I'm Michael Schmidt, senior biotech analyst with Guggenheim, and I'm pleased to kick off the next Fireside Chat with the Relay management team. With us, we have Sanjiv Patel, the CEO, as well as Don Bergstrom, and Pete Raymer. Welcome, guys, and thanks for joining us. Great. Thanks for having us. Maybe just kicking it off with the high-level questions. You know, as some of the biotech and pharma industry shifts towards AI, implementing AI, novel drug discovery technology, remind us of what makes your platform so unique and differentiated. Yeah, I mean, as you probably know, the word AI is used in basically every conversation now in any industry. Maybe we just start with just debunking some of the myths around it. I think, you know, we see people talk about pushing a button and making a medicine in the metaverse. That's not us. Our general approach to this, making medicines is difficult, and I think one of the things that differentiates us is exactly that. We are, you know, very humble about how to make medicines. Our platform consists of disparate tools, both on the experimental side and on the computational side, and they are both traditional tools and some emerging tools. We apply them, again, in a what we believe to be a unique way, in that our approach is simple, which is we make the traditional approach of making medicines just incrementally better across every step of the process. Obviously, that results in a, you know, much more efficient process at the end. We are a company that focuses on validated targets, we're not one of these companies that's applying the platform to elucidate novel biology. We focus on, first of all, understanding how to modulate the protein in question, and we'll use a range of both experimental tools, you know, obviously creating full-length proteins is difficult to do, as well as computational tools like long-timescale molecular dynamics simulation to understand how to modulate. In the hit finding stage, we'll use traditional hit finding techniques and then combine them with novel computational approaches like machine learning, DNA-encoded library screens. As we get to the kind of optimization of the molecule, we'll use traditional hand-to-hand combat and chemists in the lab, but we'll combine that with computational simulations to get us to the output more rapidly. What really differentiates us? Probably a few things. One, the platform is disparate, consisting of lots of different approaches all combined together. Second thing is the people. I mean, these approaches are very much people dependent. Again, kinda counterintuitive to some of the rhetoric that's out there. We have a very deep experience curve, and the datasets that we've developed over seven years are important and valuable. Look, I think probably the final thing that differentiates, the platform is output. We have 3 assets now in the clinic, a fourth one about to arrive in the clinic, and that's pretty much the thing that differentiates us from many companies that are talking about AI these days. Great. Thank you, Sanjiv. Really appreciate it. Maybe one for Don. Talking about your pipeline, obviously. RLY-4008 is your most advanced. Yeah ... product candidate targeting FGFR2. We've recently seen some very impressive efficacy data in the FGFR2-naive cholangiocarcinoma setting. Just remind us, you know, how does the drug's design play a role in its differentiated profile that has emerged? To what degree can those principles be applied to any other programs that you're working on? Yeah, let me start, then we'll hand it over to Don. I think it's a great example of how we think about the discovery of medicines. We're trying to remove as much biological and translational risk as we can. We saw from the kinda pan FGFR inhibitors that you could see efficacy in the tumor types that were being studied. The dosing and the dose intensity was limited by the off-target effects of inhibiting FGFR1 and FGFR4, as well as obviously, what they wanted to do is inhibit FGFR2. I think the hypothesis was if you could just create a super selective inhibitor, you could get the efficacy, dial out the off-target, and therefore then you could just titrate against the on-target toxicity. Obviously, the industry had been trying this for several decades unsuccessfully. We approached the problem using the platform. We understood the dynamic nature of all the FGFRs, and there was a clear differentiation in FGFR2 that gave us a handle on selectivity. That created the molecule, which is 200-fold more selective for FGFR2 than the associated family members. That gave us what we wanted, which is a very clean off-target toxicity profile. We see virtually very low levels of hypophosphatemia and diarrhea. That allowed us to hit the target, you know, way above 95% of the time for an extended period of time, that gave us the efficacy that we showed last year at the ESMO meeting that we showed of 88% objective rate of response in the population that we showed data on. That gives us a template for how to really go forward, and what you'll see for the PI3K program is exactly that, and you see it, you know, earlier in our portfolio, that if we can make selective molecules against validated targets, we should be able to translate that into clinical efficacy that's meaningful for patients. Great. How should investors think about the commercial potential for 4008 in cholangiocarcinoma and perhaps more broadly in other indications? Yeah. Pete, do you wanna take that? Yeah, sure. You know, the way we characterize it at this stage, still early in the development, is just to talk about a range of numbers, based off of if you were to take the alteration frequency, multiply against cancer deaths within those