we, you know, based on data, we're committed to take these programs forward, obviously, in a path that will make sense, based on the data. But, we're also, we're also interested in exploring potential collaborative arrangements. We believe that, with the data that we showed of COM701-driven activity across tumor types, and we tested mainly in non-inflamed indications, it served us in order to be able to conduct small studies, single-arm studies, and not facing a situation where the activity is considered to be driven by PD-1 inhibition. But as we move forward in this, we believe that there is, and we're focusing on two indications. We believe that there is additional potential to be tested with COM701 combinations.
It could be in non-inflamed indications with data that is based on what we've already discovered, but also in the inflamed settings, which we didn't pursue. There is a real ground there to pursue COM701. This will allow us to broaden the opportunities and to maintain our leadership as well.
And maybe this is asking it in the same way, but PVRIG, you guys had a lead, but there's increasing attention with other assets in the clinic, including, I think, some companies in China looking to go to hepatocellular, which, you know, biologically, I think, makes a lot of sense. And then GSK has one as well. So how do you maintain your edge?
So first, as I said now, exploring the potential collaborative arrangements will help us. I think that the fact that, after testing, these combinations in multiple tumor types, focusing in order to drive into specific indications, I can say that IP-wise, the data that we generated allowed us to come up with a stronger IP. We have some composition of matter claims in Europe that generate some freedom to operate challenges. And we have data protecting ourselves across indications, combinations, bispecific, and we have the expertise. We're working on this, with groundbreaking science, of this pathway and axis for quite some time. We have the systems and the tools, in order to move forward and the knowledge. So that's how we see maintaining our leadership. And by the way, we see the fact that other companies are following us. It's a good thing.
This is validating for us. This was a target that we were being asked, "PVRIG, who?" And now it is not the case. So we see it also as a validating.
Okay. So on the biology, conversation we've had many times in many formats. Genentech recently published a paper in Nature, detailing some data they recently I think you they presented at a conference last year, maybe AACR last year, that was supportive of their Fc-competent design for their TIGIT antagonist, tiragolumab. And specifically, it was around the immune-activating impact on the myeloid cells and the Tregs from their clinical study. And I wonder what's your take on the findings and how you reconcile it with your own IgG4 strategy for COM902, but also COM701.
So first, I think it's great to see another high-profile publication, a Nature publication, on the target. There's a continuous interest from many pharma companies, phase 2, phase 3, trials ongoing. So TIGIT, we believe, is an active molecule, and it's great to see the continued interest. Specifically for the question, I think that maybe it's already answered. So the proof is in the pudding. Arcus did present data from randomized study comparing PD-1 combo and PD-1 plus TIGIT with their Fc non-binder TIGIT antibody. It was extremely effective, maybe even a better safety profile. AstraZeneca are moving with their COM902-derived bispecific antibody into phase 3 now, probably also based on data that they are seeing. So I think it's already quite clear that the TIGIT with non-Fc active is active clinically quite efficiently.
Specifically for what Genentech have published, so, you know, the first part was about correlative studies of complex, the convolution algorithms in bulk RNA. Without going into all the technical details, it's really more correlative and not causative. And then the second part of the paper is about mouse. And we all know that in mouse, these Fc studies didn't work out in the past. PD-L1 was shown in the mice to require Fc binding for its activity. And we all know that the PD-L1 Fc non-active are perfectly fine in the clinic. So overall, yes, we have IgG4. We like the PD-1 blockers. We focus on pure blockers like some of the others. We kind of know for TIGIT, it's active. We see the signals of activity for COM701 as well.
So, we expect to see continuous activity and, again, potentially better safety profile for the non-Fc active molecules.
So I agree, myself. On the clinical correlates, actually, have you looked at those same markers? Do you see any correlation even with an IgG4 backbone in your studies?
