🎁 Black Friday: 50% off Unlimited and 30% off Pro annual plans
IDEAYA Biosciences, Inc. (IDYA)
NASDAQ: IDYA Β· Real-Time Price Β· USD
25.89
-0.28 (-1.07%)
Nov 20, 2026, 12:34 PM EST - Market open
← View all transcripts

R&D Day 2026

Nov 16, 2026

Summary

The pipeline spans PRMT5, KAT6/7, and CDKN2A programs, with early clinical signals and combination strategies targeting tumor drivers and resistance. Multiple IDE892 and IDE574 updates are planned for 2027, while IDE101 combination data may follow in 2028.

Yujiro Hata
CEO, IDEAYA Biosciences

Good afternoon. I wanted to welcome everyone to sunny London. Thank you for taking time out of your very busy schedule. We have a lot we are going to cover today, and so thank you again for the time this afternoon. Our presenters today, Dr. Frank McCormick from UCSF. I will also be presenting the introduction portion. We will also have Dr. Michael White, our President of R&D, and Dr. Theo Ross, our Chief Development Officer. For today's presentation, we will cover several different aspects of the pipeline. 2026 has been a year of significant importance for IDEAYA. This was the year where we filed our first NDA for darovasertib. We also presented as an ASCO oral presentation the top-line results in the first-line metastatic uveal melanoma setting.

In addition, we did note recently that our plan to advance our second program, our DLL3 ADC program, into phase III studies, both in the small cell lung cancer setting, as well as neuroendocrine carcinoma. Today, our focus will be on our third driver in our clinical pipeline, and that is in the MTAP deletion CDKN2A setting. People may have seen earlier today, we did announce the announcement of our development candidate, IDE101, which Mike will profile in depth in terms of why we are excited about this molecule, why we believe it is a key enabler in a very key genetic setting called CDKN2A, which we believe will enable key combinations, both with PRMT5 as well as pan-RAS. In addition, today, we will provide a clinical program update on PRMT5, IDE892.

Here, we will show the first time we are sharing both safety data, PK data, as well as human pharmacodynamic data.

Next, we will also cover IDE574. This is our phase I/II KAT6/7 inhibitor that we are now evaluating for monotherapy, including indications such as breast cancer. We will also show some early data as it relates to why we believe this combination is potentially compelling, specifically with pan-RAS inhibitors in the area of KRAS CRC. For those that have evaluated pan-RAS and potential resist mechanisms, we know that roughly half of the resistance mechanisms that have been associated with pan-RAS are not genetic mutation-driven. In particular, related to a specific bypass mechanism that is connected to the Wnt/MYC pathway. For those that may be familiar, RAS is specifically in the MAP kinase pathway. Mike will walk through in depth of why we believe these two pathways converge and where we believe this creates a therapeutic opportunity for KAT6/7 to be combined with pan-RAS, specifically in the indication of KRAS CRC.

As I mentioned earlier, CDKN2A, here, there is going to be several key pieces that we will highlight. One, for those that may be less familiar with CDKN2A is considered one of the most important tumor suppressor genes in cancer. As frequent and as important as other tumor suppressor genes, several you probably are aware of, including P53. The reason for this is the critical role of several key tumor suppressor proteins are impacted specifically by CDKN2A, which we will cover here in a little bit, including P14 and P16 as well. As we look forward and as we think about some of the key data that was presented in 2026. 2026 was really a remarkable year, not just for pan-RAS, but also for PRMT5. We saw with pan-RAS, obviously key data that was presented at ASCO, where we saw doubling survival in the refractory setting of pancreatic cancer.

We also saw in 2026 the first proof of concept data combining pan-RAS with PRMT5, generating what we believe is key first clinical proof of concept that we believe may create a potential registrational path forward in pancreatic cancer. Now the CDKN2A portion comes into play, which is the connection between MTAP and CDKN2A. CDKN2A and MTAP are both located on the 9p21 chromosome, and when MTAP is deleted, CDKN2A is almost always co-deleted. What is the significant importance of that? First, we know that, as I mentioned earlier, CDKN2A's role as a key tumor suppressor gene, but we've also generated data, which we're going to show to you today, which is one of the most significant combination synergy effects we've actually observed is with this mechanism of CDKN2A and pan-RAS.

We believe this is partly linked to, which Mike will go through in detail, two tumor suppressor proteins that we've seen impacted when you combine these two mechanisms together, and that's related to P21 and P27. This data has never been presented or shared before, period, in any forum that we're aware of, but we think this is critical data that really supports the mechanistic rationale of this combination. Why should we all care about this? The reason why we should care is because CDKN2A deficiency, which is defined by deletion, methylation, and mutation, is prevalent in roughly 70% of pancreatic cancer. Now you can imagine a situation where we combine this agent with pan-RAS, address the majority of that patient population, and as I noted earlier, MTAP and CDKN2A is almost always co-deleted.

Now we can use this as a plug-and-play combination across any MTAP-deleted solid tumor indication with these two now wholly owned assets. With that, we will then proceed to the next section. That provides sort of a high-level summary and hopefully a couple of key touchpoints that you're going to hear across this presentation today. When you take a step back on what is driving our strategy and combinations in the area of pancreatic cancer, CRC, and lung cancer, specifically with the genetic settings of MTAP, CDKN2A, and KRAS in mind, it's really driven by these two core areas. One is this concept of tumor heterogeneity, which as we know today, that cancers, in particular pancreatic cancer, KRAS, yes, we believe is a key driver oncogene in that indication.

However, these patients have other significant genetic mutations that we believe are driving those cancers, specifically MTAP deletion and CDKN2A. The targets that we'll focus on here as part of this discussion will be largely focused on PRMT5 and CDKN2A, and the key genetic mutations that we'll cover, as I noted earlier, will be focused on MTAP, CDKN2A, and the KRAS setting. The second area is around this concept of tumor plasticity or adaptive plasticity. Essentially what this really refers to is changes in the tumor that are causing resistance to pan-RAS specifically that are not driven specifically by genetic mutations. As I mentioned earlier, we believe there's a significant interplay between the MAP kinase pathway and RAS and the Wnt/Myc signaling pathway, which we believe creates this therapeutic opportunity with KAT6/7, and we'll go through that data as well.

So as you can see, in 2026, we believe we have now built one of the broadest pipelines in precision medicine oncology today across several key areas, including IDE892, our PRMT5. We will cover here why we believe we have a potential best-in-class molecule. As noted earlier, we will cover several key parameters, and really the three critical product profiles that we believe are important for first-in-class agent. First is around drug-drug interactions, in particular around CYP3A4. So we will share some data, including some early clinical data that we have in combination with a CYP3A4-mediated agent of IDE397, our MAT2A inhibitor. Second is around the importance of having a non-brain-penetrant molecule.

We believe because of the mechanism and role of both PRMT5 and MAT2A, specifically in RNA splicing, is that you do not want that activity specifically in the brain because of the known role of RNA splicing in cognition specifically.

Then the third component is around MTA cooperative selective binding versus SAM cooperative binding, and ultimately we believe where this will translate is what type of therapeutic window can we generate it specifically as it relates to myelosuppression. So Dr. Ross here will cover the safety data to hopefully establish that we believe we are seeing a very clean safety profile as well. Next, for KAT6/7, we will also share, as I mentioned earlier, some early clinical data. We will share early safety data. We will also share some PD data as well as some case studies for several of these programs. So you will get to see some early clinical efficacy data across these programs. The last part that I will mention, which we will also have a short section on, is around how we are implementing AI in our drug discovery process.

And we have utilized this both for our CDKN2A program, KAT6/7, as well as what we are doing in the next generation oncogene area as well, which Mike will cover here shortly. So one of the key concepts that we want to cover in today's R&D presentation is what we believe could be potentially a new patient treatment paradigm as it relates to how we think about rational combinations in precision oncology. And this is the idea to not just hit the key driver oncogene, but through the combination also addressing the key tumor suppressor loss gene. As noted earlier, we have now clinical proof of concept about combining both a RAS inhibitor with a PRMT5 inhibitor in the MTAP/KRAS indication of PDAC, which we believe represents about 40% of that population.

