2026.09.23 | GP | Pharma | Gandeeva GND-323 Solid Tumors

Inhibitory Molecular Glues Unlock Previously Undruggable Protein Interfaces?

Primary:

Gandeeva Therapeutics, Inc.

Associated:

RVMD

ERAS

Nested Therapeutics, Inc.

Viewpoint:

Academic KOL - Physician Scientist

Moderator: Chinmay Rahane (CR)

| Sep 23, 2026 | 22 Min Read

Piro Lito (PL), Professor - Weill Cornell Medicine

SU MMARY

Overview

The discussion focused on Gandeeva Therapeutics' preclinical asset, GND-323, which is being developed for solid tumors. The conversation revolved around the challenges and potential of molecular glues, particularly in targeting the MAP kinase pathway. The expert provided insights into the requirements for evaluating such therapeutics, the role of AI and Cryo-EM in drug discovery, and the complexities of therapeutic index and resistance.

Key Insights

The expert expressed skepticism about the current claims of Gandeeva's AI and Cryo-EM capabilities without concrete data. They emphasized the need for detailed information on the target and binding partners to assess the scientific and investment potential of GND-323. The potential for Gandeeva's platform to advance drug discovery was acknowledged, but the expert highlighted the gap between AI predictions and Cryo-EM validation. The expert stressed the importance of understanding the therapeutic index and resistance mechanisms, noting that clinical trials are essential for determining these factors.

Competitive Landscape

Company/Product

Positive Aspects

Negative Aspects

Gandeeva Therapeutics

Potential novel target with AI and Cryo-EM integration.

Lack of disclosed data on target and mechanism; skepticism about AI- Cryo-EM claims.

Revolution Medicine

Collaboration on cyclophilin RAS glue; experience in MAP kinase pathway.

Not specified.

Verastem (RAF-MEK glue)

FDA-approved for different indications; experience with molecular glues.

Considerable toxicity in patients.

Quantitative Insights

No specific financial figures or metrics were discussed in the transcript.

Clinical Data

The expert mentioned the need for pharmacodynamic endpoints in preclinical models, such as ERK phosphorylation, to evaluate the efficacy of molecular glues. They also discussed the importance of clinical trials in assessing therapeutic index and resistance.

The expert concluded that more specific data from Gandeeva would be necessary to make a comprehensive assessment of GND-323's potential. They expressed interest in further discussions if more information becomes available.

CR: A very good morning and thank you so much for joining us today. My name is Chinmay and I will be the moderator for this online call which will focus on Gandeeva Therapeutics GND-323 in solid tumors. As a reminder, we will be only discussing publicly available information. What are you able to tell us a bit about yourself, your clinical practice and your research focus?

PL: Yes, sure. I'm a physician scientist at Memorial Sloan Kettering Cancer Center. My area of focus has been RAS- driven cancers. I've been working in this space since early 2000s and I've been heavily involved in the preclinical and clinical development of various RAS-directed therapeutics and my work spans the full spectrum of translational research beginning from basic structural biology and biochemical studies all the way to clinical trials and then reverse translation back into the bench. Yes, I'm happy to be on the call and I'm happy to discuss the questions that you have.

CR: Got it. So as you may be aware, we're discussing a preclinical asset. So there's a bit of information available publicly at least is quite less. Gandeeva currently discloses GND-323 only as an inhibitory molecular glue for solid tumors without publicly naming the target proteins, tumor genotype or even lead indication. What's the minimum biological information you need before you could judge whether it deserves serious scientific or investment diligence?

PL: I mean we would need to know some basic information as to what the target is and I'm assuming this is a molecular glue approach so what is the binding partner so essentially what is the ternary complex that is being formed and then with that understanding then one would go back and then do the necessary due diligence with regards to whether or not there are additional agents that target the same complex or not, and whether there is a comparator that would essentially establish a particular path of drug development, because then one can use the comparator as a benchmark. If there is no comparator, if the target is completely novel, then that can be both a blessing and a curse, and it can be a curse because you have to begin everything anew and you have to establish the entire platform without a reference. But it is a blessing because then there would not be any other competing agents in the clinic, so that might enable the clinical development path to be a little bit more straightforward and less competitive. But, yeah, I'm not aware that there is any additional insight with regards to the target, and I think the only clue is that it's a MAP kinase pathway protein, but the MAP kinase is a very broad term, so it's very hard to speculate on that. There are some existing molecular glues that target members of the MAP kinase pathway, including 2 glues that my laboratory has been heavily involved in describing their mechanisms of action. This includes the RAF-MEK glue that we published in 2014, and more recently, the work that we have done in collaboration with Revolution Medicine, which is a cyclophilin RAS glue, and those could be considered as some examples in the field, but again, not knowing whether the company has this as a potential target or other targets in the MAP kinase pathway.

