Generate Biomedicines, Inc. (GENB) June 2026 Conference Transcript

Good afternoon. Thank you, everyone, for joining us. Really pleased to have with us the Generate team. With us, we have Michael Nally, CEO, and Jason Silver, CFO. To start here, Mike, would you introduce Generate, your platform, your clinical programs, and the catalyst path for the next 12 to 18 months?

Yeah, thanks, Alvin. It's great to be here and excited to share a little bit more about Generate and the story. So Generate's a company that was founded in 2018 with a fundamental premise that, you know, a data-driven approach to protein design could lead to superior outcomes. At the time, the prevailing paradigm was largely biophysical-based tools to design proteins. And, you know, it was a convergence moment in the field where, you know, the Transformer paper was coming out of Google at that point. We saw masses of increases in compute. and two foundational data sets, the Protein Data Bank and the companies like Illumina that had sequenced every species on the planet, allowed us to start to understand sequence, the structure, and our relationships. And the company was founded to add a third leg to that stool, which is function. And so the company operates with a very tightly integrated dry and wet lab to generate experimental data at a pace and a scale that was previously unimaginable. feed that data back to train models to design novel proteins that nature hasn't discovered. And so over the past, you know, seven years now, we've been able to put five molecules into the clinic. The productivity of the platform is pretty extraordinary in terms of the efficiency in which we can find new proteins, but also the scalability of drug discovery. Our lead program is an anti-TSLP antibody that entered Phase 3 in December of last year. It was designed basically five years ago today, and so the ability to go from kind of concept to Phase 3 in less than five years is, I think, pretty impressive for a chronic care program. And the advantage is it binds the target with exquisite affinity, So 100 femtimolar binding, you couple that with an extended half-life, and we believe we'll have an every six-month dose anti-TSLP antibody. We're looking at it in severe asthma in the first instance. We'll have regulatory interactions over the course of the summer in COPD, and so we're excited by that program. We then have a couple of oncology programs that are just entering the clinic right now, the first of which is a payload neutralizer program. So this is a kind of a unique concept where we've seen enormous advances in antibody drug conjugate technologies, and this molecule selectively binds the cleave payload of MME-based ADCs to hopefully reduce the rate of peripheral neuropathy. In the PAD-SEV phase three trials, you saw about a 65% rate of peripheral neuropathy, Much of that leading to either discontinuation, down-dosing, or dose stoppages. And it is a clear in terms of the adoption of these ADCs, especially in the community setting. And so by selectively binding, we believe we'll be able to clear the toxic part of the regimen while not interfering with the efficacy in a very targeted fashion. So we think it will open up the therapeutic window in a meaningful way. The next program is a CAR-T program. So we're working with the Roswell Park Cancer Center. They have been pioneers in a number of different CAR-T domains. They hired a key scientist from Sloan Kettering, Rainier Brenchens. Rainier had asked us, could we use our technology to optimize the binding and the framework for CAR-Ts in solid tumors? and we will be testing what we believe will be one of the most potent CAR T's on a target in ovarian cancer, MUC16, imminently. And so, you know, what we're seeing with the technology is a huge breadth of potential impacts across these different domains. And I think we're just scratching the surface of the potential of the technology where we're, in the first instance, using it to optimize existing molecules where we think the technology will ultimately have greatest impact is being able to drug on druggable domains with biologics.

And maybe just stepping back with regard to the platform here, we heard the term AI as an umbrella for different strategies and technologies, particularly in healthcare. What do you see as the key differentiator of your platform versus traditional companies or traditional computational biology and those employed by other companies in this field?

Yeah, I think maybe I'll start and then Jason, please, you know jump in I think there's a lot of noise in the field I think you know many people are declaring that AI will be a panacea for drug discovery and development and well we think it will be transformational in certain in biology and so with a great degree of humility but where we see the real impact and you know what we're seeing very tangibly at Generate is we now are able to design molecules that were almost impossible to discover using kind of immunization campaigns and as you all know you know when you the way we've historically found biologic agents is you either immunize a human or mouse or llama and then you try and manipulate that protein to have drug-like properties what we're seeing with these technologies technologies is an ability now to de novo generate desired protein sequences that perform in a way that we want. And so we think we're migrating from what has been an artisanal craft of drug discovery, borrowing from nature, to an endeavor that is much more based on programming biology. And I think the key for us has been this tight integration of experimental capability and computational capability. And if you look at and you think about the domains of generative AI that have had the greatest impact to date, there's a reason Demis and Cebus focused on gaming first. There was a rapid verification engine with gaming where you had a scoring mechanism that allowed you to continue to train the model. Coding, math, both have rapid verification engines. Biology, you know, at its core, doesn't have that verification engine, and we don't have the Internet to train models on. And so having this tight intersection of experimental capability with cutting-edge computational tools allow us to have that verification engine to really refine the answers and get to answers that were previously almost unimaginable.