Then we will now showcase a second example of how we are targeting both a key tumor suppressor loss, this time CDKN2A, in combination with a RAS inhibitor. And we think this continues to validate this hypothesis of potentially a new way to really pursue combinations to drive benefit to these patients. So a lot of this I have covered already, but this, I would say, hopefully highlights the depth and the breadth of the combination activities that we are pursuing. As you can see, we have extensive efforts across PDAC, planned activities across also non-small cell lung cancer and CRC. As people may be aware, we have two collaborations that are ongoing with Roche, one with our pan-RAS inhibitor, specifically to treat MTAP deletion PDAC patients, and then more recently, a combination with KRAS G12D to target KRAS G12D pancreatic cancer patients.

The first combination, we anticipate we will start dosing our first patient by the end of the year, and then our KRAS G12D should be dosing patients by our anticipated goal is early next year. Look for us to continue to push these combinations forward, both through wholly owned combinations that IDEAYA controls, as well as more collaborations that we anticipate forming through biopharma partnerships. With that, we will go ahead and introduce our next speaker, Dr. Frank McCormick. Dr. McCormick does not need much of an introduction, has been truly one of the key pioneers in RAS research for many decades. I have known Frank for a long time, and I have to say, I think this is the first time I have seen him wearing a tie. Thank you, Frank, for doing that.

We really appreciate your time and hearing your insights both in what the opportunities in RAS have been and also what are the future key challenges as we look forward into the future.

Frank McCormick
Professor, UCSF

Thank you very much for the introduction. It is great to be here with all of this team. Good afternoon to all of you from rainy and cold San Francisco. I will speak to someβ€” It is a beautiful switch, I am sure you all will agree. Switches back and forth from the off state into the on state, depending on whether it is bound to GTP or GDP. This protein does not present any obvious pockets to which a small molecule combined until very recently. The technology was just not there to find drugs that hit RAS proteins directly. Next slide, please . Back in 1992, when we started Onyx actually to target this pathway, the pathway looked like on the left. We have RAS proteins cycling between the off state and the on state.

In the tumor cells with mutant RAS, the protein is resistant to gaps, which normally turn RAS off, and therefore signals downstream persistently. The major output of this pathway that is relevant in cancer is the increased expression of cyclin D1 and related proteins, which activate CDK4 and CDK6, push cells into S phase, activate CDK2, and get cells through the S phase of the cell cycle. In normal cells, this is a very precisely regulated process, and when the cell cycle is over, everything resets to normal, to zero. But in cancer cells where the pathway is on persistently, you get accumulation of these cyclins causing inappropriate activation of CDKs and contributing to progression of a disease.

Although CDK4 and CDK6 inhibitors have been in the clinic for quite a long time, including one we developed called palbociclib, Mike and I really feel there was a huge opportunity here to take advantage of our new understanding of this part of the cell cycle and how it is misregulated in KRAS cancer cells to find better ways of intervening in the CDK area of the cell cycle. Back in 1992, we developed sorafenib to try and target RAF kinase to block RAS signaling, and you can see in that pink box on the left there, those are all the RAF inhibitors that have been tested in human tumors. None of them work in KRAS cancers. MEK inhibitors haven't worked either, nor have ERK inhibitors. RAF inhibitors don't work because, as shown in that little cartoon on the right, RAF inhibitors actually paradoxically activate the RAF kinase.

RAF kinase is kind of booby-trapped so that when you try and inhibit it actually makes the complex more active. MEK inhibitors and ERK inhibitors are just too toxic. So far, all efforts to target the hyperactivated MAP kinase pathway in RAS cancers have not been successful. Next slide, please, Mike. Mutant RAS proteins do activate a second pathway, the PI3 kinase pathway. They are not supposed to do that. They don't do that in normal cells. That is the job of RAS-related proteins, as shown here. But when mutant RAS proteins such as G12D and G12V, the two of the most common alleles of KRAS in human cancer, actually have an increased binding to PI3 kinase alpha, and in combination with high expression and the activated state of these proteins, they can cross over and activate the PI3 kinase pathway inappropriately.

This leads to hyperactivation of mTOR and other downstream pathways. This is relevant in this discussion because activation of mTOR affects the expression of ornithine decarboxylase and SAM decarboxylase, two rate-limiting enzymes in polyamine synthesis, which actually sensitizes cells to PRMT5 inhibitors, as Mike will show you in just a moment. We can exploit the hyperactivation of PI3 kinase alpha in new ways to take advantage of that inappropriate activation by mutant RAS. Next slide, please. Up until 2013, efforts to target the MAP kinase pathway had been unsuccessful. The PI3 kinase pathway is still a work in progress, but we had no other way of targeting KRAS in our toolbox. Until in 2013, Kevan Shokat at UCSF found a way of targeting the G12C allele of KRAS, as I am sure you all know. This is a brilliant idea because it killed two birds with one stone.

First of all, he took advantage of the cysteine residue, which is obviously the mutation in G12C, to make covalent compounds that bind to this residue. That gets around the affinity problem, and also these compounds are specific for the mutant allele, so that gets around the specificity problem. It also opened up a new pocket which hadn't been seen previously, the switch II pocket, where that drug binds in this picture, also known as the Shokat pocket. Now there are 28 knockoffs, essentially, of Kevan's original idea in the clinical pipeline, mostly targeting the off state of G12C. On the right pie chart, you can see the red segment there shows that G12C is the most common allele in non-small cell lung cancer because it is the target of the hallmark exposure to cigarette smoke.

So that launched a whole wave of knockoffs and drug discovery in that pocket. Next slide, please. Eight years later, Mirati published a paper which described a non-covalent compound that binds in the same pocket with picomolar affinity and with selectivity for the G12D allele, something which we'd have never dreamed of back in 2013 when we started this whole initiative. The off rate for this compound is so slow, it's actually very hard to measure, so it binds so tightly. But it does also bind to the off state of the KRAS protein, which is a liability I'll mention in just a moment. But there are 27 knockoffs, essentially, of this original compound in the clinical pipeline, and they're all progressing through phase I and phase II clinical trials.

As you can see in the little summary there, these two alleles, C and D, have been successfully targeted. V has been much more difficult, and that's because the G12V protein is actually smaller and tighter than the G12D and C proteins. Still a bit behind in targeting G12V. Next slide, please. I'll skip that, Mike, in the interest of time, if you don't mind. We have the direct and the covalent and non-covalent KRAS binders, and then the tri-complex inhibitors, which Yujiro has already mentioned. This is one which has hit the headlines when it got a standing ovation at ASCO when they presented the data on the right, showing almost a doubling of survival in pancreatic cancer.

This is from daraxonrasib, a tri-complex inhibitor, which takes advantage of a cellular protein called cyclophilin to present a small molecule to bind to RAS and then block its function. A very different mechanism of action to the direct inhibitors. This particular drug is rather interesting because it's not even a KRAS-specific drug, it's a pan-RAS drug. This raises all kinds of interesting questions about the role of inhibiting other RAS isoforms in normal tissue and in immune cells and in the stroma, and whether these effects are beneficial or detrimental to the overall therapy, we don't yet know. Also, this drug has the sort of very unexpected property of stimulating hydrolysis of GTP on mutant RAS proteins, which helps turn them off and may contribute to the efficacy of this compound.

But it does inhibit normal RAS proteins, and as I'm sure you're well aware, many of you are well aware, it does have severe side effects. We think that this is a fantastic start, great news for patients with pancreatic cancer, but most likely, I think, the wave of RAS drugs, which are allele-specific, will be safer, better tolerated, and hopefully even better efficacy coming down the pipeline to be determined. Next slide, please. If we sort of very briefly summarize mechanisms of resistance, there are mechanisms which relate directly to the specific compounds. I mentioned that most of the drugs in the pipeline that target KRAS directly inhibit the inactive GDP-bound state, the off state of KRAS, and trap the protein in the inactive state. Unfortunately, tumor cells can get around this quite easily by hyperactivating upstream signaling to push the RAS protein into the on state.