histologies, and then do the same thing for incidents. If you do that, you come up with across alterations in FGFR2, you have roughly 11,000-35,000 patients per year in the U.S. Within cholangio, that's probably about 1,000-1,500 of those patients in the U.S. You know, the non-cholangio fusions are probably 4,000-8,000. With mutations, you're in the 10,000-15,000 range, and the rest of that being in amplification, so 5,000-8,000. Yeah. All right. Let's maybe talk about some of those opportunities. First, again, circling back to cholangio, where you are now in a pivotal phase II study. Just remind us of how we should think about, you know, enrollment, pace or timing of the trial, and then the FDA's been more actively managing confirmatory phase III trials these days. How do you think about that? I would say, first, we've guided to having the pivotal study fully enrolled by the second half of this year. We, you know, continue to confirm that guidance. We've had an end of phase I meeting that helped inform that pivotal trial design, specifically around dose selection and size. At that point in time, that was in the spring of last year and aligned on a 70-mg dose once daily and 100 patients in that pivotal cohort. You're, you're right. You know, we continue to be active interpreters of how the FDA is behaving in similar contexts. Probably the closest analog we have for us right now is the futibatinib accelerated approval. They got accelerated approval that was with Taiho in October of last year. The post-marketing requirement for a confirmatory study that they have is to run a randomized study against two doses of futibatinib. We will have our own data and context-dependent interactions with the agency to ultimately determine what our obligations will be, but that's the state of affairs as we see it today. Okay. Great. As we think about the potential of FGFR2 inhibition outside cholangio, you know, I know you've had some responses in breast cancer. You know, maybe talk about how you think about that and what investors should expect to learn perhaps from the full dose escalation data or later this year and again, sort of how do you think about next steps outside cholangio? Want to start with that one? Yeah. Just a reminder, we started the dose expansion cohorts, which are three tumor-agnostic cohorts within non-CCA fusions, non-CCA, non-CCA mutations, and non-CCA amplifications. In our dose escalation experience across doses, we had very few, if any, patients actually treated at the 70 milligrams once daily dose in those other histologies and alterations, but we saw some early signs of antitumor activity. We do believe that given the profile of the molecule we've seen to date, some of the anecdotal evidence we've seen from other non-selective inhibitors, there is likely going to be an opportunity outside of cholangiocarcinoma. These expansion cohorts are signal-seeking in nature and will be the initial evidence of where that opportunity might be, and it would inform further development from there. You know, we've seen the, the most probably obvious other opportunity a priori is probably across fusions. We've generally seen in targeted oncology that fusions tend to be histology-agnostic in terms of the driverness of those alterations. I think as you get into mutations and amplifications, that's where we've seen historically that it tends to be more context-dependent in specific diseases. Yeah. We'll show this data in the second half of this year. We would anticipate it being a relatively robust number of patients with nominal follow-up, and it would be able to start to demonstrate where those other opportunities could be. Right. I guess mechanistically in the non-cholangio, among the fusions, I'm sorry, among the mutations and amplification patients, are there any particular tumor types that are, you know, mechanistically thought to be more responsive? Or is it really open at this point? I think it's fairly open at this point, Michael. You know, one of the questions that we'll be looking to answer is what is the context that determines monotherapy sensitivity to FGFR2. We've shown, as you mentioned, a response in a breast cancer patient from a dosage escalation experience. I think as we look historically at the development of precision oncology agents, you've seen that in the cases of mutations or amplifications, there can be tumor context dependence. Think back to the RAF class where you saw activity in RAF mutant melanoma patients but not colorectal cancer patients as monotherapy. I think, you know, as Pete mentioned, these ongoing expansion cohorts are signal-seeking. you know, we believe based on what's been reported anecdotally, the limited experience we have from dose escalation outside of cholangiocarcinoma and from our preclinical data that there are opportunities outside of cholangiocarcinoma. I think we need to define whether there are specific either histologies or genomic contexts that are more sensitive to FGFR2 inhibition. Yeah. All right. Great. Maybe switching over to your PI3 kinase alpha inhibitor program, which, initially, you know, your most advanced compound there is RLY-2608. You have guided to initial data in the first half of this year. You know, maybe just level set in terms of, you know, what's the, how much opportunity is there to improve upon PIK3CA? We've heard a little bit about that at our breast cancer panel yesterday that there's a lot. You know, what degree of improvement are you looking for, I guess, for your compound? I mean, I think it's kind of analogous to what we talked about earlier in the conversation, which is obviously we believe that creating exquisitely selective molecules can provide therapeutic benefit for