Relay Therapeutics, Inc. (RLAY) Earnings Call Transcript & Summary

Thank you so much. Good afternoon. Welcome to Barclays Global Healthcare Conference. We're in Miami and do e-mail us, so get us on Bloomberg, if you have any questions. My name is Peter Lawson. I'm one of the SMID-cap biotech analyst at Barclays, focus on oncology-related names and really delighted to have with us from the management from Relay Therapeutics, Sanjiv Patel, President, CEO; Don Bergstrom, President of R&D; and Peter Rahmer, Chief Corporate Development Officer.

And I guess first question just with kind of pertinent kind of looming data in the PI3K alpha space, kind of what should we expect to learn from this data update? Is it something investors should be laser-focused on? Is it something that's going to kind of help enrollment or is it more kind of investigator-related data you think?

Well, thanks for the invitation, Peter, the opportunity to get out of what was a very cold and windy Boston. And so we're thrilled to be. In terms of the data on PI3K alpha mutant selective, I think we're excited about sharing it at the AACR meeting that's upcoming in Orlando. And yes, I think it will be informative. As you know, this is one of the largest unserved precision oncology unmet opportunities out there. And given the number of patients that could benefit, I think the data will hopefully be informative. As you know, this target has been served by the nonselective PI3K alpha inhibitors and the dose intensity for those therapies is limited by their off-target toxicities. And so the goal of this program we set out to work on it in research a few years ago was to try and dial out those off-target toxicities such that we could maintain the dose intensity, and that would then translate into a step change in efficacy. And so this first data disclosure, we've been guiding towards for the last year is the first in-human disclosure. The goal is to be able to demonstrate PK, PD, safety and then early signs of efficacy and maybe hand it back to Pete and just talk about some of the expectations that investors could expect.

Yes. And I think it's just as Sanjiv suggested, the goal of all our programs generally the true line is that using traditional tools, there's a lot of efficacy being left on the table for some of these targets. And by using our unique platform, we tend to be able to gain insights that allow us to drive to greater selectivity and therefore, greater tolerability and safety. And then because of that, it can eventually drive to greater efficacy over time. So this initial disclosure similar to our 4008 disclosure for our FGFR2 inhibitor when we first showed data. It starts to answer the first part of the question, which is, have we been able to translate our preclinical findings of having developed a more selected PI3K alpha pan-mutant selective inhibitor into early clinical data? And you can see that because these off-target toxicities, you can measure those and start to see at clinically relevant doses, are those meaningfully differentiated from what we've seen with alpelisib and inavolisib. And you can tell for clinically relevant doses because by our PK modeling and ex vivo PDSA and early antitumor activity, which acts as a pharmacodynamic biomarker and the data sets, we can start to understand if once we are able to choose a dose and in a larger end, will we be able to drive to greater efficacy.

And so this update, what the key thing here is to show safety and that you don't get this glucose spike that's essentially cured the PI3K class for those physicians?

Don, if you want to take that?

Yes. So the elevated glucose is certainly part of the toxicity question. And that is the toxicity you see with the nonselective PI3K-alpha inhibitors that is most directly related to inhibition of wild-type PI3K alpha outside of the tumor. But when you look at the overall tolerability profile, it goes beyond just hyperglycemia. And you see AE such as diarrhea and rash that with hyperglycemia are the 3 most common toxicities that lead to discontinuation of alpelisib. And we think that these are AEs that are also due to inhibition of wild-type PI3K alpha, but also potentially due to some of the residual activity that drugs like alpelisib have against PI3K delta and other family members. So as we're looking at the safety profile, certainly looking at glucose is an important marker for us for proving this selectivity hypothesis. But I think for being able to look at the profile of 2608 in light of the future development of the asset, you have to look across the totality of the safety data and really assess whether we're able to generate a safety profile that's meaningfully differentiated across the board.

How do we kind of kind of understand that collection of safety data points? I mean, I guess, longer run, it's going to be discontinuation rates if we look at. But for this set of data, what's the best way to triangulate a balanced diarrhea, rash, glucose spike, what's good, what's bad, what -- how do we fit into those quadrants?

So I think you're looking, first of all, just in terms of key toxicity, what's the overall rate of Grade 3 and Grade 4 AEs, and then you're looking specifically at those 3 key, AEs as I just mentioned, hyperglycemia, diarrhea, rash, what is the rate that we're seeing with 2608. And then while our data will be early, we can start asking a question of whether you're seeing the ability for patients to stay on drug? What's the rate of dose modification or discontinuation.