CR: Right. So Gandeeva's core platform thesis is that AI can identify transient PPIs and Cryo-EM can visualize druggable interfaces. The traditional drug discovery myths. How often do you expect such computationally-related interactions to translate into relevant disease drug targets actually in the clinic?

PL: I mean, this is... Personally, I am a bit skeptical on this claim, and I say this with a grain of salt, because we do not have any actual data from the company, from Gandeeva, to know how to really evaluate their claim. One would need

to have access to a lot of their development pipeline data to be able to determine with certainty whether the claim that they are making is substantial or not. But from a generic standpoint, from a general standpoint, you know, there is a little bit of a challenge even in the way that you phrased the question because if indeed there is a transient protein- protein interaction that is being identified by AI, by virtue of the transient nature of the interaction, I imagine that would be extremely difficult to confirm such an interaction utilizing Cryo-EM because you must have some stability of the complex formation in order for the Cryo-EM to give you anything really. So again, unless Gandeeva has figured out some special sauce with which to further elucidate and validate these types of transient interactions, I'm a little bit skeptical about that. Of course, the AI field is moving at a tremendous pace, so if they have developed a proprietary algorithm that is able to identify these features a lot better than existing algorithms, then of course it could be a tremendous step forward. Similarly, if they've been able to appropriate AI and implement AI within their Cryo-EM pipelines, maybe they have developed a major breakthrough that will enable us to get even better outcomes from Cryo-EM than we have currently been able to. But based on what is currently available right now, I see that there is a big gap between predicting a PPI interface using AI and then validating the same interface using Cryo-EM.

CR: Right. So speaking of Cryo-EM, what would Cryo-EM actually need to contribute specifically?

PL: Yeah, I mean, it's hard to say this, right? Because there are a couple of ways that one can go about here. Cryo-EM can be helpful in the sense that it can validate the prediction of the AI, although in my opinion that seems to be a bit of an overkill. Why would you need Cryo-EM? You can use regular x-ray based crystallography with 2 of the key domains involved in the interaction and that's a lot simpler to accomplish than it is with Cryo-EM. Of course, Cryo-EM can be better in organizing conformational transitions and more dynamic exchanges that happen within binding interactions. Of course, Cryo-EM could provide tremendously deeper resolution and key aspects of the particular interaction. And by resolution I do not mean the structural resolution, I mean resolution in the sense of a better understanding of the dynamic range of the complexes and the different conformations involved in the complex. And of course Cryo-EM could be helpful here, but again, it's not as straightforward as it might sound, where you predict something with AI and then you immediately validate it with Cryo-EM. It's not as straightforward as that. Of course, if the AI component here has been implemented within the Cryo-EM pipeline, then the AI here is essentially being utilized to evaluate the different poses and the different volumes that are being generated by Cryo-EM. And then AI then becomes the tool to help interpret the Cryo-EM information and particles. That could be, again, a key conceptual advance. But again, I do not know that Gandeeva has really achieved this or not. For that, one would need to look at their data to be able to determine with certainty.

CR: Right. Basically, I mean, I'm still working with Cryo-EM, right? They say that their Cryo-CADD workflow repeatedly cycles Cryo-EM structures back into medicinal chemistry. What kind of metrics would convince you that this produces better therapeutic molecules and not just data, structural data?

PL: I'm not sure that I understand your question because I don't think there is anything particular that Cryo-EM would provide here that would convince me, right? The Cryo-EM is just a tool. Of course, if one is proposing to develop or has developed a molecular glue, there are certain features that one must demonstrate. So they must demonstrate that these 2 proteins, protein A and protein B, form a ternary complex with a ligand present. And then one would need to understand where exactly is the interface. And then to understand the interface, one would need to have structural information and whether that structure is obtained with Cryo-EM or the more standard X-ray crystallography, either one would be fine, but the structure would need to be understood so that we understand the interface of the interaction. And understanding the interface is important because then it can help us optimize the ligand to increase potency, to increase the various characteristics, and of course to begin to predict potential mechanisms of resistance, etc. So Cryo-EM per se here is a tool to just reveal the structure. So right, if Gandeeva has figured out a way to make Cryo-EM a lot more easy to perform, that would be a tremendous advance, but that's not the sense that I'm getting from, again, the very non-specific information that you shared with me.