I would just add a few things that are differentiating, I think, for us. Number one, we're large molecule only. Two, we're modality agnostic. So this is not just an antibody company. We're working across eight different modalities in terms of what our products are, including with our partners in Amgen and Novartis. And also we've spent a lot of time and effort and money on cryo-EMs, so looking at structure, not just experimental verification of what comes out of our machine learning, but also we can look structurally at protein-protein interaction space down to a few angstroms or even lower than that, so at an atomistic resolution, so that we can not only look at assays and functional and biophysical, but we can actually look at what's happening at the binding site. And that's real-time, so very high throughput. So maybe a couple years ago, it could take three to six months to actually create the structure from using cryo-EM. We're now solving at least a structure per day, per scope, and we have four microscopes, which is pretty unique in the industry.

Maybe starting with the lead program here. So GB0895 is in Phase III development in severe asthma. Speak to the bar that you hope to achieve on exacerbation rate reduction in the overall population and the low EOS group.

Yeah, it's a great question, Selvene. You know, if you think about respiratory disease, biologics still aren't standard of care. We've obviously seen the first-generation biologics drive about a 20% uptake in the severe asthma population, but the vast majority of asthmatics that have severe disease still don't have a suitable alternative to ultimately control the deleterious effects of the disease. And so for us, we think T-cell has shown enormous promise in the domain. What we've seen and what we believe our molecule will ultimately show is that we'll be able to dose this almost in a prophylactic sense, where you'll give an administration every six months. And on a blended basis across all EOS levels, we think the efficacy will be in that roughly 55% range. The way that will balance out is in the high EOS population, which is where molecules like the IL-5s are dosed. We'll see roughly a 70% reduction in exacerbations. And the trial is specifically powered to detect a 40% reduction in that low EOS cohort. And one of the inherent advantages of T-slip has been the ability to reach those of that low EOS population. And so we think both in the design but also based on the qualities of the molecule that those sort of rates are very achievable.

And speak to what gave you enough confidence to advance directly from a phase one to phase three.

Well, I think part of it is what we saw in the phase one data. And we designed this molecule with purpose. I talked a little bit about programmability a moment ago. We had a certain set of specs that we wanted to make sure that this molecule reached. And when we revealed the phase one data, it basically hit on every one of those parameters. We always thought this was a potential pathway, and largely we were inspired by the work at GSK with Depomocumab. So we certainly want to give due credit. GSK was really clever. They went into mild to moderate asthmatics in phase one and then used that data based on the ability to effectively modulate the biomarkers to go directly to phase three. And so we know this is now a validated pathway. The Depomocumab has been approved in the U.S. and Europe. We went to the same sites in the U.K. and Germany. We wanted to make sure that we were not only showing that we were safe, that we were, you know, the ADA rate was low. We also wanted to show that we were modulating the key pharmacodynamic biomarkers in a relevant fashion. And what we were able to show in that study was that we were able to modulate IL-5, IL-13, EOS, and Pheno at a level that was commensurate with, if not slightly numerically higher, than what we saw with tezapelumab. When you couple that data with the fact that you can effectively model target occupancy at our 300 milligram level versus what you see with tezapelumab at the marketed dose of 210 milligrams, we were able to show over the course of six months that we have the same or better target occupancy at all time points to tezapelumab. And the third piece of data that was important in this choice was actually based on the TESI phase two data where they did their dose finding. They explored an eightfold range of doses, and they showed no difference in exacerbations across that eightfold difference. And so for us, you know, it came down to a pragmatic choice. We could have run a phase two study, looked at, you know, multiple different dose levels, showed the same exacerbation reduction rate, and then made a pragmatic choice. That would have cost us $50 million and spent two years, or based on our ability to modulate the key biomarkers in a commensurate way, the ability to show that we're engaging the target at the same or better level while binding the same epitope, and based on the TESI data on a pragmatic choice, we said we're better off just making that pragmatic choice based on that data.