So the number of patients that they had, and looking at the correlation, which are not very strong even in their study, it's really difficult to say. I can say that in general, when we combine COM701, we seem to see different biology. When we see COM701, we see activity in PD-L1 low indications. In, from what we published until now, we didn't see PD-L1 as a predictive biomarker in general. While for TIGIT antibodies, TIGIT PD-1, it seems to be the case that you need PD-L1 high patients. So I would say that we are looking all the time. I would say that in general, when COM701 is in the picture, the, predictive biomarker seems to be a bit different. And also, again, the number of patients, we are not there yet in terms of numbers to really see these correlations. But it could be.
I mean, we have some signals of myeloid activation in the blood in few patients in our non-FC active molecule. So yeah, we'll see.
Another biology question, and you guys have gone back and forth on this, with your own strategies, is, you know, you're going after pure IO regimens in very cold tumors. So you set a high bar, and then you make it even harder by having just 100% IO. I think other companies, with their TIGITs, have abandoned the pure IO and going just for chemotherapy combinations. They have a little bit: is it reducing the tumor burden? Is it stimulating a sort of in situ vaccine? And I wonder what gives you confidence that you can move forward IO alone versus adding a cytotoxic?
Yeah, so, you know, first of all, it's a very good question that obviously we ask ourselves as well. We think I mean, what we know is that PVRIG has a very different biology from the other checkpoints. And that's the reason we are trying to pursue a bit of a different strategy because we know that, and again, without going into all the details, but PVRIG, that unique expression on stem-like memory T cells seems to be able to, when you block it, really drive T cells into less inflamed tumors. So as always, we first make our decision based on research and biology. So we have this unique biology of PVRIG. Then we have the clinical and translational data.
I mean, we have seen from the early days in clinical testing, we have seen activity in places where PD-1 or PD-1 plus TIGIT or even other checkpoints are not really working. We've seen it clinically. We've seen it translationally. When we treated the MSS-CRC patients, 70% with liver metastasis, we saw immune modulation in tumor microenvironment following PD-1 plus COM701, which resembles what others, others have shown in inflamed settings. We saw a complete modulation of tumor microenvironment in this patient, which is really surprising for us, and along with the actual clinical data. So we have these signals clinically and translationally that when you add COM701, the outcome is different from other checkpoints. Obviously, at the end of the day, we'll have these ongoing trials at the moment. We'll have to see how they materialize. Maybe eventually, we will need to add TKI or chemotherapies.
At this stage, with the excellent safety profile that we are seeing, we think it's a good strategy to follow the triplet in these indications, and we'll see how the data will develop.
Let's switch in the last five minutes and talk about COM503, the IL-18 binding protein, and maybe use that as a way to talk about your computational platform. So I wonder if you could tell us the story of IL-18 binding protein. And, you know, we talked about, and I'd like you to relay it, how your unique computational integrated with wet research led to that program that ultimately Gilead recently licensed.
Yes, we have this when we are pioneers in computational discovery of new targets. I mean, we discovered TIGIT around the same time that Genentech did and then ILDR2 and PVRIG, so we're repeatedly able to discover new targets and bring first in class. Specifically for IL-18 binding protein, so we are asking a computational query. Can we identify in the tumor microenvironment a resistant mechanism to PD-1, which is employed by tumor-associated macrophages? So we looked computationally to see, first of all, if we have upregulation on specific subsets of macrophages compared to others. For example, IL-18BP was upregulated following PD-1 treatment, again, suggesting a negative, immune resistance mechanism to PD-1, and then also other attributes. So, so first, we identified IL-18BP as a potential cancer-resistant mechanism before it was discussed by others. It was quite a few years ago before any publication. So this was first, computational identification.
