Greetings, and thanks for joining us to have a conversation with Eran Ophir, CSO of Compugen. Compugen is a clinical stage biotech company that uses AI and ML to design first-in-class I/O therapeutics. The company has established two partnerships with AstraZeneca and Gilead. To discuss the company's development strategy in 2025 and beyond, I welcome Eran to this fireside chat. Eran, thank you very much for joining us and having a conversation with our audience today. Hi, Alkhid. Thank you for having us. To start off, can you give us an overview of Compugen, Compugen's history, I mean, and what's the area of focus for the company at this time? Sure. Compugen is a clinical stage company focusing on immuno-oncology. As you mentioned, we're actually a pioneer in computational discovery of targets. We use our computational platform to identify novel drug targets in the field of immuno-oncology. We've been doing it for quite a while before all this AI hype arrived. We're using computational tools, including machine learning and other tools, to identify novel drug targets. We actually identified TIGIT around the same time Genentech did. Later on, we realized that to make the full potential out of our computational capabilities, we should actually develop clinical and developer capabilities to be able to take our discoveries all the way to the clinic. This is what we do. We have now two fully owned assets in clinical testing, COM-902, which blocks TIGIT, and COM-701, which blocks PVRIG checkpoints. We also license the right to use COM-902 as part of the bispecific Rilvegostomig of AstraZeneca, a TIGIT PD-1 bispecific that they are now testing in multiple phase III studies. More recently, using Unigen or computational discovery platform, we identified another first-in-class asset, an IL-18 binding protein as a target for a blocker antibody. This one was licensed to Gilead last year, and it is now in phase I that we are leading. We also have multiple other non-disclosed assets in the early pipeline, all of them in the field of immuno-oncology, all of them derived from Unigen or a computational discovery platform. Very good. On the platform itself, which you call Unigen, what's the model? What's the platform? Also, can you highlight some of the validations that you have gained on this model? Sure. First, this is an AI conference, and companies and people are using AI in different stages of the research and development. They use AI to optimize drugs, they use AI to identify biomarkers. What we do is quite unique because we go to the very first stage of identifying the targets themselves. We use AI and other computational capabilities to identify new drug targets. Eventually, if you look at the clinical trials, phase I, phase II, phase III, most of them are around the limited number of targets. We bring novel targets that we identified computationally. I think this is one thing that is very unique, and talking about how we employ specifically AI in our models. We have this, like in any AI, I would say that the center of it is a big database that we build along years of multi-omics analysis of tumor environments. We use AI to build the database. We use AI tools to harmonize and QC the database. Obviously, we use AI tools like large language models to ask the query themselves to identify these novel drug targets. Now, along the years, not only are we able to discover computationally the targets, but we develop the capabilities, how to validate the targets, how to identify quite early on which is a good development candidate, and then take this all the way to the clinic. I think that eventually the proof is in the pudding. Many people claim to do many things with AI, but our platform is validated because we're quite of the few companies that are able to identify multiple first-in-class assets, some of them validated by deals with big pharma companies like AstraZeneca, Gilead, Bayer in the past, Bristol Myers. Some of them also have some initial clinical proof of concept. I think the uniqueness is in the fact that we're focusing on bringing first-in-class drug targets and that the platform is validated. Very good. Thanks for that. As we started off discussing about Compugen, you talked about your partnerships. One of the major partnerships has been AstraZeneca, where they are working on your TIGIT molecule. In general, how did you come about identifying TIGIT? TIGIT has had a unique development history so far. How does your TIGIT antibody differentiate itself against the rest of them? If you can just start talking a little bit about the molecule, then maybe later we can discuss the commercial potential and the economics of the deal. Sure. As most of us know, TIGIT had a complex development history. It It started off with some of the big pharmas with actually validating data in phase II randomized studies. Later on, there were also some failures, especially in phase III, and people started to doubt the target. We need to remember that not all TIGIT