Also, just overexpression of KRAS basically outruns the ability of the drug to trap it in the on state. These are two major mechanisms of resistance for that particular class of drug. The tri-complex inhibitors, at least in preclinical models, have a liability through loss of or mutation of cyclophilin A, and also by increased expression of KRAS. To what extent these contribute to resistance in the clinic is yet to be determined. We will see. For targeted therapies, we always expect to see mutations in the drug-binding pocket. That has been noted in some patients on some of these drugs already. Reactivation of the pathway by mutations in NRAS or downstream mutations have been seen and are expected, but they are kind of fewer than one would have predicted, perhaps.

As Yujiro said, a major resistance mechanism is through tumor plasticity, which has been seen in other types of cancer, but is really a serious issue in KRAS cancers. When in lung cancer and in pancreatic cancer, we see state changes, which make the cells less RAS-dependent and less easy to kill. These kind of state changes and plasticity can, we hope, be addressed using inhibitors of KAT6/7, as Mike will describe, through the interactions with stem-like pathways in tumors. Finally, future directions. Combinations, obviously, of the 98, whatever it is, compounds in the pipeline, all will be tested in combination. Some are already being tested in combination with the obvious partners, cetuximab and checkpoint inhibitors.

I think smarter combinations, such as those with PRMT5 and specific genotypes, and with CDK type inhibitors and KAT6/7 inhibitors, will definitely contribute to the overall response rates and duration of responses in many of these cancers. There will be a new focus on colorectal cancer in the future, but so far, most of the KRAS drugs have focused on lung cancer and pancreatic cancer, where G12C, D, and V alleles are the dominant alleles. Colorectal cancer is still a massive unmet need with really nothing to offer patients that really helps. The first KRAS drug approved, actually in combination with cetuximab, the approval was withdrawn because there is no overall clinical benefit. A lot of work to be done in colorectal cancer. In the U.S., the incidence of colorectal cancer is rapidly increasing, particularly in people under 40 who do not undergo screening.

There is a really big and growing unmet clinical need. One of the big problems with colorectal cancer is that the first initiation mutations in the APC gene activate the Wnt pathway, which makes cells more stem-like. Again, this presents an opportunity for intervention to reverse this effect using the KAT6/7 inhibitors that we will hear about in just a moment. These tumors are also more heterogeneous because KRAS is not necessarily the primary driver, another issue we need to address. Finally, the field is moving towards earlier intervention, partly through advances in early detection, but also stimulated by papers such as the one below, which I would urge you to read, which has been discovered that even in young people, in biopsies of pancreas from young people, you can see mutations in KRAS and related pathways, even in very early stages.

This presents a possibility that we can intervene early in these kind of non-malignant, pre-malignant states in the right combination, the safe combinations, and prevent the progression to the more heterogeneous and difficult to treat tumors later by early intervention. This is definitely a field I am interested in, and the field, I think, is moving in this direction. With that, I will stop and hand over to Mike.

Michael White
President of R&D, IDEAYA Biosciences

Really appreciate that. For those of you in the audience and online, Frank will be staying for the question and answer session, assuming that I talk fast enough for us to be able to get there in the first place. With respect to that, we will get into the PRMT5 program. As Yujiro mentioned at the beginning, we are sitting with unprecedented mechanism-driven combination opportunities in front of us in MTAP cancers. In order to be able to help meet that opportunity, IDE892 was actually designed with intent to be maximizing utility for combination therapy. That means we had to produce a molecule with an exceptional therapeutic window and a molecule that has no drug-drug interaction liabilities.

When we are thinking about this, before I get into the details of those properties, some of which will be shared with you by Theo, and I think I have got both microphones on here, guys, so you might want to switch to the lapel mic. I want to point out three mechanisms that are key underpinning components of our prioritization of the PRMT5 RAS combination partnership that we have with Roche. The first is all the way over here on the left. This was introduced by Frank. This is a key interaction between RAS and the one-carbon metabolism pathway, and that is at the level of mammalian target of rapamycin. mTOR activation by PI3 kinase and RAS is a major adaptive and precision medicine bypass mechanism because it makes tumor cells harder to kill.

However, something that is very unique in the setting of MTAP null tumors is mTOR is actually increasing MTA accumulation because it activates polyamine synthesis. MTA accumulation, of course, will sensitize tumors to an MTA cooperative PRMT5 inhibitor. If you guys could control this mic here, that would be great. This means that a combination therapy. Shall I use this mic or shall I use this one?

I am fine now. Okay. This means that a combination of a PRMT5 inhibitor together with a RAS inhibitor should actually shut off mTOR as a bypass pathway because it flips everything upside down and actually sensitizes the tumors to the MTA cooperative PRMT5 inhibitor. In the middle is another demonstration of the unique context of MTAP loss in the setting of RAS cancers. This is an observation that came from a very large dataset using RNA-seq to profile the response of 50 cancer cell lines to drugs, including darax onrasib . What you can see here is that MTAP null cells have an altered response to darax onrasib compared to their wild type counterparts, and this includes activation of an epithelial to mesenchymal transition, which is known to promote resistance.

Another reason why we have distinct combination opportunities in this particular setting related to this MTAP co-alterations with RAS. Then finally, over here on the far right, I think it is important to remember that the effector pathways that oncogenic RAS uses to transform cells ultimately impinge on transcription factors that drive those regulatory events. That activity, with respect to those transcription factors, is supported by PRMT5 in at least three locations. One is methylation of histones that allow transcription. The next is alternative splicing that allows for production of appropriate protein. Then finally, through methylation of ribosome components, translation itself. We believe that this support of RAS signaling suggests that a combination of a RAS inhibitor and a PRMT5 inhibitor can both shut off pathway activation as well as shut off pathway output. Now, one of the key properties that we were seeking toβ€” Sorry.

One of the key properties that we were seeking to solve with IDE892 in the setting of this combination is to have the capacity to maximally inhibit PRMT5 activity in tumor cells, but not normal tissue. MTAP loss, the synthetic lethal opportunity here is all about MTA. SAM-bound PRMT5 is the active enzyme. MTA is a SAM-competitive PRMT5 inhibitor. To make a tumor-specific inhibitor of PRMT5, you need to lock MTA onto PRMT5 without disturbing SAM-bound PRMT5 in normal tissues. To achieve that profile, we designed an 892 super cooperative binding with MTA, as well as anti-cooperativity with SAM to push it off normal tissues. You can see over here on the far left, in the presence of MTA, IDE892 affinity for PRMT5 increases by 1,400-fold. That super cooperativity is coupled with anti-cooperativity.

We can compare that here when we look at these different metabolites using a fast surface plasmon resonance assay. As expected, MTA addition enhances the binding of PRMT5 with IDE892 versus control. SAM is doing the opposite, anti-cooperativity. So that super cooperativity plus anti-cooperativity delivers exceptionally specific activity in the MTAP null state versus the wild type. The top here, you see no anti-tumor activity, no perturbation of splicing in the wild type MTAP setting. You knock out MTAP, you switch no activity to regressions and strong inhibition of alternative splicing. This profile, as you would expect, delivers very nice combination benefit in multiple models. These are all KRAS mutant models. Each one of these models, we see either partial control or relapse on daraxonrasib monotherapy. In every case, we see deep and durable regressions with IDE892, irrespective of monotherapy activity.

I am going to turn the floor over now to Dr. Ross so that she can explain what we are seeing in the clinic right now, and don't get confused like I did with respect to which clicker actually moves the slides.

Theo Ross
Chief Development Officer, IDEAYA Biosciences

I am actually not going to touch that, having seen your experience. Thank you. Now, for the human experiment, which is the most exciting part, in my opinion, based on really good science, we have a development strategy, which is shown here. It is kind of our map of how we are looking at the PRMT5 inhibitor, MAT2A inhibitors, and the combinations. On the top, you can see the monotherapy escalation and expansion in a phase I, looking for safety, efficacy, exposure, et cetera. On the bottom are our combinations where we are combining with the MAT2A inhibitor, so PRMT5 inhibitor, MAT2A inhibitor. We are in combination now in parallel with our escalation, as shown there.

We are targeting our FPI, as mentioned by Yujiro, for our combination with pan-RAS inhibitor with Roche, as well as the G12D inhibitor. Of course, the IDE101 that will be discussed by Mike coming forward. It is just to come there.