And critically, we hit the same epitope as tecipelumab, which is a big, important point for the FDA and other regulatory agencies in allowing us to go to phase three.

As we think about the potential to improve upon test buyer here, speak to, so clearly there's biannual dosing, and speak to if that's just the improvement, what physicians are saying about uptake and if they're comfortable with long-term use of an agent. And then secondly, is there any potential to improve upon efficacy?

So, I mean, I think when we started the program, and this goes back to the earliest days, what we were able to show in the preclinical models was a 20-fold improvement in affinity and a 5-fold improvement in potency. The question always has been, you know, how well does that potency marker translate into humans? What we're seeing based on the phase one data is relatively a commensurate reduction in the biomarkers. And so we think the base case should be comparable efficacy. But until you run the studies, you never know, and maybe that five-fold improvement in potency preclinically will translate into something. But I think the base case certainly should be comparable efficacy. I think if you look across all immunological conditions, and we've seen this now as these immunological markets have matured, you see these first-generation agents oftentimes be shorter duration, and then over time you migrate to longer-duration agents as the market becomes more sophisticated, as patients become more demanding around the types of therapies they're looking at. And, you know, the psoriasis market is a classic case here where now you see somewhere between a 40% to 60% penetration with biologics, and the long-acting biologics are the disproportionate beneficiaries in those sort of spaces. And so we think asthma will evolve in a very similar fashion. One of the things that's really important to note is, you know, severe asthmatics visit their doctors twice a year. And if you look at the current compliance records, and GSK presented some beautiful data at ERS last fall, only 20% of patients on biologic therapy are maximally compliant. And so the ability to drive a better level of compliance, given the fact that you're dosing every six months, but also you're dosing in alignment with the current doctor's visits, we think could ultimately have a huge real-world benefit of these sort of therapies. And we saw this, if you go back to the osteoporosis market, and you probably remember when Prolia was launched, right? The bisphosphonates were great, and Prolia was not able to show a difference in hip fractures in the Phase III protocol. But as soon as you got to the real world, we saw a marked improvement in these sort of key markers. And so we believe over time, GB0895 will not only just have a convenience benefit, but will be cost-sparing to the system.

Got it. As you mentioned, in COPD, you plan to report full Phase 1B data this year, and we'll also understand your registrational development pathway. At this point, are you leaning more towards a Phase 2, 3, or a Phase 3?

We're certainly going to propose to the FDA that a Phase 3 is the appropriate next step. So we'll meet with the FDA in the coming weeks. And, you know, the range of alternatives, you know, reach from a single phase three trial of a single dose to a 2B3 design. And, you know, I think, you know, given the changes at the agency, it's a little bit hard to know what is the preferred stance. Clearly, with the change in leadership, MACRI had been a big proponent of the single phase three trial. In that case, what you would do is you drop the P value down to the 0.01 level. You'd do about a 1,200 patient or subject study, and we think that could be a valid path, but ultimately that will require regulatory alignment, and so we'll get that alignment in the next few weeks. And then, as you rightly point out, we'll look to maybe top line the data around the second quarter earnings time and then present at an appropriate medical meeting in the fall.

And how much time would that save you? And I guess, secondly, your confidence that you'd be appropriately suppressing the target at this dose?

Yeah, so, again, if you go back to the data that we've shared to date in COPD, what you're seeing is a comparable reduction in the relevant biomarkers between the asthma and the COPD study, right? So, you're seeing roughly this 50% reduction in a number of these key biomarkers, less so on phenol, but you're seeing kind of the similar reductions. You're not seeing any improvement when you go up to 600 milligrams, importantly. So the lines are relatively... When you couple that with the fact that we are, you would be left on the table. But that ultimately will be the conversation, and that will be what, you know, the phase three study will be ultimately to elucidate.

And you're probably talking about saving two years if you go directly to phase three versus have to do a phase two, three trial and a substantial amount of cost.

Can you discuss the target population in COPD and your confidence on TSLP and the mechanism in this particular indication and how it differs from the two approved antibodies, Dupixin and Nucala?