CR: Right, so what I'm getting at is basically if someone claims, I mean if Gandeeva claims that their workflow translates cryo-structures to medicinal chemistry, I would expect something like a hit rate or time from hit to candidate, something like that, or maybe even potential improvement.

PL: Okay.

CR: with a generic candidate.

PL: But see, that to me sounds like a strawman argument because our limitations right now, especially with molecular glues, are not obtaining the structures. The structures can be obtained unless we're talking about very big proteins, right? And if we're talking about very big proteins, then the Cryo-EM could be tremendously helpful to accomplish that. And if indeed Gandeeva has demonstrated what they claim to be true, then there would be a big step forward, I think. But again, the proof lies in the pudding, so we would need to see the data to know to what extent they have demonstrated what it is that they are claiming versus to what extent this is wishful thinking, you know?

CR: Mm-hmm.

PL: Right.

CR: Mm-hmm.

PL: [Continuing...] CR: Yeah, got it. Makes sense.

PL: No problem.

CR: What's the therapeutic index evidence you would require for an inhibitory glue whose pharmacologic dependence on bringing 2 proteins together basically potentially creating a complex that does not exist in normal cells?

PL: Well, I mean, this depends on what the target is and what the target proteins are, because whether or not this interaction does not occur in normal cells under physiological conditions, it doesn't mean that the glue will not have a potential for toxicity. And let me give you an example, because again, this is very hard to answer a question of therapeutic index from a conceptual perspective. So let me give you a couple of examples here of 2 molecular glues that have been developed in the MAP kinase pathway, and both of which have been tested in the clinic. And before I continue with my example, let me just say that it's very hard to predict therapeutic index in the preclinical realm. The best way to understand and to realize therapeutic indexes is to carry out clinical trials and to see what happens in patients. But the 2 examples that we have with molecular glues in the MAP kinase field are what I was alluding to before. 1 is the RAF-MEK glue that was initially developed by Chugai and now is licensed by Verastem. And again, we describe the mechanism of how they work, which is to bring in MEK in a complex with RAF, and the MEK is catalytically inactive in that complex. So now, another example is of a glue that was intended to bring cyclophilin in contact with RAS. And again, this interaction of cyclophilin with RAS does not happen under physiological conditions. It's a synthetic complex that is mediated by the small molecule ligand. But in both of those examples, both with the RAF- MEK glue as well as with the cyclophilin-RAS glue, there is considerable toxicity in patients. So even though the cyclophilin-RAS interaction does not happen physiologically, when you bring those 2 together, there is still toxicity. So I don't think that one can rely on this notion that because the 2 proteins do not interact with each other in normal cells, that will give us therapeutic index. I don't think that's a sufficient argument to make right now because we do not have any sufficient evidence to justify that argument. So I think it will need to be examined on the basis of what the target is and what the binding proteins are and what are the dynamics of the molecular glue itself. Now in both of the

examples that I gave you, even though there is toxicity, both agents have been FDA approved for different indications. So despite there being toxicity, there is a clinical path forward for molecular glues that will lead to them being approved. And of course, toxicity can be identified and doses can be identified where patients still exhibit meaningful clinical benefit despite having some toxicity. But again, the therapeutic index cannot be defined simply by knowing which of the 2 proteins come together via the glue. One needs to know the biology underlying them, whether any of the 2 proteins are mutated or wild-type, etc. So there are many factors that contribute to therapeutic index.

CR: Right. And speaking of pharmacodynamics, what would be the test and the evidence which would convince that the protein alteration actually works? Would you go for paired biopsy or some assays to get DNA or...

PL: Yeah.

CR: Will you...

PL: Are you talking... go through blood?

CR: Yeah. Are you talking... I'm assuming that you are referring to clinical?

Yeah.

PL: Some of the dynamic endpoints or preclinical?

CR: Yeah. Both. We could actually... yeah.