Do you think you've taken a hypoglycemia down to kind of grade 1, grade 2, or do you think it's too early to determine that yet?

I think you got to look at it in its totality. Obviously, the goal of this program is to try and get a step change in efficacy. And ultimately, the regulatory hurdle is going to be PFS. And so I think, obviously, the goal is to try and maintain as much dose intensity as we can, but I think it's a kind of multiparameter optimization question in the end. But yes, the goal is to try and reduce it down to as low as is possible. And obviously, that will then translate in the dose intensity, which we hope will then translate into efficacy. Because you can -- like you can't just focus on just hyperglycemia. You can in the context of understanding the translation of our selectivity window because that is the most translatable toxicity associated with PI3K-alpha wild type. But we have a good analog for what happens if you held all other safety and tolerability of the same, but just lower hyperglycemia in the AKT inhibitor from AstraZeneca captive. That's exactly what that agent generally did has a very similar safety and tolerability profile to alpelisib, except for a lower hyperglycemia and a higher diarrhea rate and has resulted in exactly the same efficacy of our policies. That is not our goal. Our goal is to have as Sanjiv has said a couple of times now, a meaningful difference in efficacy for patients with PI3K-alpha mutations. And so that's why it's truly the totality of the safety profile that will.

Yes. So we should really be looking at all 3 of those safety profiles being nudged down.

I think -- look, we're -- your comparator data sets coming from Phase II data sets from alpelisib and inavolisib or large Phase Ib data sets and a recommended Phase II dose. And we are obviously going to have a much smaller end coming from the dose escalation patient data set. And so I think you have to be able to look at our data set compared to that proportionately adjusted for the difference and, and have confidence that once we get to a dose in a larger end that, that will hold up. So again, I don't think we're looking for incrementalism here, and that will probably be the ones in which we try to look at this initial safety and tolerability profile.

Got you. Okay. Just if you could remind us the number of patients, we will we see recommended Phase II dose ACL?

We haven't guided to the specific number of patients that we'll see. It will come from both the dose escalation monotherapy dose escalation in combination dose escalation. And we've said that the goal is to have a robust enough and to be interpretable against all these parameters that we just discussed. And in terms of a recommended Phase II dose, we -- that was not a hurdle for making this disclosure. So -- and if and when we choose and recommend a Phase II dose and move into expansion, we would announce that. And so by proxy, we have entered is not a requirement at this disclosure.

That's really helpful. And just -- I guess, when we think about these patients going into the study, are there any that have kind of background level of hypoglycemia that you're kind of battling against?

I mean it's a western population. And as you know, given the demographics, unfortunately, it's just a fact of life that there is a background level of grade 1, grade 2 hyperglycemia in the populations that are being studied in any clinical trial. And so Don, if you want to specifically comment on that.

Yes. I think our trial, we are excluding patients who have uncontrolled diabetes from coming on the study. But we will allow patients on study who have a hemoglobin A1c up to 7%. So we will allow patients in the study who have borderline glycemic control.

What do you want to see whether it's this data set or continuing data sets this year went to move the drug forward?

I mean I think we've been consistent around this since the start of the trial last year, which is we want to be able to show that we can sustainably maintain a dose intensity above the IC80 and do that -- to do that, obviously, you need to have -- be able to leave the target. So we need to be able to demonstrate that through PD. And that then is in the context of differentiated safety profile. And I think if we could we get confident around those 3 things, then that would give us confidence that if we could then enroll a larger number of patients in the defined patient populations that we're looking at and find a recommended go-forward dose, you would start to see the step change in efficacy that patients require.

And at AACR, would we see kind of the dose escalation portion that's in combination with fulvestrant as well?

Correct. We'll show both the monotherapy dose escalation and the combination dose escalation data. And for reference, just to make sure everyone knows, the monotherapy dose escalation started in December of 2021. The combination with fulvestrant started in April of 2022. It was when we started that combination when we guided to data in the first half of this year because we are very focused on making sure we had about a year's worth of experience in that combination cohort because obviously, that's where the breadth of the data exists with [ price of the ] inhibitors. And so that's where you're going to be able to make the probably the most robust cross-trial comparisons.