Yeah, so I think we were really encouraged by the TESI phase 2 data in COPD. What that data was able to show was a significant reduction in the greater than 150 EOS cohort. It didn't show a therapeutic effect in the less than 150, and, you know, the percentage reduction in the greater than 150s was about 37%. In the greater than 300s, it was about 46%, albeit with a small n. What you've seen with DUPI and with NUCALA has been, you know, DUPI has been around a 30% blended exacerbation rate reduction. NUCALA has been roughly about a 20% reduction. So if you could replicate what TESI has shown in the Phase 2 trial, we think you'd be positioning toward a best-in-class advocacy profile. I think the one unknown is the TOZO data that AstraZeneca will show with their IL-33. What's been encouraging, I think, for COPD patients is it's the first agent that has shown an effect in that low EOS cohort. What the magnitude of effect is, we'll have to wait and see. Obviously, we've had other IL-33s with slightly different approaches not yield consistent results. And so I think it's a great thing for patients that, you know, the TOZO data will come out and, you know, we'll obviously have to look and see where that efficacy bar is set based on their presentation at ERS.

For your MMAE targeting antibody, when do you plan to present data next year? and why do you see 50% reduction as kind of the key window here?

So 50% reduction is important, and frankly, the FDA and our conversations with them even encouraged us to potentially go up to 80% reduction. The reason we're targeting 50% reduction is in our preclinical data, we've shown that at 50% reduction, you actually retain the tumor killing but get decrease in skin toxicities and others in mice. an example, and in non-human primates. And so for us, that 50% level we think is the right level with flexibility to go up if we need to. The trial design for phase one, which we dosed our first patient actually this morning, and all the sites are activated for that trial. The trial design is we'll first do a dose escalation in terms of what dose will give us that 50% reduction and free MMA. Once we determine that, which we believe will be probably close to the end of this year, maybe it falls into 2027, we will then open a cohort of patients who have grade one peripheral neuropathy on pad seven Keytruda, give them our MMA neutralizer, and see if we can stop or slow the progression to grade two or irreversible peripheral neuropathy. We haven't made a decision in terms of which data, if any, we would present first, whether it would come in steps, or ultimately we'll present the whole package. We're pretty confident that we should be able to be in a position in 2027 to present all of those data, but we'll make that decision in terms of whether we present the dose before we actually move to the second cohort of the trial or we present all at once.

And, Salveen, just one point, like when you look at that 80% threshold that Jason alluded to, above 80%, you do disrupt the bystander effect. And so part of what you're trying to do is find how do you open up that therapeutic window in an optimal way so that you reduce the rates and the burden of neuropathy and some of the other negative side effects associated with infortimavidotin while make sure you're optimizing for tumor killing. And so we think the ultimate path will require us to not only show the reduction in neuropathy but also show a non-inferiority on overall survival. And so, you know, that kind of is the blend, and that's why this window of 50% to 80% becomes so important.

How much of a reduction in neuropathy would you expect at about 50%?

It's a really good question. I think there's two parts to that, right? So there's the rates of neuropathy, and so we think you could expect somewhere upwards of 50% reduction, but it's also the progression of neuropathy, right? So once you get to grade two neuropathy, you have irreversible side effects. And so part of what we're trying to figure out, especially as Infortumab-Vidotin continues to post really stellar data in a number of different tumor types, but also in healthier patients, so in the adjuvant and neoadjuvant setting, the risk benefit of irreversible side effects actually changes as you go into earlier populations. And so for us, it's not only just the absolute reduction, but also the ability to slow the progression or stop the progression so that you can maintain, you know, on drug for longer with these sort of agents.

In addition, the two other important pieces, one, in the community, you're seeing probably less than 50% of patients with urethelial cancer get treated because physicians are less comfortable dealing with the side effects given all the different toxicities. So that's kind of point number one. And then, two, in the academic center with 65% peripheral neuropathy, it's not only reducing the peripheral neuropathy. Ultimately, as Mike mentioned, if we're looking at non-inferiority, there's been some recent data to suggest that patients on PADSF get true for a longer period of time actually convert from PR to CR. If that's true, then by giving our reducing toxicity medicine, that could enable the window for patients to receive the drug for a longer period of time and open that therapeutic window and ultimately lead to a better efficacy for them.

Right. And you have your own ADC preclinically that would go into the clinic at some point. Do you have any updated timelines around that?