Moving to safety for IDE892, early safety data shown here. It demonstrates potential best-in-class safety. We have 31 patients dosed in five cohorts in escalation, and 58% of those show any treatment-related AEs, 18 of 31 patients, and 6% of those, only two with grade 3 TRAEs. The most common treatment-related AEs are shown in the middle panel there, vomiting, nausea, diarrhea, fatigue, less than 20% for all of those. On the far right-hand side, we have broken it out by dose levels. As I said, we have gone through five dose levels. It started expanding in the dose level 3. We continue to escalate in dose levels four and five. Context for the grade 3 TRAEs, we have two patients. One had a transient neutropenia that upon re-challenge didn't happen, but we still, of course, have to monitor that, and an iron deficiency anemia.

Again, very limited toxicities here, all grade 1 and 2 in general, other than those two grade 3s. No grade 4 or 5 treatment-related AEs, no skin rash, which is very important for our RAS combinations. RAS, not rash combinations. RAS. No CNS-related toxicities, which is important. We will show some preclinical data coming up where there is little exposure in the brain in rats and dogs, and hopefully in humans. Without the CNS toxicity, we think that is the case. No DLTs at this point. We have not reached a maximally tolerated dose. From a safety point of view, so far so good in the human experiment with IDE892. We also have really nice exposure. Shown on the left-hand side is our PK data, human PK data at our first four doses, dose levels one, two, three, and four.

At dose level 3, and if you are color-blind, you probably cannot see it, but there is the top line, which is dose level 4, and then the next line, which is green for those who are not color-blind. You can see that the exposure over 24 hours continues to be above the effective cellular dose in preclinical doses. So exposure increases in a dose-dependent manner, as you can see. Half-life supports QD dosing, which is what we have. As I mentioned, our target EC90 coverage is favorable with QD dosing. There is no evidence of a time-dependent PK, no noteworthy accumulation observed. We have initiated mono expansion in that dose level 3 that is above the EC90.

On the right-hand side, using very preclinical CNS exposure experiments in dogs and rats, you can see at a low dose and a high dose in the dog toxicity study that the IDE892 exposure in the brain, in the CSF, is very limited or not detected at all in the dog. In the rat, only 2.4% is in the brain compared to the plasma. With regards to PD data, which is achieved by measuring SDMA, or symmetric dimethylarginine suppression, you can see on the very left-hand side, serum SDMA levels starting at 0% and going down. So 25% suppression, 50% suppression, 75% suppression, goes down in a somewhat linear way as you go across in terms of IDE892 steady state levels, AUC. The green bar there is the medium percent SDMA change at 200 mg to 600mg QD for the BMS navlimetostat drug.

We are in that range as we go down, and as the dotted line is their maximal change. So we have nice PD, nice data, pharmacodynamic data showing that our 892 escalation, that we have strong exposure to plasma SDMA response relationship. Interestingly, our combination with IDE892 plus IDE397, the MAT2A inhibitor, or the open circles, that is our dose level 1 combo, shows increasing SDMA suppression at a fraction of the IDE892 expansion dose. In terms of activity, in our dose level 3 cohort, we have a patient with a third-line cholangiocarcinoma. This is a 64-year-old gentleman with both MTAP, CDKN2A/B mutations, KRAS mutations, all of our friends, SMAD mutations, ARID1A mutations, a very mutagenic tumor, not responsive to XELODA, gemcitabine, cisplatin plus pembro.

In the second line, you can see that we had a pretty rapid response, which is unusual for PRMT5 inhibitors at six weeks in the lung, which are, we have two lesions, right lower lung shown here. Baseline is on the left-hand side in the right-hand side panel. You can see the red outline there. You can hardly see it in the right lower lung, a little down there, but you can't see it at all at week 12 on the lower part of the lower lung. Then you can see the 60% reduction in the upper lesion. This patient remains on treatment, and had a confirmed PR by 12 weeks. Cholangiocarcinoma, highly resistant. We're looking for more responses as we go along. It's very early in the study, on average four months before you see a PRMT5 inhibitor responses.

One of the questions is, this MAT2A inhibitor, PRMT inhibitors, these are vertical inhibition. Are we going to see an increased response rate? Our hope is that we'll go from a 30% to a 60% when we have both together. Of course, there's always the question of, are you going to have any drug-drug interactions? We have tested that for IDE892 in a reversible CYP inhibition assay here on the left-hand side, where you can see seven major cytochrome P450s are only inhibited after a very high dose of IDE892, more than 45 micromolar. There's no time-dependent or irreversible inhibition of the seven major cytochrome P450s.

In humans, in our first dose level of the combination of IDE892 and IDE397 shown on the right-hand panel here, if you look at IDE892 alone or IDE892 together with IDE397, that's the top line, dotted and not dotted are right on top of each other, so no drug-drug interaction there. Then for 397, the lower lines, the same thing, 397 alone or together. This is very important because IDE397 is a CYP3A4 substrate. In fact, the experiment, as you all know, has been done by Amgen with a different PRMT5 inhibitor, AMG 193, where they combined that with our IDE397, our MAT2A inhibitor. PRMT5, MAT2A inhibition combo. This is an example, another cholangiocarcinoma, thought it would be nice to show you that, of a 35-year-old woman with MTAP deleted, tumors that have MTAP deletion, CDKN2A/B deletion, BAP1 mutation, et cetera.

This individual, unfortunately, was given multiple therapies, Durva plus GemCis, cytarabine and oxaliplatin, five-FU, leucovorin, and irinotecan, and no response to those prior therapies. With this combination, you can see a PR did not occur until week 18, which is kind of what you would see with this pathway being inhibited. It takes about four months, but she did have a very nice response at week 18, confirmed by RECIST 1.1 at week 21. Baseline is shown on the left-hand side of the images here. You can see all the liver tumors, very large, big, massive tumors, and then another one right below it on the right-hand side. This is your week 18 PR, and you can see one of those is hard to see. The women stayed on this treatment for more than a year.

Just to summarize our program summary clinically and upcoming milestones, IDE892 dose expansion evaluation continues. Dose level 3, 4, and five are what we are looking to expand in to determine our recommended phase III dose. IDE892 and IDE397 dose escalation is ongoing. We are targeting expansion in the first half of 2027 in MTAP-deleted non-small cell lung cancer. IDE892 pan-RAS combo, hopefully our first patient in will be in December. We are targeting that by the end of the year. IDE892 KRAS G12D combo, done by Genentech, run by Genentech, that FPI will be in the first quarter of 2027. There will be multiple catalysts and clinical data updates in 2027. 2027 is going to be a very exciting year for what we think is best in class MTAP PRMT5 inhibitor. We will have IDE892 monotherapy safety and efficacy update that will be more mature with time.

We will have our combination PRMT5/MAT2A inhibitor safety and efficacy data. We will have our pan-RAS combination safety and efficacy data for PDAC. We will also have plans for our registrational studies in monotherapy and combination, and hopefully initiate those by the end of 2027. With that, I will hand the baton back to Dr. Michael White, who is going to talk to us about AI and machine learning and getting better drugs and our KAT6/7 pathway.

Michael White
President of R&D, IDEAYA Biosciences

Thank you, Theo. I know you guys do not want to hear about AI, but our molecules do come from somewhere. These are difficult things to make. Thought it might be useful for you guys for us to introduce a little bit of this to you. We have talked about in the past how we use AI/ML to really as a force multiplier to drive productivity in our IND engine. Our philosophy is to build that on top of a firm physics foundation related to small molecule discovery, and at that center for us is using free energy perturbation as the ability to calculate and predict small molecule protein interactions. This is a quantum mechanics-based modeling that assesses these interactions atom by atom, taking into account molecular motion, taking into account the surrounding water, and taking into account energetic loss or gain from these binding events.

As you might imagine, the combinatorial complexity that is associated with those calculations is enormous. We have not been able to do very many. We wanted to unleash this and do this at scale. In order to accomplish this, we invested heavily in methods that allow us to reduce the complexity of the system by minimizing the sizes of the protein domains that we are using for these interactions and constraining their edges so that we account for that minimization, by optimizing our ability to calculate for each molecule in our library the surface energies that are associated with that molecule, and by using active learning loops to be able to enhance the productivity and the target-based accuracy of these models through our programs.