Okay. Yes. And so how do we think about the bars in each of those 2 groups, essentially, dose escalation, et cetera?

The monotherapy dose escalation, that is simply a part of the study that allows us to understand dose and safety and tolerability. The -- that is an all-comer solid tumor, obviously, with PK mutations, but not enriched for tumor types in which we apriority would suggest or think would have -- we'd be sensitive to single-agent PI3K-alpha mutation inhibition. For example, colorectal cancer, non-small cell lung cancer, in those tumor types, PI3K alpha is commonly co-mutated with KRAS G12C. And so a single agent in those settings, we wouldn't presume would be effective by themselves. And so in the combination, though, we are going into a patient population in a regimen where there is a good amount of efficacy data from the present inhibitors. And so these early data sets at clinically relevant doses will give us a pretty good hint into the profile of 2608 in that context. All you're hearing is caution against is that efficacy data specifically in this setting, but in really any person human data set a year into the study is hard to interpret. And you really need to get to a recommended Phase II dose in a much larger patient population before you can start to really define that.

How many doses have you've been through so far?

I haven't disclosed that. But suffice to say, we've said that the -- we were going to be starting the combination dose escalation with fulvestrant once we've reached a biologically active dose in the monotherapy dose escalation, we define that to be crossing over the IC50 mutant inhibition. So fair to say, we at least covered that as we started to start the combination. And then we likely had some dose escalating to do to get to clinically relevant doses.

Would we get durability data as well?

We will disclose what the median duration of exposure is at that time, but inevitably, it's going to be small. But there's not going to be a lot of follow-ups, especially at the clinically relevant doses.

Got you. And do you want to see, I guess, as this data pans out, do you want to see a step change in efficacy as well as kind of a step down in safety as well?

That is essentially the goal of the program. It is only one goal, which is efficacy. And so that's exactly.

Now we think the way you get there is greater dose intensity. How you get to greater dose intensity? You have an agent that's more tolerable, which patients can stay on for longer periods of time. So the trade-off will be -- if you can provide an opportunity for an agent that is more tolerable that covers the target for -- at a higher level for a longer period of time, it should drive to greater efficacy.

Got you. Okay. Is there -- do you have a sense from physicians, your own internal goals, if like the percentage of Grade 3 events that -- which would kill the drug or you feel it's acceptable to move the drug forward?

The existing inhibitors, alpelisib and inavolisib sit somewhere between depending on the study, 40% to 70% Grade 3 adverse events. As Don said in the opening, the hyperglycemia rash, diarrhea are clear toxicities leading to discontinuation. All these things are contributing to the lack of the ability to maintain dose intensity with these agents, which is leading to a lack subpar efficacy. So I don't think there's one particular what you're going to say it's at 30% now and so 5% grade 3 is okay. It's the totality of the safety and tolerability profile, I think that we really need to meaningfully differentiate that.

Okay. And as we think about expansion opportunities, where else is there good rationale not trying to discredit the scale of the breast cancer opportunity, just where else?

Yes. So as we're designing the first human protocol for the monotherapy cohort specifically, we tried to design cohorts that we felt based on our understanding of the genomics of the tumors, our preclinical data and the experience with PI3K inhibitors in these patient populations to try to identify cohorts where we thought there'd be the greatest likelihood for single agent activity. And the 4 patient populations we identified consist of 3 patient populations that are sort of conventionally defined. So clear cell ovarian cancer with PI3K alpha mutation, squamous head and neck cancer with the PI3K alpha mutation and cervical cancer with the PI3K alpha mutation. But then there's a fourth population that we defined as well, which is a pan histology population that is defined by having 2 mutations in PI3K alpha. And this is a phenomenon that was first described by Jose Baselga about 6 years ago when he was still at Memorial, where they identified patients who had 2 PI3K alpha mutations that are actually in the same copy of the PI3K alpha gene. And that leads to a significantly hyper-activating PI3K protein that then renders the tumor very sensitive to PI3K alpha inhibition. That is estimated to represent about 10% of the overall PI3K alpha patient population and represents the fourth cohort that will enroll in their expansion cohorts in monotherapy. Now as Pete alluded to, we did not, as we're performing monotherapy dose escalation, perform any enrichment for just those patients. And we have other patients we anticipate will be represented in the monotherapy dose escalation as well, not just from those 4 patient populations.