patients. In this case, the non-selective PI3Kα inhibitors have shown some degree of efficacy, but they've been limited by some of the toxicity, hyperglycemia, diarrhea, rash, fatigue. The hypothesis is that if you could increase the selectivity, create more sustained inhibition of PI3Kα mutant, you could create an opportunity for greater efficacy with a differentiated safety profile that's been limiting some of the pan PI3Kα inhibitors. I don't know if you have any... Yeah. You know, I think you summed it up well, Michael. There's lots of room for improvement. This is significant not just from the safety perspective, but from the efficacy perspective too, because these are two sides of the same coin. Some of the acute toxicities you see with agents such as alpelisib and inavolisib like hyperglycemia ultimately define the upper limit that can be used in dosing those agents. You see issues with chronic tolerability as well, with diarrhea, rash, other AEs that limit the ability to stay on that dose. For example, 300 mg for alpelisib, you know, most patients end up being dose modified. Many patients actually end up discontinuing due prior to disease progression. Ultimately, both of those together are efficacy limiting. This comes back to the hypothesis that if we can design more selective inhibitors, they give us more sustained, you know, more robust and more sustained inhibition of a validated target like PI3Kα. Through leading to a better tolerability profile, you can potentially unlock additional efficacy as well. Right. Then there's, you know, you're obviously not the only company who is working on these molecules. There's compounds out there from Lilly, from Scorpion on the private side. Roche has one that's more advanced. Any comparisons of your drug against those others and any high-level thoughts in terms of, you know, how those might be differentiated? Yeah. Pete, do you wanna cover that? First, this is a very large unmet medical need. We're talking about tens of thousands of patients in the U.S. alone. It's great to see multiple drug developers trying to help solve that for patients. We, you know, I think we're all rooting for each other in making a difference here. As we think about trying to compare these molecules, inavolisib from Roche, probably more in the alpelisib camp, another orthosteric inhibitor, very similar profile. You know, no selectivity between wild-type and mutant and maintaining animal or activity against the other isoforms. As you get into the categories of, you know, Loxo, Scorpion, us, we are taking a different approach, using allosteric mechanisms to try to generate mutant selectivity. In the case of Loxo, they've generated what appears to be a very selective molecule against H1047R, one specific mutation that's about 1/3 of the patient population that we can address. I think we're roughly at the same place in clinical development. We were a little bit ahead. Then on the Scorpion molecule, again, the caveat being we don't know the chemical structures of either Loxo or Scorpion, so we have not been able to benchmark those molecules head-to-head in the internal assays. All we can do is make cross-group clinical comparisons and interpret the data the way you have. The Scorpion molecule looks to be more pan-mutant selective with roughly the same selectivity window that we have just, you know, about a year and a half or so behind. Yeah. As we think about the first upcoming data disclosure from your program, could you just help us set expectations on what we're looking for in the data? Yeah. We've been guiding to this data coming up in the first half of this year, continue to be on track to do so. The I take you back to the progression of data we saw from RLY-4008, the FGFR2 inhibitor. The goal of that initial disclosure a year into the study was to demonstrate clinical proof of mechanism and then get two A dose into a homogeneous patient population at a larger end and start to really define the efficacy profile of the molecule. It's gonna be a similar trajectory here. We're about a year into the dose escalation experience across monotherapy and combo. The combo started just in April of last year, the goal would be in this initial data set to start to demonstrate clinical proof of mechanism. We would define that by, in the context of clinically relevant doses, being able to evaluate the safety and tolerability profile, at least acutely anyways in this early data, and being able to demonstrate that we've been able to dial out some of those off-target toxicities because those are the clinical biomarkers for selectivity in this context. Demonstrating that we're able to get to IC80 inhibition or greater of the mutant while staying below levels of wild-type inhibition that leads to the dose-limiting toxicities that we've seen with the other agents. I think that's the type of data you'll be able to see: acute safety tolerability, PK, modeling, ctDNA, ex vivo PDSA, and then early antitumor activity, which in totality of the story would help to start to give you confidence that over time, at a dose in a larger N in a homogeneous patient population, we should be able to demonstrate meaningfully differentiated efficacy. Can you talk a bit about what types of patients you expect to enroll in the phase I? Yeah. The phase one is running with two parallel dose escalations, a monotherapy dose escalation. That's in all solid tumor patients with a documented PI3K mutation. That's not restricted to any given histology or any given tumor type, not enriched for any patient populations that we think may be more sensitive to single agent PI3Kα inhibition. That's a cohort that will largely be informative for establishing PK, PD, and safety profile. We