antibodies are the same. Mainly the issue of the Fc. Some, especially one of the, I think, the first movers, Merck and Genentech, that also later on had difficulties, used an Fc-active TIGIT molecule that is associated with depletion of cells and potentially with some increased side effects that I think was part of the reasons that they had these challenges. While others like us, AstraZeneca, Arcus, Gilead, which are now in phase III as well, chose a non-Fc active, which is meant to block TIGIT like you do with PD-1, Keytruda, Nivolumab, without depleting anything. It seems to be that this kind of asset has a much better safety profile, less discontinuation rate. Given the fact that, well, TIGIT was shown to be active in pre-randomized studies, but like some of the other checkpoints, the magnitude of effect is probably not huge as we thought at the beginning. Having an active molecule, which is also safe to be able to combine it, makes TIGIT drugs which are non-Fc active potentially better in terms of risk-benefit. This is exactly how COM-902 was developed to block TIGIT. COM-902 also has a best-in-class potential. It's very strong, high-affinity antibodies. This is probably one of the reasons AstraZeneca chose this specific TIGIT molecule to be incorporated as part of their TIGIT PD-1 bispecific. They already shown data in gastric, in non-small cell, in non-randomized studies to this point. The efficacy seemed to be impressive and definitely better, well, taking into account comparison across trials, but better than historical data on PD-1s alone, and a very good safety profile. This is probably one of the reasons that AZ are using this Rilvegostomig molecule, Rilva, as a new I/O backbone. They take this molecule, which is a good, safe I/O compound, and they combine it with different ADCs, chemotherapies, and now opened eight phase III trials using this molecule as a backbone and combining with different combinations. We are really eager to see how this will develop. Two questions from that. One, you said AstraZeneca has a large set of clinical programs on the back of R ilvegostomig. What sort of indications are they going after, and what sort of data would we be seeing in the next couple of years? Then on the economics itself for Compugen, how much can you tell us about what the milestone payments are or when they could be coming? Sure. They are following mostly non-small cell, different trials, some are in the early settings, some are in the late-stage settings, and gastrointestinal. They are combining with either chemotherapies or ADCs from B7-H4, TROP2, Enhertu. I think what is nice to mention is that even though we are quite certain that TIGIT is active and that Rilva is more active than Pembro and other PD-1s, in some of the studies, they are actually not even comparing necessarily directly Rilva to Pembro. It is a bispecific antibody. Nobody can ask for contribution of component. If you are adding Rilva to Enhertu and compare it to Pembro plus chemo, you are joining the force of Enhertu and Rilva. Potentially, if the treatment has a good risk-benefit profile, you can get potentially approval based on that. I think this is also something worth to mention. They plan to show this year first glance of a non-randomized study yet of Rilva combination with ADC. That probably will show us why they chose to move so aggressively in so many phase III studies. The phase III readouts themselves probably no later than 2026, 2027. This is for the trials. About the economics, the deal was a deal of $200 million total. We received until now $30 million, and we're eligible for up to additional $170 million based on development, commercial milestones. What is even maybe more interesting, the fact that we are eligible for up to mid-single-digit royalties. Given the fact that AstraZeneca are predicting target of sales more than $5 billion of sales, obviously, this will materialize. We'll be eligible for an annual income of royalties, and this will be a very strong driver for Compugen to be able to continue and drive our pipeline from this income. Okay. Talking about income from the pipeline, you also signed a very large deal, I would say, with Gilead on GS-0321. What's the molecule and what's the deal structure with Gilead? Yeah, so this was a really interesting one because we're looking, using Unigen, the computational platform, into a tumor environment to identify the resistant mechanism of tumor-associated macrophages. We're doing it for quite a while, and we developed a new algorithm, and we had this output of suppressive mechanisms. One of them was this IL-18 binding protein. This is also where sometimes non-artificial intelligence, some used to call it human intelligence, is also important. Our expertise in immuno-oncology and development enabled us to pull out this one, this target, out of the list. What is unique about this target, even though we're not looking for a cytokine program, this observation took us into the cytokine world