Got you. And I guess the other question we keep getting is around scopings molecule in the space. How are you differentiated? And how do you do those kinds of cross preclinical explorations of differentiation?

Do you want to take on, Pete?

Sure. So the 2 main next-generation inhibitors, if you will, in the space come from Loxo, Lilly and the other one from Scorpion. The Loxo Lilly molecule from our understanding, interpretation of their preclinical data is specific to H1047R -- so just one of the hotspot mutations makes up anywhere from about 30% to 40% of the patient population, depending on what tumor type you're looking at. So that one enters the clinic early after we did, but addresses a minority of the patient population. Unclear to us is we think that as we contemplated what mechanism to bring forward to fully serve this patient population, we kind of had both options at our disposal because we started with full knowledge of the target and having solved the first full length structures of both wild-type and mutant form on the protein. We are understanding of a pocket that would allow us to get H1047-specificity and announced our pocket that we discovered that allows us to get pan-mutant selectivity. And as we progressed through our decision-making as to which mechanisms bring forward, we decided to bring the pan-mutant selective forward first. Now we do have an H1047R molecule also in discovery, in addition to multiple pan-mutant selective molecules. And that is because of the size and importance of this opportunity is too great in the depth and breadth of our knowledge here. It is so large that we will continue to prosecute inhibitors for this target until we definitively define the best efficacy for patients clinically. The other molecule from Scorpion is -- looks like to be a pan-mutant selective inhibitor that looks and feels a bit more like our pan-selective inhibitors, it's just 18-plus months behind.

Is there anything to triangulate from their preclinical data, I mean when you look at the molecule and…

We don't -- neither of them, neither Loxo or Scorpion nor us frankly, have published the structures of our molecules. And so none of us can do head-to-head comparisons in the same assays to really start to give you a more in-depth understanding. It's all just conjecture and interpretation of each other's preclinical data.

Got you. I guess the other very exciting story is of course FGFR2 inhibitor and kind of data in the second half. What should we expect to see from the, I guess, the non-cholangio expansion cohort?

Well, I mean, you're exactly right, which is obviously last year, we showed what we believe to be pretty definitive evidence around the efficacy of our selective FGFR2 inhibitor in cholangiocarcinoma fusion patients. And obviously, the question there is what is the breadth and depth of the opportunity outside? And I think in the second half of this year, we will share data that we'll start to answer that question, and maybe, Pete, you can detail the types of data that we will share.

Sure. So we initiated those expansion cohorts in December 21, at 70 milligrams once daily. The target enrollment for those -- each of those 3 cohorts, non-CCA fusions, non-CCA amplifications and non-CCA mutations was 30 patients each. In the second half disclosure, we would expect to have those close to fully enrolled, if not fully enrolled because the goal here is to be signal seeking to start to define areas of further development for 4008. We do anticipate there's likely an opportunity outside of CCA and this data set will start to tell us where to focus our efforts.

Got you. I guess maybe the final question just on the kind of the background AI platform, I mean how much do you rely upon that now is if you kind of migrated from AI platform company into a full development company and what do you do with the AI platform?

So first of all, I mean, the company is built as an experimental plus computational approach. Everything that we do is based around both coming together. And so we focus on validated targets. And so we're not one of these people that are using computational methods to at least state novel biology. We're focused on making medicines. And then we're focused on combining experimental approaches to novel kind of approaches to structure biology with both computation approaches that involve physics-based simulations as well as using machine learning and algorithms to help us make the process of making new medicines, both more efficient and effective. And so we break that process down into many, many steps. And essentially, the computational approaches that we deploy in each of those different steps are all about making the experimental work we do more efficient in helping us answer what is the next question or the next experiment that we need to do. And so that all remains fully intact and the fact that we've now put 3 experimental therapies into the clinic and we're about to put a fourth one into the clinic starts to validate that. And so as we build out the development organization, I think all that does is start to test the biology of these molecules that the platform has delivered. And so the platform gets ever more experienced and then hopefully becomes ever more effective, producing molecules against the hypothesis that we're trying to go against. And I think the development organization is testing the biology. And so I think they are related, but one doesn't change the other.

Perfect. Thank you so much.

Great. Thank you.

Pleasure talking to you.

Thanks again for the invitation to get back into the thought. Thank you.