do have a breast cancer dose escalation running as well. That's a combination with fulvestrant that is focusing on patients with PI3K mutation with any PI3K mutation with breast cancer that's previously been treated with a CDK4/6 inhibitor, but has not previously been treated with PI3Kα inhibitor. As these are all hormone receptor-positive patients that are being treated with fulvestrant as the combination agent. Will we see data from the combination as well, or just the monotherapy cohort? Both. Both. Okay. You'll see data from both dose escalation portions of the study. Right. Maybe remind us how you think about the sort of next steps or the even a path to registration longer term for 2608? Okay. No one, Don. Yes. you know, I think there are a couple of different paths that could be available for 2608. There could be single agent precision oncology paths forward with well-defined patient populations with high unmet need, where in the late line single agent sensitivity to PI3K with a, you know, measurable response rate could give you the opportunity to continue single arm development to be able to support an accelerated approval. In the breast cancer setting, it's likely that the path that you'd focus on would likely be in the setting of a combination, potentially with fulvestrant or with an anti-estrogen and CDK4/6 inhibitor therapy to move earlier into the treatment settings. We think, you know, we will further define the path forward, in the breast cancer setting in earlier line patient populations as we continue to develop the agent. Okay, great. Great. Maybe switching over to the rest of the pipeline, you announced last year a few new compounds moving towards, you know, IND. You have a CDK2 inhibitor, as well as an ER alpha degrader. Maybe just talk about the opportunity for these, and again, where you see, you know, where you see differentiation potential for those drugs. I mean, just taking a high-level step back, we're trying to create, you know, a sustainable franchise and not seeing it kinda asset by asset. As you can see with our PI3K portfolio, 2608, you know, is the first of our pan-mutant PI3Kα inhibitors. 5836 will go into the clinic in Q2, then we obviously have a H1047R specific PI3Kα inhibitor. Obviously this is our foundational asset, what we're trying to do is build a more comprehensive holistic therapeutic solution for patients. That's why you saw us announce the CDK2 and the ERα. The goal of those is to be used in combination with our PI3K, and obviously we have unnamed programs and the general approach there is increased selectivity, which you can see is a theme now in our programs should allow us to generate, you know, a more tolerable safety profile, which over time should then translate into greater efficacy, and in this case, greater combinability, which is one of the challenges for the patients that we are trying to address. I think what we're not trying to do is create truly differentiated. Obviously that if that is how it plays out, that's how it plays out. We're trying to create here a franchise of molecules that go together as a holistic solution for patients. Yeah, that makes sense. It sounds like a breast cancer focus for sure. Yeah. You talk about your PI3K kinase inhibitor portfolio. How should we think about your next-gen pan-mutant inhibitor? Yeah ... relative to the lead product? Where does the selective drug fit into sort of the longer development path for. Yeah for the group? Maybe, Don, you can take the profile, and then Pete, I mean, you can just take how they fit together. Yeah. RLY-5836 was designed really with chemical diversity in mind. It's, you know, I think best practice in drug discovery that as you develop, you know, an interesting molecule, you have a structurally distinct backup to be able to protect against any unpredictable risk that you may run into the clinic. What you see companies typically do is sit on their backup. In the case, in our case, as we're advancing RLY-2608 towards the clinic, given the magnitude of the opportunity for Relay Therapeutics and the magnitude of the unmet need for patients, we felt it wasn't prudent to sit on a backup, and we essentially developed it in parallel, put it through non-clinical development actually before we ever started dosing patients with RLY-2608. It gives us the opportunity to really be able to have two shots on goal, if you will, for a mutant-specific PI3Kα inhibitor. RLY-5836 profile, again, it's chemically distinct from RLY-2608. It's got some different metabolic properties, but otherwise has a very similar profile to RLY-2608 with regard to mutant selectivity, both biochemically and in cells, and then ultimately in animals. Shows a very similar profile, being able to achieve full anti-tumor efficacy in PI3Kα-driven models without seeing the associated hallmarks of wild-type PI3Kα inhibition like hyperinsulinemia. Great. I think the long-term strategy here, Michael, is we're trying to create the best PI3K molecule for patients, ultimately maximize the efficacy you can with this mechanism. That might be done with RLY-2608, it might be with RLY-5836, it might be with H1047R or the ones behind that. We'll stop once we feel like we've maximized the efficacy for patients. Sounds good. With that, I think it's time to wrap up. I guess we're looking forward to seeing phase 1 dose escalation data for FGFR2 and PI3K in the first half, and then expansions in from both in the second half, correct? Expansions for RLY-4008 specifically. The RLY-2608 data in the first half is all we've guided to thus far. Got you. All right. All right. for the PI3K alpha program. Well, looking forward to that. Thank you so much, guys. Thanks very much. Excellent. Thank you.