PL: Okay. So then let me begin with preclinical because it's a lot easier to establish here. And since we are working, let's assume that the glue here involves 2 members or a member of the MAP kinase pathway. I mean, the first pharmacodynamic endpoint in preclinical models would be to evaluate what is the effect of the glue on the signaling output of the pathway. So let's say this is a MAP kinase, then the output in question here would be ERK phosphorylation or ERK transcriptional output, something like that. And then one would need to know whether there is a dose-dependent reduction in ERK activation downstream of the glue target and whether this inhibition can be reversed if there is, for example, genetic ablation of one of the glue targets to determine that the molecule is selective for this. Or, for example, if we understand what the interface is, then engineering interface point mutations, then one would want to see that whatever the effect of the glue was on pathway inhibition is now reversed when one introduces this specific interface mutation. So I think from a preclinical standpoint, that would be the most essential feature to have to determine that there is indeed an inhibitory effect as a consequence of this molecular glue complex. And of course, there are other preclinical studies that would need to be implemented to determine that there is indeed complex formation, et cetera, et cetera, and I'm happy to discuss those if you see fit. Now once this has been established preclinically, like once we have pathway inhibition, then the next level would need to determine whether this pathway inhibition correlates with antiproliferative effect. So imagine if we are targeting again with the glue, the MAP kinase pathway, we would need a series of cell lines that have dependency on the MAP kinase for their proliferation and then treat them with the glue to determine whether there is an IC50 effect and what is the IC50. And then if that is obtained, then the next step would be to determine what happens in vivo with either cell line or patient- derived xenograft models and once all of these have been established, then one can then move to the clinic and then depending on where what is the space and what is the competitive nature of the target within the clinical trial, then one can begin to envision about targeting but doing pharmacodynamic endpoints with on-treatment specimens or cell-free DNA or other means.

CR: Thank you very much.

PL: Obviously, the best pharmacodynamic endpoints to have in patients would be a biopsy-based, like an on- treatment biopsy-based approach, so that one can then be able to assess the same parameters that were tested preclinically can also be tested in the clinical specimens, like for example, ERK output by ERK phosphorylation or by transcriptional programs, etc., those can be determined in patients as well. However, it's not always easy to obtain clinical specimens shortly after treatment has started because a lot of factors that need to be considered here include, for example, what is the half-life of the inhibitor, whether it's administered orally versus IV, whether we're talking about a solid tumor indication that has biopsies that are easy to obtain. For example, melanoma, it's very easy to obtain biopsies. For example, pancreatic cancer, it's harder to obtain biopsies. You see what I mean, right? So it depends a lot on the clinical context to what extent one can do this. Now, of course, you mentioned cell-free DNA. Cell-free DNA can be helpful, but that doesn't necessarily inform so much the pharmacodynamic endpoint. It might provide us with a clue towards the minimum residual disease after treatment, but not so much pharmacodynamic endpoint. One can use it as a proxy. One can use cell-free DNA reduction upon treatment as a proxy for radiologic response thereafter, but still I would not feel very comfortable with relying on cell-free DNA alone as the sole pharmacodynamic endpoint. I would need some of the other measurements that I just mentioned.

CR: Understood.

PL: Yeah, yeah.

CR: And molecular glues may also create resistance to mutations in either protein partner. Do you think there should be also resistance experiments before the assay enters the clinic?

PL: Yeah, I mean, so this depends a little bit again on what the target is and how saturated the field is with agents that inhibit the same target. If, for example, this is a completely novel target without a lot of competition in the field, one could imagine that prioritizing the pharmacokinetic properties of the glue and then maximizing entry into the clinic and then carrying out the first phase of clinical trials that may be okay. To start with that and then try to determine mechanisms of resistance based on clinical performance in patients because if you're talking about a target that has not been previously validated the most important thing is not so much whether there will be resistance down the line, the most important thing will be the proof of concept, the clinical proof of concept. Is there a clinical response to begin with? So I think in that particular example, it would be better, I would imagine, to advance the clinical development strategy so that the clinical trial phase one is carried out so that we have a dose and we determine whether there are some initial clinical benefits before we think about resistance. If, however, we're talking about a target like, for example, the RAF or RAS or MEK, for which we already have proof-of-concept evidence with small molecule inhibitors or other glues, then in that particular instance, I probably would want to see a lot more preclinical assessment of mechanisms of resistance. I think because it becomes very important to understand whether resistance to the molecular glue happens as a consequence of similar alterations that cause resistance to the other agents for the same target. I hope that makes sense. So I think it would depend a little bit on what the target are. But again, these questions are, it's very difficult to answer these questions in a meaningful manner in such vague terms because a lot will depend on what the target is and what is the competition, the relevant competition in the particular space, etc. And I think I could provide you with a lot more concrete information if I had access through the actual data from the company.

CR: Right. That's a challenge for me as well. So Gandeeva also operates a substantial Cryo-EM service capability and has collaborations with Zymworks. Do these partnerships like with Zymworks materially de-risk the therapeutic pipeline or they mainly validate if Gandeeva is a high-quality service provider basically?