One example we are going to be talking about in a second, the KAT6/7 program, we actually accelerated that eight months versus what we would have otherwise been able to do because of the way we were able to crash through large numbers of designs and hone on those that would be the most predictive. Once we had this on the table with respect to a high bandwidth free energy perturbation pipeline, our next step really has been to release the potential of generative design. Generative design, as you guys may know, is purely computer-driven designs that can bring orthogonal options to the table that your chemists would have otherwise been aware of. That is a good thing.

The bad thing is that there are so many options, the majority of which are not going to survive the make and test cycle, that it is very difficult to hone in on a predictive molecule that you had not already thought of. The way that we are handling this is in this flowchart here. We are applying our FEP capabilities as a solution here. We enumerate the AI-generated designs against a particular target. We randomly select a collection of those, assess their binding potencies by FEP, build a model on that through iterative training to be able to enrich the selection of molecules that actually do have predicted affinities that would be productive for the program.

We use that model now to go back to the larger pool of predictions and extract those that have the highest predicted affinities from those AI-generated designs, and amongst those designs, we optimize for the drug-like properties using our HARMONY AI/ML prediction model so that now we can take thousands of designs and focus on those that are going to be most productive for advancing the program. This is a capability that is heavily in use for the next generation programs that we have coming on board, and has made a big contribution to the CDKN 2A program that I will talk about in a minute. This next slide, last piece on the AI/ML. The next step for us is to enable virtual screening.

We would like to be able to do this in order to do parallel assessments of many new targets with very attractive mechanistic properties, but which are outside of the space of targets that are druggable or the druggable class of targets. In order to be able to do this, virtual library screening, something that has been out here for a while. The caveat to virtual library screens, again, is parsing the signal from the noise. They do not work well when you are using a first-in-class target, and the reason is this docking step. These docking steps in virtual screens are defined by structural features within the pocket that have constraints on how a small molecule can interact with that pocket. Those constraints can only be identified by a crystal structure of your target of interest together with a known ligand.

If your target of interest is first in class, no known ligands, no constraints. That means that you generate tons of false positives in your docking, and you are unable to rank the affinity of molecules against that target to select them for synthesis and testing. We are putting into place methods to be able to solve that problem by determinism of constraints in these pockets through physics-based methods in order to eliminate false positives that come from the absence of pre-computed docking constraints. The final piece here is the ranking in order to score these thousands of hits with respect to what is going to bind best. The models that do that both best is the free energy perturbation calculations.

That is very difficult to apply at the scale of the hundreds of thousands of calculations that are required to parse through these screens, and we are solving that problem with some unique three-dimensional ML-derived descriptors that allow us to filter out low binders before making those calculations so that we can focus on the best targets. Now coming to IDE574, our dual KAT6/7 molecule is a very challenging profile to achieve. KAT6 and KAT7 are histone acetyltransferases that function by opening the chromatin at specific points that support the ability of transcription factors to bind targets that specify lineage identity. In tumor cells, KAT6/7 activity collaboratively acetylates histones to a large extent that unleashes lineage-dependent transcription factor activity that drives neoplastic transformation, that drives disease progression, and that drives adaptation to therapy. So IDE574 was designed in order to stop that from happening.

Here are three examples of preclinical models within the prioritized clinical development strategy that IDEAYA has for IDE574. Here is an example of an estrogen receptor-driven breast cancer model. We get complete regressions with 574. Here is an example of an NKX2-1-driven lung adenocarcinoma model with KEAP1/STK11 mutants. We also get regressions here. Here is an example of a TCF7L2 transcription factor-driven colorectal cancer model that gives us regressions. These blue lines are a KAT6/7 molecule. That is the Pfizer molecule. It cannot achieve this activity in these models because KAT6 inhibition alone is not sufficient to inhibit these transcription factor pathways. You need to hit both KAT6 and KAT7. Here on the upper right-hand side is a pharmacodynamic representation of what you get when you put KAT7 and KAT6 inhibition together in the same molecule.

Here, using a breast cancer model, you can see superior suppression of estrogen receptor signaling pathways by dual inhibition that you cannot achieve with either a KAT6 selective or a KAT7 selective. This also comes with superior inhibition of Wnt pathway activation, as noted by Yujiro, as well as superior inhibition of epithelial mesenchymal signaling. The other piece that we are very excited about is the role of KAT6 and KAT7 in supporting the transition, the adaptive transition to drug-tolerant persister cells that can serve as reservoirs that allow relapse when therapy is withdrawn. Here is an example of a classical model of looking at the emergence of drug-tolerant persister cells. This is an EGF receptor mutant lung cancer model treated with osimertinib. You get these complete responses. You take osimertinib off, and the tumors come back because of these adaptive transitions to drug tolerance.

If you have that in combination with IDE574, it does not happen. We can block that emergence of drug tolerance in the combination setting. As Frank alluded to, the real message for the update today is how we can take advantage of the role of KAT6/7 in allowing lineage transitions and adaptation as a combination partner to be able to intercept what we believe are KAT6/7-dependent resistance mechanisms in combination with RAS inhibitors in lung and colorectal cancers. Here is how we fell into this in the first place. The real opportunity behind this combination came first from our evaluation of the combination of IDE574 with many of the known and emerging assets in the bile cancer space.

Long story short, daraxonrasib turned out to be the winner there by a long shot, both with respect to its breadth of activity across models independently of the KRAS allele, as well as the depth of the synergy that we observed. With respect to the latter, if you zoom into these synergy scores and look at the inhibition matrix, the growth inhibition matrix, you can see most of these numbers are above 100%. The reason is that we convert a cytostatic response into a cytotoxic one in combination, so stopping cells from proliferating versus killing them. We do the latter in combination, which is a good thing.

If we evaluate the synergy scores and we ask, what are the transcription programs, the baseline transcription programs that are associated with the best synergy that we see in these models with daraxonrasib, we see expression signatures that are highly reminiscent of features that are being observed preclinically and clinically with respect to this lineage transition that Frank just talked about, where you move from a RAS-responsive, WNT-low, MAP kinase-high tumor to something that looks more like an intestinal stem cell, which is now WNT-high and MAP kinase-low. We think IDE574 destabilizes this transition as well as kills cells once they arrive into it. Consistent with that model, we see very nice collaborative inhibition of MYC when we put IDE574 together with daraxonrasib in vivo.

We see that in vitro as well, associated with poor clonogenicity, probably because we are affecting both MAP kinase-driven MYC as well as KAT6/7-dependent, WNT-driven MYC in order to see that enhanced response. What happens when we evaluate this hypothesis in vivo? You are seeing here very robust regressions observed with IDE574 together with daraxonrasib. On the left here is turning loss of tumor control into regressions, and on the right is turning loss of regressions into durable regressions.

If we look at the signatures that are associated with this at steady state after just three days of therapy, you can see that the combination is delivering deep suppression of Wnt, which is important because that's a lineage-specific tumor driver pathway, and we are also getting superior suppression of the PI3 kinase/mTOR pathway in this combination setting, which is very interesting in colorectal cancer, given that that is a tyrosine kinase-dependent bypass mechanism in this tumor type. If we look at the individual gene level, we find that the real standouts that we get in combination versus single agent is reduction of MYC and activation of PROX1. These two genes are known as master regulators of lineage transition. PROX1 inhibits it, MYC activates it. We think that IDE574 controls these directionally so that we maintain the RAS-sensitive state, preventing that transition into the intestinal stem cell-like state.

It was very gratifying for us to see very similar mechanisms at play in lung adenocarcinoma. Again, much synergy across alleles. Most of the models are showing very strong synergy scores, again, converting a cytostatic response into a cytotoxic one. In this particular case, when we asked for the features that were associated with the best synergy, we see, for example, AGER1, otherwise known as RAGE, which are biomarkers for alveolar type 1 cells, which is a RAS-refractory cell state, even in the presence of RAS mutations. It has been shown by a number of groups, both in mice and in humans, looking at this biomarker analysis of the cohort studies with sotorasib here from Tyler Jacks, modeling RAS-driven tumor formation in mice, that RAS mutants help to maintain an NKX2-1 high alveolar type 2 cell state. This is the sensitive cell state.