Along with computational identification, we also developed along the years all the means and tools to be able to validate this target fast and efficiently. So once we identified as a potential target computationally, then the wet lab is coming into play. And we have these tools really to be able to, you know, we're quite small, and we are able to discover again and again targets that the big pharma is not able to identify because we have these skills and know-hows, how to look at the target, to ask ourselves, "What is the most burning question on this target that we want to answer in the most efficient way?" Then we have all the tools and systems to validate the target. And once it's validated, then we start really moving aggressively toward the clinical development. This is what we're doing now with IL-18BP together with Gilead.
Can you talk about it as a target? I think that most other approaches are going after IL-18, you know, dosing that as a cytokine with various modifications versus you guys going after the binding protein. Why that strategy?
So IL-18 is a great cytokine. But we all know, I mean, cytokines are great tools. I mean, they are approved clinically, but it's also very poor drugs in general, cytokines, not only IL-18. And people are trying for ages to engineer them and to give them systemically in different ways. But eventually, we have this challenge that you give the cytokine systemically, probably inducing first toxicity before you even reach sufficient amount in the tumor to really mediate anti-tumor effects. So this therapeutic window issue is common to all cytokines. What we are doing is very different. We are not using a cytokine moiety, but rather a drug, an antibody blocker moiety. And we exploit the fact that you have a lot of IL-18 in the tumor microenvironment naturally there because it's inflammasome-induced.
You have IL-18 binding protein, which is an inhibitor of IL-18, which is complex to it in very high affinity. So in the tumor microenvironment, you have a lot of IL-18 floating in an inactive way. And when you use our antibody, in the periphery, it's quite low. So what we have shown preclinically, when you use our antibody, you unleash, you inhibit the inhibitor. Now IL-18 in the tumor microenvironment is really active. The periphery is completely clean. And what we are able to do is exactly the opposite of what most other cytokines are doing. We're modulating the tumor microenvironment without inducing toxicity, without modulating the periphery.
On the computational side, and you've been a computational company for a long time, and there's a whole new crop of AI companies, can you talk about how, you know, your skills and your platform are differentiated over this or compared to the new crop?
I'll just start, and then I'll later on relate to it. I think that it's really on a few components. This is the computational capabilities where people are saying today, "AI, we believe that it's not only AI. This is AI with human expertise, the integration with drug development capabilities and wet lab, and the proof of concept." We have targets that we've identified and moved to drugs in the clinic. But, Eran, you'll probably want to relate more into the details.
Yeah, so, you know, so some companies are single technology companies. We are an AI company. We're a transcriptomic company. So we are have this expertise developed along the years. And we are fast to adopt new technologies, but we employ different co-technologies. For example, we are one of the first to employ 3D looking at tumors in really in the three-dimensional space. So when we thought this is the right technology for us to explore new targets, then we used the technology. So obviously, we're using AI. Obviously, we're using different, kind of technologies. But eventually, the secret sauce is really the human expertise developed along the way, from computational biology, target discovery to, as mentioned before, validation tools. And all of this eventually brings to a continuous flow of targets to feed our pipeline, which are, again, all new biology, first-in-class targets.
Is all your computation focused internally? Early on as a company, you did more external partnerships. I wonder where you are in that journey.
It's a very good question. Yeah, in the beginning, the idea for the company that was a compute-pure computational discovery company was to enter into collaborations, on the outputs or on discovery collaborations. We were there when there is a huge skepticism in the field on computational discovery. And we realized that in order to generate revenues for the company, we need to translate on our own the discovery capabilities into drugs and then enter into collaborations. I think that in the last two years, the field started to look differently on companies that are computational companies, what is called AI companies. By the way, I think that a little bit with more naïve on the space and the fact that there is an attrition rate moving from computational discovery to drugs at the end of the day.
And I think that Compugen have done this process, a few times already. Today, when we are having a pipeline that is more mature, with diversified pipelines stemming out from our computational discovery capabilities, we're not ruling out, looking carefully on whether we want also to externalize also these capabilities. But that's it will still continue to feed our own pipeline.
Okay. Well, thank you both. I appreciate it. I think we can conclude there.
Thank you very much.