PL: I mean, again, it depends a little bit on what is Gandeeva providing to the collaboration, right? If it's just providing the standard run-of-the-mill Cryo-EM approach, I wouldn't be that excited about it. But I have to assume, and I underline, I assume here, they're doing a little bit more than that. If they figured out such a methodological innovation

to implement AI into the Cryo-EM workflow, and they have evidence to suggest that this implementation does indeed result in a more rapid analysis or in a more robust structural characterization, of course, that would be a tremendous benefit, not just for Zymworks, but for other entities out there as well. But again, one would need to have a little bit more insight as to what exactly is Gandeeva doing to be able to answer this specifically. And then, the partnership, one again would need to know exactly what is the nature of the partnership and what is one entity providing to the collaboration to be able to provide more insight or more assessment of that.

CR: So, and lastly, based on the, you know, GND-323's discovery status, the undisclosed target interference, the iGlue mechanism, their AI platform and current pipeline breadth, what evidence would move Gandeeva into further diligence territory versus just a watchlist or even deprioritization?

PL: I'm sorry, would you mind, I missed the first part of your question. Would you mind repeating?

CR: Right, so this is just an overall summarization of what we discussed. So, you know, GND-323 is a discovery stage. We don't have a lot of data about it. The target or interface is not disclosed yet. The mechanism of the iGlue is not quite clear. They have an AI Cryo-EM platform, basically a combination of that. So what would move this company into further diligence territory versus deprioritization?

PL: Got it. Well, a couple of things. The most important thing to know is to know what the target is and what is the glue partner and how is this glue being achieved and is this a glue that is just a protein-protein interaction glue or is this a glue that is more sort of like degradation or things of that nature? But knowing what the target and the mechanism of the glue are would be very important. Then the second consideration would be for the specific target. Does the glue complex affect only the mutated form of the target or both mutant and wild-type form? Of course, if it affects only the mutant and spares the wild-type, that could go a long way towards a therapeutic index down the line. If, however, it affects both mutant and wild-type, or if the target is not mutated at all and it's just downstream of a mutated alteration in cancer, then the consideration becomes different, and one would need to know the dynamics of that effect to determine how exciting that is, and then the third component is again what other modalities are existing to target the same target. For example, if the glue is targeting something that doesn't have any other inhibitors available, the pathway and the level of excitement is different as compared to a glue that is targeting, let's say, RAF, MEK or RAS for which we already have agents available to us. The other component in the due diligence consideration that would allow one to determine whether this has potential or not is to understand what is the workup and diligence that Gandeeva has conducted in characterizing and validating some of their claims. Specifically with regards to the glue, do we understand, do we have a structural understanding of what the interface is where the molecule binds? How many analogs have been explored to understand the structural activity relationships and do we have a sense that Gandeeva has maximally optimized the molecule to have the best possible, the most potent formation of the ternary complex? Do they have an understanding of the various kinetics involved in the complex formation? How selective is the complex formation? And then in preclinical models, how potent their anti-proliferative and anti-tumor effects are in cell culture or in in vivo models, to what extent xenograft models, patient-derived xenograft models have been used or patient-relevant models have been used, and so on and so forth. And of course, if they have clinical data that defend the rationale, especially if this is a previously untargeted target, I think all of those. If one puts them all together, one can come up with an assessment as to whether a company like Gandeeva would merit further consideration, the level of excitement of such consideration, etc., etc. So anyway, it is... to determine whether it has potential or not cannot be judged simply on what is the big picture overview. We need to know the specifics and we need to know how well and how strong the evidence is to support the specific context of the glue. Does that make sense?

CR: Yes, it does.

PL: Okay.

CR: So Dr. Lito, thank you so much for joining us today. It was great having you on the call, and thank you so much for answering all my questions well.

PL: No, thank you as well so much, and I'll be happy to connect again in the future because this sounds like an exciting company and I would want to hear a little bit more about exactly what it is that they are targeting and how, and I would love the opportunity to do a deep dive on their platform and provide some more meaningful impressions on their program. But anyway, thank you so much for chatting with me and I appreciate your time.

CR: Sure. Thank you so much and have a nice day.

PL: Thank you. Bye.

DISCLAIMER: The statements or opinions expressed today are those of the Advisor and not Guidepoint, who disclaims all liability for the content provided. The Advisor may not disclose material nonpublic or confidential information or any information that would cause the Advisor to breach any duty or obligations. Guidepoint is not a registered investment advisor and the information provided is not intended to constitute investment advice.