Under treatment, there is a lineage transition from AT2 cell state into this, sorry, yes, into the AT1 squamous cell state that becomes RAS independent. When you put 574 together with daraxonrasib, you see the opposite happens. We suppress these AT1 squamous markers, and then NKX2-1 gets activated by 50-fold in this combination to push cells into this NKX2-1 high alveolar type 2 cell RAS-sensitive state. What happens when we try this in vivo? Here's the lung cancer model, K-1 mutant KRAS G12C, daraxonrasib alone, lose control. Put this together with IDE574, we get complete regressions. Those complete regressions, again, are associated with deep suppression of transcription factor pathways. This signature is reading out the c-Jun/AP-1 pathway. This signature is reading out the NRF2 pathway.

What we think is happening here is the combination durability is a function of restricting adaptive lineage transition from this AT2 cell type to this AT1 cell type. Turn it back to you, CDO, to give them an update of what we're seeing in the clinic.

Theo Ross
Chief Development Officer, IDEAYA Biosciences

Thank you so much, Mike. Amazing data, KAT6/7 dual inhibitor, first one, hopefully, first in class, best in class. We're just neck and neck with Pfizer, so super exciting. Great data to start with. This is our clinical development strategy shown here. Again, monotherapy and clinical combinations going in parallel. We're in monotherapy dose escalation now, and our expansions will be in breast, CRC, and NSCLC, as mentioned by Mike. In terms of our IDE574 clinical combinations, we've started our combination with fulvestrant, and so far so good. We're having discussions with regards to combining with pan-RAS, other in CRC and NSCLC. I'm going to show you just two pieces of data, just our early safety data and an early clinical response. With regards to safety, shown here, we have dosed 32 patients in the monotherapy. This is only on monotherapy at this point.

Our combinations are a little bit too early for this, but we have five dose cohorts. 56% any grade TRAE, 18 of those 32 patients, and three of those 32 patients had grade 3 TRAEs. As we would expect, they're all hematologic in nature. The most common TRAEs are no different than what we see with the KAT6 inhibitors alone by Pfizer and others. Those are dysgeusia and anemia at 25%, and neutropenia and fatigue. All other TRAEs that we've observed are grade 1 and 2. There's no grade 4 or 5 TRAEs. We've not yet reached our maximally tolerated dose, and our dose levels are shown on the right-hand panel. Our projected clinical efficacious doses are our last two doses, dose level 4 and 5. The grade 3 TRAEs are two anemias. One anemia alone, it was a DLT.

That patient received a transfusion and continued at dose level 3. We have a neutropenia, and then anemia that I mentioned, along with a thrombocytopenia at dose level 3. So relatively manageable safety profile and no different than what has been seen for KAT6 inhibitors, which is something we're very excited about. Then this early clinical response that was seen in a patient with breast cancer, actually a fifth line dosing. This patient's a ER-positive, low HER2-neu lobular breast cancer, 79-year-old woman, who received her first scan. On the left-hand side is her baseline scan. It's of her liver. You can see outlines of two liver lesions, one in blue and one in red. The week six scan is shown on the right-hand side there, and it's obvious that she's had a nice, strong initial response at minus 40% for both her lesions together.

Then exciting, on the very right-hand side, we have some early PD data, which is acetylation of histone H3 subunits, lysine 23 acetylation and K14 acetylation for KAT6 and KAT7 respectively. You can see on the upper part, the early doses, dose level 2 and dose level 3, and then on the bottom, dose level 2 and dose level 3. So a nice inhibition of acetylation of our targets. On that point, I'm going to reintroduce Mike. He's going to talk about the exciting CDKN2A pathway and how he's targeting that with his amazing machine learning and all the really great stuff they do in discovery.

Michael White
President of R&D, IDEAYA Biosciences

[inaudible] β€” being a poster child for jet lag. This is the last section now, CDKN2A section. This is something that is a huge opportunity for us, in my opinion. Yujiro already described to you how CDKN2A and MTAP are always lost together. That is basically because CDKN2A drives MTAP loss, because it unleashes the mother of all tumor suppressors here, P16 and P15. What is the normal job for those? The normal job for those is to restrict entry into cell cycle and proliferation. What is the frequency of that? Well, you are talking about co-deletions in PDAC, that could be up to 40%. Like Yujiro said, we also have mutations in CDKN2A. We also have promoter methylation in CDKN2A. We also have chromatin inactivation in CDKN2A. That all adds up to almost everybody.

Almost all pancreas cancer patients have CDKN2A deficiency. Almost all pancreas cancer patients have RAS mutants. Almost all cancer patients with pancreas cancer have both CDKN2A loss and RAS mutation. This is important because, as Frank described, the major consequence of RAS effector pathway activation is to license cell cycle progression, and a major node of bypass mechanisms, both intrinsic and acquired bypass mechanisms, is activation of the cell cycle. This is druggable if you find the right profile to attack the CDKs that have been unleashed upon loss of P16 and P15 that drive cell cycle progression. What does our profile look like? Our CDKN2A program, potential first-in-class dual CDK inhibitor for CDKN2A cancers. I have this chart on there, this schematic on the upper left here to remind you guys that the licensing of proliferation really means activation of the E2F transcription factor.

The E2F transcription factor gene expression program is a great surrogate to understand how well you are doing that. We use that to be able to evaluate with tool compounds what our ideal profile would look like with respect to dual CDK inhibition. We wanted to make sure that we would inhibit estrogen-dependent cell cycle progression in breast cancer. You can see that in the top, in the middle panel. IDC427 is a tool compound that has the same CDK profile as IDE101, our clinical development candidate. This CDK4-driven tumor, CDK4-selective inhibitor does a pretty decent job. CDK2 doesn't do much at all, but we are superior to both of those. If you look at a cyclin E-driven model of cell cycle progression in ovarian cancer, that is really a CDK2 model. It can do it a little bit. CDK4 doesn't do much. Our compound squashes it.

CDKN2A loss-driven model, lung cancer models, pancreas cancer models, neither CDK4 or CDK2 can be sufficient. Our profile strongly inhibits E2F transcription factor pathway. If we look at IDE101, you can see in the next column, we extinguish S-phase progression in estrogen-driven breast cancer models. We extinguish S-phase in cyclin E-driven ovarian cancer models. We extinguish S-phase in CDKN2A loss-driven models in pancreas cancer, and that corresponds to single-agent activity. We are doing all of this without tripping myelosuppression that is CDK6-driven. You can see that in the figure on the lower left. This is an assay using human CD34 cells, which is a hematopoietic system that we can cause to produce myeloid cell populations by adding chemokines and cytokines. If you put palbociclib in here, you kill everything. Its IC50 is about 100 nanomolar, so as you know, myelosuppression is dose-limiting tox for palbociclib.

CDK4-selective, atirmociclib, very nice molecule. You get a nice shift away with respect to that EC50. Our molecule, IDE101, just as clean or better than atirmociclib. Very strong inhibition of multiple oncogenic pathways driving cell cycle progression without the liability, we think, of the myelosuppression that you would otherwise get if you were inhibiting CDK6. What about the combinability with RAS? We see nice synergy, upper left. Many models, it does not matter if you have MTAP, it does matter that you are CDKN2A deficient, but they are all CDKN2A deficient, so that is a good thing. Not all models respond. Wild-type models, the KRAS wild-type models, we do not see synergy. There is an example of what the synergy looks like in the middle there. On the top right, we are doing washout experiments here to see how long inhibition of proliferation can persist after drug withdrawal.

What you see is daraxonrasib, those cells come right back after you take the drug off. IDE101 does a much better job keeping them off, but the combination really keeps it off for quite a long time. This gets back to what Yujiro introduced at the beginning, that made us start thinking about cellular pathways that are getting engaged by the drugs that are also helping to suppress cell cycle. When we did immunoblots, we saw exactly that. IDE101, you can see in those immunoblots, that is inducing the p21 tumor suppressor. Daraxonrasib, that is inducing the p27 tumor suppressor. The two of them together, you get both. What are those? That is the other arm that can suppress cell cycle progression at multiple stages in this cycle. So we see compensation for CDKN2A loss by activation of the CIP/KIP tumor suppressors.