because this target is an inhibitor of a cytokine which is naturally occurring in a tumor environment, a cytokine called IL-18. By using this antibody, and IL-18 is really high in a tumor environment and low in the periphery. By using this antibody, what we have shown preclinically and the reason the target was attractive and pulled this relatively, as you mentioned, large deal in the preclinical stage was because we showed that potentially this target could overcome many of the challenges that other cytokine therapeutics have encountered, namely systemic toxicities, challenging therapeutic window. What we have shown is that we are modulating with a cytokine that is naturally available in a tumor environment. We are modulating the tumor environment in the periphery like completely, completely clean. This is exactly the opposite from what you see with cytokine therapeutics. It was a preclinical asset, but it was a competitive process in which Gilead eventually, we signed a deal with Gilead. It was an $850 million deal. We received $60 million upfront. Recently, upon IND approval, we received additional $30 million. What is also interesting about this deal is that even though Gilead licensed the deal, we are taking it into phase I. Not only did they trust our discovery and computational capabilities for this first-in-class asset, they also trusted our development and clinical team to take efficiently, and I would say with flexibility of a biotech, to take this drug into phase I. In later stages, Gilead obviously will take over. As mentioned, total deal value of $850 million and tiered royalties up to a low double digit. Perfect. In terms of the antibody itself, how did you identify this target and what's the mechanism of action of the clinical candidate? The mode of action, it's a blocker antibody. It's a high-affinity blocker antibody that displaces IL-18 binding protein away from IL-18. We're actually inhibiting the inhibitor. In the tumor environment, there's naturally a high level of this complex of IL-18, IL-18 binding protein complex together in an inactive form. Our antibody displaces IL-18 BP away from IL-18. By that, enabling IL-18 to activate very potently the tumor environment. Again, without modulating the periphery. This is really a unique MOA for cytokine therapeutics in general. Okay. In terms of the clinical indications, what do you think are the appropriate indications that GS-0321 can talk about? We have shown and published that the expression profile of this IL-18- IL-18 BP complex is quite broad in many tumors. We do think that probably one needs some basic level of inflammation or T cells, NK cells to respond to this endogenous cytokine. We think the potential is broad. We start in dose escalation in all comers and solid tumors. In the part two of expansion, probably going to expand in specific indication based on the biology and based on the signals we'll see across the first part of the clinical trial. Great. Now, I'd like to pivot to your lead program, COM-701. This is a PVRIG antibody. The PVRIG is also known as CD112R. Can you help me understand what is this target and why is it a good target of ovarian cancer that you are currently evaluating? Yeah, this is a very good question because checkpoints historically have been very poor in ovarian cancer. PVRIG is a checkpoint, but it's a very different one. First, it had a very high expression. PVRIG and the ligand, PVRL2, are very dominantly expressed in ovarian cancer. This was a target indication for us from day one. Being highly expressed is definitely not enough to modulate the relatively resilient tumor environment of ovarian cancer. What we also have shown and published quite extensively is that PVRIG has a very different biology and actually focuses in different stages of the cancer immunity cycle, inhibiting stem-like memory T cells, which have very strong proliferative capacity. What we have shown is that if you use COM-701 to block PVRIG, you can really unleash anti-tumor immunity. We showed it in a few indications in patients which are typically not responding to immunotherapy, like patients with ovarian cancer. In the very last line patients in the Platinum-resistant settings, in patients who failed everything, including patients who came from other clinical trials, we were able to show some monotherapy signal. We did not test many patients in monotherapy, but we did show monotherapy signal both translationally with the MOA of COM-701 and actual clinical long-term durable response. Also in combination, mostly triplet combination, we saw quite strong data compared to historical of PD-1 plus TIGIT. Therefore, we think that it is not only the biology, but the clinical signals in the last line patients are supporting COM-701 activity in ovarian cancer. Now, looking at our results, what we identified is that some of the patients had really long durable responses, excellent safety profile, but some of the patients, and unfortunately, more than 50%, and we