That is giving us very deep control of the proliferation regulatory system in that combination because we are hitting the drivers directly with ATP competitive inhibitors, and we are inducing the cellular tumor suppressor function at the same time. What happens when we put this in vivo? Very interesting data here. This compound, IDC318, this is an in vivo tool compound, has the same profile as IDE101, except IDE101 has better drug-like properties for humans. What you can see in these studies here, KP4, that is the most difficult model for daraxonrasib to get on top of, according to the publications that are coming out of RevMed. We put that together with IDC318, our dual CDK inhibitor. All those are regressing. Almost all of them are complete regressions. We see the same thing with a different kind of RAS inhibitor. This model, MIA PaCa-2, using sotorasib because that is G12C.

We get nice activity with sotorasib, but that comes back. If you put the combination on board, it stays down. Why? Look at that Phospho-Rb up there in the upper right. In combination, we basically extinguish Rb phosphorylation, so we are keeping Rb on to be able to control these pathways. If we look at the gene expression programs that are induced in these tumors on therapy, we see deep suppression of the DREAM targets, these are the proliferation genes, in combination, deep suppression of the G2/M checkpoint, and also deep suppression of MYC pathways. The other piece of data that I want to show you is, what if we let tumors relapse on daraxonrasib and then add our CDK inhibitor on top? You can see the answer there. In combination, in both of these models, we see very strong combination activity, deep regressions.

The daraxonrasib arm, which is the red one, after those tumors got big, we put our molecule on board, and then they shrunk. They look almost as good at the end of the day as if the combination had been added at the very beginning. Some mechanistic work on the right. What is actually happening here, we are using derived daraxonrasib-resistant MIA PaCa cells. They are not cross-resistant to IDE101. If you look at the resistant cells versus the parental, you see that these daraxonrasib-resistant cells have lost the ability to control p27. If we add IDE101 on top of that, we regain cell cycle control independent of that p27 response. So we reverse resistance-mediated cell cycle escape. There, I am done. Yujiro, I will let you close it out.

Yujiro Hata
CEO, IDEAYA Biosciences

Well, great. Thank you, Mike. Thank you, Theo, to the presenters. As you see, we went through quite a lot of data today, and hopefully getting a sense of the breadth of the pipeline and really digging in deep on the R&D strategy for the company. I will not go through all of the specific items that was mentioned earlier, but just needless to say, as you can imagine, we have a lot of data catalysts that we are anticipating in 2027. First, as mentioned earlier from Theo, we do anticipate an IDE892 PRMT5 monotherapy update in 2027. We also expect a combination update for both the IDE892 pan-RAS combination and PDAC, the IDE892/397 combination, and MTAP deletion as well. In addition, we are planning to update on IDE574, our KAT6/7 inhibitor, in terms of a monotherapy update, and also potentially in combination with the SERD inhibitor, fulvestrant.

A lot of updates to come. As also noted earlier, we hope to be able to transition the MTAP program towards registrational studies by the end of next year. Lastly, with CDK2A, it probably won't be surprising to note that we are getting a lot of interest on this program. We do believe, in addition to what we are doing, KAT6/7 and CDK2A, expect for us to be doing more combinations with RAS molecules, including indications such as pancreatic cancer, as well as CRC. With that, thank you, everyone, for joining us for our R&D Day, and we will now transition to the analyst Q&A portion of the event.

Maury Raycroft
Analyst, Jefferies

Thank you. This is Maury Raycroft from Jefferies. [inaudible]

Yujiro Hata
CEO, IDEAYA Biosciences

β€”Pan-RAS would be relevant, KRAS G12D as well, and of course, what we are doing with our PRMT5 inhibitor, Maury. I think there, our first objective is get the IND filed. I think what we can say there is the GLP tox studies are going exactly as planned. We anticipate that will get filed in the first half of next year. We will obviously prosecute the monotherapy, but I would say based on the level of interest we have been seeing, we would like to get combination studies for the reasons Mike just went through. We think this asset positions us very uniquely, especially in the field of pancreatic cancer. We would like to get those studies going as soon as possible.

Maury Raycroft
Analyst, Jefferies

Got it. That makes sense. Maybe one other question just on the 70% of PDAC patients with the CDKN2A deficiency. Can you break down that population by homozygous deletion, inactivating mutation, and epigenetic silencing? Do you expect the drug to work across the board?

Michael White
President of R&D, IDEAYA Biosciences

Yeah. I have got the mic. Thanks for that question, Maury. So we have that breakdown on the slide, so you guys can take that home. It has got the percentages on there. We have done extensive evaluation across multiple datasets to be able to come up with those values, and you should be able to reproduce them yourselves. With respect to the assessments, as you saw from the slide where I was listing out the synergy that we are seeing, cellularly, we are seeing that synergy across all of those mechanisms, not just homozygous deletion. So CDKN2A deficiency is a real thing in the absence of an actual structural variant like homozygous deletion.

Tyler Van Buren
Analyst, TD Cowen

Great. Thanks so much for the presentations. Tyler Van Buren from TD Cowen. I wanted to ask a couple. I guess the first one is on PRMT5. How do you expect to prove clinical differentiation with IDE892 relative to the first-gen PRMT5 inhibitors that have come before it on the efficacy front, because safety looks good, especially given the high bar set by the Tango combination most recently at ESMO, and when should we expect those data? On the RAS combinations, I guess moving beyond the pan-RAS combo, the CDKN2A preclinical data at the end there looked pretty striking. So I guess, how soon can we get the first combination clinical data from that that really validates that, and are you thinking about a triplet as well?

Yujiro Hata
CEO, IDEAYA Biosciences

Yeah. In terms of PRMT5, Tyler, I would say there, our focus is around safety. If you look at the various clinical datasets that have been presented on PRMT5, including fairly recently at ESMO, at least our view on the data is it looks across multiple clinical data updates that have happened over the last several years. So we think really the key point of differentiation will be about safety. We obviously want to see a consistent level of response, but based on what we are seeing from the pharmacodynamic side, we feel good about that likelihood. Safety is going to be paramount because we know the agents you are combining them with, first, as Mike highlighted, is ideally you do not have a drug-drug interaction, and that includes not just for pan-RAS, but also for KRAS G12D.

Second, as we mentioned around some of the AE profile parameters, we know GI, there is GI that is observed with pan-RAS, for example. We know skin rash is another one to see. We know that those AEs are observed with other PRMT5 inhibitors. The CNS side as well. I think once you start talking about longer treatments, frontline strategies, CNS, we think tox liability could be a clear point of differentiation versus our peer. Front and center is we feel if we can deliver a clear efficacy dataset in combination with pan-RAS without seeing drug-drug interaction, we think that is a significant win. We do believe multiple groups would view it that way as well.

In terms of CDKN2A, and when we might be able to generate data on that, and particularly with pan-RAS and also possibly PRMT5, that is probably more of a 2028 type event, I would think, on the combination side. But we will move as quickly as possible.

Michael White
President of R&D, IDEAYA Biosciences

I can address your third question there about the triple. This molecule was the hardest molecule we have ever made. I think you guys just cannot wait till we say more about this molecule. This molecule also was made with a very, very strong line of sight on properties. We have properties here that are going to facilitate combination opportunities, and we definitely want to take advantage of those as quickly as possible. This is a situation where triple, that is asking for a lot from a clinical development standpoint, but if we have the best molecules, certainly 892 has a profile that benefits that, 101 has a profile that benefits that as well, and we seem to be clean on the liabilities that are associated with the RAS molecules that everybody is putting out at the moment.

From the combination standpoint, don't forget about KAT6/7. There is a situation where we have additional combination thoughts with 101 and KAT6/7. Then KAT6/7, I want to make sure that folks recognize that what we are seeing there, you need to have a RAS mutation, and you can combine in lung. You need to have a RAS mutation, and you can combine in colorectal. That is going to be a very, very big opportunity for the company.

Charles Zhu
Analyst, LifeSci Capital

Excellent. Let's try this. Okay, great. Charles Zhu from LifeSci Capital. Thanks for the presentation and for taking the questions. Two, one, two. First one being, can you talk about the current status of biomarker testing and identification for CDKN2A loss, whether through its the various methods of how that is lost? That's one. And two is, fully appreciate the emphasis on non-brain penetrance for PRMT5 in an effort to avoid CNS toxicities. But just looking at your MTA cooperativity and SAM negativity profile, you seem to be driving a wide window systemically. Why the worry in the brain if you did have a brain penetrant molecule? Thank you.

Michael White
President of R&D, IDEAYA Biosciences

Take that last one there, Charles. One of the things I was showing you was the impact of PRMT5 on alternative splicing. Alternative splicing, in order to have an impact in tumors, you need to crush it, right? You need to get that pathway off in the tumor. In order to impact alternative splicing, a little bit can do that, right? That's a very, very sensitive readout. You really want to avoid that as much as you can in normal tissues. Yes, we have a really big window. Do we want to go in the brain? We purposefully wanted to stay out of the brain because of the role that we think alternative splicing plays in cognition, in learning and memory, confusion. These are things that patients notice, so that is not the direction that we're going.

Of course, in pancreas cancer, CNS disease is not a player in that setting, so that would be more in the context of non-small cell lung cancer. The first question you were asking about the biomarker, that one slipped by me.

Yujiro Hata
CEO, IDEAYA Biosciences

Yeah, maybe I can take that, Charles. In terms of the two main settings we're going to be focused for this asset, one is in combination with PRMT5 and MTAP CDKN2A co-deleted tumors. There, we're going to likely just rely on MTAP because they're almost always co-deleted, 95% plus or more. In the area of pancreatic cancer, I mean, our view is the frequency is so high, we probably won't be patient selecting. At least our perspective is that first combination, if it's with a pan-RAS inhibitor and PDAC, we would just do it in unselected population.

Theo Ross
Chief Development Officer, IDEAYA Biosciences

Do you want me to go? Yeah, or anything else. No, I mean, CDKN2A is measured in almost all of the NGS testing out there, so it's a pretty easy biomarker.

Laura Prendergast
Analyst, Stifel

Hey, guys. Laura Prendergast from Stifel. Congrats on the exciting new asset. Excited to see this move into the clinic. Tango has shown that a PRMT5 combination seems to alleviate some of the pan-RAS-driven rash AE adverse event. Do you expect this to be a class effect as you move forward with a different pan-RAS and your molecule? What doses are you guys planning on starting at? I know you haven't disclosed your recommended phase II dose yet, but for Roche's pan-RAS and your PRMT5, should we expect you guys starting both molecules near the recommended phase II dose?

Yujiro Hata
CEO, IDEAYA Biosciences

Yes, maybe I can take that, and I'll give Mike for the rash question. In terms of the combination with pan-RAS, our opening strategy here will be starting the pan-RAS molecule at their expansion dose. We know what that is, and so that's where we'll be starting. With PRMT5, as Theo showed, we have now three expansion doses, so we'll likely be starting at a dose we know will be efficacious. It's very possible that our first cohort is our final cohort, because we anticipate that will be a very active dose. They'll both be at expansion doses, which is, I think, will enable us to move very rapidly. So far, we don't anticipate overlapping AEs based on the data you saw and based on what we expect from a pan-RAS inhibitor. Do you want on the rash mechanism?

Michael White
President of R&D, IDEAYA Biosciences

Yeah, so what you might be getting at, I think, is a historical anecdote with PF-9999, which was the Pfizer SAM competitive PRMT5 inhibitor, where they reported a patient where they had some reduction in rash in that patient. I don't know if anyone ever figured out what that was about. Is that about impairing the humoral response or the innate immune pathway activation? Anything happening in that setting would be happening with a molecule that wasn't tumor-specific. So, unless it only required a very, very light touch, I wouldn't expect us to be provoking mechanisms like that one. The other thing for us was from a combination setting, we're expecting to go with full dose intensity with everybody, so nobody has to dose reduce because of the window that we have.

We think there are only a handful of molecules like ours that have that SAM anti-cooperativity and the MTA super cooperativity, as well as the absence of those DDI issues so that we can safely use full intensity of the RAS inhibitor.

Greg Renza
Analyst, Truist Securities

Great. Go od afternoon, good evening, IDEAYA team. Thanks for hosting the event. Congrats on the progress. This is Greg Renza from Truist Securities. Two questions for me. Maybe starting number one with Mike, perhaps circling back to IDE101. Mike, I think you've certainly piqued our curiosity when it comes to the mechanism of action. I'm just curious, it's described as a dual CDK inhibitor. What perhaps could you tell us about the specific targets and how they're different from CDK targeting with the agents that are approved or under development? Then the second question maybe for Yujiro. I think IDEAYA has been steadfast in the combination and partnership strategy over the years. With this portfolio and the portfolio in its current state, what is the right level of partnership presence that you want to pursue with your programs under development? Thank you.

Michael White
President of R&D, IDEAYA Biosciences

Your first question, you saw Yujiro raise his mic up. That was because he was going to smack me on the head if I answered your question. What I can say is that this is a potential first-in-class. I am unaware of any molecule that has the selectivity profile that we have. Pretty excited about that one, but more to come later.

Yujiro Hata
CEO, IDEAYA Biosciences

Yeah, in terms of the collaboration strategy, Greg, look, we're thrilled to be working with Roche and Genentech on two combinations, for pan-RAS and G12D. At this time, we have a firm commitment to them to generate that data as quickly as possible. We do think we have a great PRMT5 inhibitor. We think they have a great both pan-RAS and KRAS G12D inhibitor based on the data we evaluated. There's clear alignment between the two organizations on the clinical strategy. I don't know, Theo, anything you'd add on? You've been taking point.

Theo Ross
Chief Development Officer, IDEAYA Biosciences

With regards Oh, I don't need that. With regards to partnership, I think they have two good molecules and they're good partners and we're aligned with our clinical development plan, and they're aligned with, if it works out, that we'll have a registrational study ongoing in 2028 in the frontline pan-can. So, pretty excited about that.

Yujiro Hata
CEO, IDEAYA Biosciences

Yeah, maybe just the last part I will add, Greg, is, look, we also appreciate there are other molecules, inhibitors out there. So, look, we are having different discussions ongoing, not surprisingly, in parallel. We just want to be mindful about what we do. But I do anticipate that we will have more combinations that will be ongoing, and hopefully you'll hear more about that through 2027.

Andy Berens
Analyst, Leerink Partners

Hi. Andy Berens, Leerink Partners. Thanks for hosting this. It was very helpful. I'll ask an AI question. No, just kidding. I'll stick to what I know. For the PRMT5 program, I guess, we've seen some of your competitors say that they think that the wild type, there's reason to believe that it could be active more broadly than loss of function. Just wondering if you thought that's the case. Then, there are some situations that obviously brain activity could be helpful. Could you retro-engineer IDE892 to make it brain penetrant? Then lastly, are you thinking about going broader than just the Roche compound, maybe with daraxon to PDAC? Thanks.

Michael White
President of R&D, IDEAYA Biosciences

Well, so the first part of your question around MTA cooperative PRMT5 activity in the wild type setting, I think that might work if you don't have a window for MTA-bound PRMT5 versus SAM-bound PRMT5. MTA is chewed up like crazy by MTAP. It doesn't accumulate in wild type cells. So I think what you would be looking at there would be some maybe an opportunity that would be associated with kind of like a first generation PRMT5. If you had some of that activity, maybe you would see that in a wild type setting. I do not anticipate that we will see any activity whatsoever in the wild type setting with our MTA cooperative PRMT5 molecule plus RAS. It's simply not how it works.

Yujiro Hata
CEO, IDEAYA Biosciences

I think, Andy, your just other question on combinations and daraxonrasib, I think was your question and whether we would consider that. Look, we appreciate it's approved as a standard of care in pancreatic cancer, so I do think in the pan-RAS area, if we were to consider an additional, I think it would likely be daraxonrasib for obvious reasons. I think here, it would really be from a mindset of generating proof of concept data, and that that would be valuable data set either way to have in hand. We are in the planning phases internally about our broader clinical combination strategy, but if we were to mention one, that would probably be the molecule for reasons I think everybody would understand.