eventually to have a success in Platinum-resistant or in any trial, we need to have improvement in median. Many of the patients relapse very fast, some of them so fast they probably never even had a chance to respond to chemotherapy-free regime like COM-701 in combinations with other I/O agents. One option would be to go with a biomarker that we are working to identify a biomarker, but this is still a work in progress. Another option would be to enrich by moving into the earlier settings of disease, the Platinum-sensitive setting. In these settings, we're taking patients who already responded to the Platinum, so they have low tumor burden. is also data to show that the Platinum chemotherapy sensitizes the tumor environment to the unique biology of COM-701. We have a more active immune system. To put a drug which is active and safe with durable response in these maintenance settings, we thought this is the exact place where a drug like COM-701 would fit. I see. Can you please remind us, what's the current standard of care in the maintenance setting of the Platinum-resisitive patients? What do you see as the market opportunities for COM- 701? We start by going into the second, third line. The patients normally in the first line receive Platinum. They receive maintenance, Bev or PARP. Unfortunately, many of them relapse. They receive maybe another maintenance, maybe depends on the HRD, BRCA status. Eventually, when they get to the second or third line, they receive the sixth cycle of Platinum. They already received PARP or Bev, or some of them are not eligible for many reasons not to receive PARP or Bev. Actually, they have nothing. All they can do is to receive the Platinum. If they respond, they sit and wait without any maintenance or maybe some off-label maintenance sometimes until they become eventually Platinum-resistant. This is the exact patient population we are looking into to start with. Patient in the third or second line, not eligible for Bev and PARP maintenance, and they will receive in our study COM-701 monotherapy. We compare it to a placebo control, including very solid historical control because we know that this patient normally has a PFS of around 5.5 months. Perfect. In terms of interim data that we can expect in the second half of 2026, what should we think of as a successful piece of data? Beyond that, what sort of development plans do you have for the 701 molecule? Yes. This is an adaptive trial design, which allows us a lot of flexibility. In the interim analysis in half to 2026, we plan to have results of 40 patients treated with COM-701, 20 patients treated with placebo. Again, with very solid historical control, we expect to see around six months or 5.5 months PFS for the control. For us, a successful study will be around nine months or three months improvement over placebo. We have a few options. One, since it is a randomized blinded study, if the results are really good, we can discuss with regulators and potentially even go for accelerated approval based on that study. Another option would be, after understanding the magnitude of effect, to start now a phase three study. Another option would be to add combinations like Avastin, to add COM-902 or PD-1s. I think we have a few options to move forward based on a successful readout in 2026. Perfect. Last couple of questions for you, Eran. One is, as investors, what should we be looking out for either from you or from your partners, whether it is AstraZeneca or Gilead? Also, in terms of cash runway, what sort of funds do you have and how long do you think that could fund operations from here? Sure. In Q2 2025, we plan to initiate this randomized adaptive platform trial of COM-701. As said, we are already ongoing now the phase I study of COM-503, or called now GS-0321. In 2025, also, our partner AstraZeneca plans to share early data for a rilvegostomig combination with ADC. Again, the readout of the COM-701 study is going to be in half two 2026. In addition, since the TIGIT, as mentioned before, the TIGIT sentiment out there is quite negative, but we might be influenced with any positive outcome. There are some other potential positive outcomes this year from other companies which are not our direct partners that also could serve as a driver for our own company. In terms of the cash balance, we ended 2024 with a solid cash position of $103 million. With very conservative assumptions of no royalty, no milestones, no other income, this will allow us to go into 2027, supporting all our clinical plans, supporting all our early pipeline. In general, I think we're in a good and solid cash position. Perfect. No, that's a great place to be, especially in this financial environment. Yes, we're grateful for that. Good luck with everything, Eran. I hope to see you soon, if not in New York City, maybe in some of the conferences that we both could be attending. Good luck to you. Thank you very much for joining us and talking to our audience today. Thank you so much. Thank you.