Well, so this is based on the observation that our technology allows you to optimize for internalization. And so there are a number of antibody drug conjugate targets that have been materially overexpressed on cancer cells and not expressed on healthy cells. And, you know, a number of these sort of first-generation attempts have struggled because they don't get enough of the payload into the tumor. Sometimes it's been, you know, kind of perceived to be, oh, well, that's just a poor internalizing target. What we've been able to show now in the preclinical setting with our technology is that by modifying the CDRs alone, you can drive logarithmic improvements in internalization. And so that opens up these kind of, in some ways, partially validated targets to a better internalizing antibody drug conjugate. And so we have one target that is kind of approaching development candidate nomination. We have a second target that is in earlier stages of exploration. But we think, based on this technology of being able to optimize internalization, there a whole array of not only antibody drug conjugate targets, but also potentially other protein-based therapeutics where getting into the cell in an optimal way becomes, they become more viable with the technology.

Jason, perhaps speak to the partnership outlook here for your overall portfolio.

So we have, as you know, partnerships today with Amgen and Novartis on the corporate side. We have MD Anderson and Roswell Park on the academic center side. The Amgen and Novartis programs are really getting toward the tail end of those programs. We've made an enormous amount of progress, frankly, even in the last year in both of those programs and the outstanding programs. And so we think by the end of this year or so, early 27 or end of this year, we should be sunsetting those programs, having met the criteria that we believe will enable us to receive additional milestones in both those programs. And those, as you'll recall, are complex in terms of what we've been asked to solve. These are more than a decade, in some cases, outcomes that are either measurable or being able to test biology just because traditional techniques are unable to get to the result they want. And I think our technology has enabled us to find those results, and, frankly, it also gives us an opportunity to utilize those technologies in other areas outside of the specific target that we're working with those two companies. We will and are continuing to explore additional platform-type collaborations like those. There will be target-specific collaborations. I think over the course of the next, you know, 6 to 12 or 18 months, we certainly will be in a position to announce at least one more of those, and we'll continue to pursue similar. Other collaborations that we're exploring and, you know, potentially would do are much larger collaborations that utilize either a specific technology with broader applications or technology with broad therapeutic applications, maybe in areas outside of our core focus, which, you know, right now we're focused. We've been on infectious disease, immunology, and inflammatory diseases. So you can think about cardiometabolic or neurology as an example. And our technology that we're working on today, much of which we've learned through doing work with our existing collaborations, will enable things in those domains in particular, like crossing barriers, pH-dependent binding, increased internalization, things like that. Other things, we have other programs that we have decided to deprioritize, just given capital constraints of the other existing clinical assets. So we certainly would be in position and consider whether we would do licensing transactions or some kind of collaboration around those as well. Those are the core collaborations.

The other thing, Salveen, though, I think if you step back and you think about the technology, why we are so excited about it is partially because drug discovery has been this artisanal craft. And when you put the computer at the center of the discovery effort, all of a sudden you introduce scalability dynamics in discovery that have not existed before. Challenging ourselves on right now is we're able to put a number of programs into the clinic, but we also know that the technology could actually produce much more than we could consume ourselves. So the partnerships with Amgen and Novartis were the first step in terms of how do we realize additional value from the scalability of the technology. As we look forward, we think there are a number of other vehicles that we owe it, certainly to our employees, to our shareholders, to further explore. First is, are there opportunities to what we would call create modality intellectual property? So are there generalizable principles that govern certain biological functions? You know, one common example is the YTU mutation for half-life extension, but are there similar sort of mutations that would govern barrier crossing, that would govern internalization, pH-dependent binding? If you can find those, you can prove them and then enable the field in a very capital-efficient way. The other thing that we spend some time thinking about is could we use the technology to spin off independent companies? And so part of our, I think, responsibility is to sit there and say, well, given a unique level of productivity, how do we effectively monetize it in an optimal fashion without requiring all the capital in one entity?

That's the last question here.

How do you defend your algorithms from potential commoditization of AI over time? i think we've always believed that the algorithms will commoditize yeah and again i it's it doesn't mean to suggest that you don't need to keep pushing the frontiers of algorithm development but much as i would look at like the llms one day claude's the best the next day gemini is the best the next day chat gpt is the best um we think um you know the techniques that you lie behind the models will ultimately converge, but what's really important is how you combine those techniques with the experimental interface in a virtuous cycle of learning, and I think where we've been fortunate is that our computational scientists at Generate believe in the necessity of experimental verification. Our experimentalists believe in the potential of these sort of technologies, and so it's this intersection that I think is much more defensible, especially with unique and differentiated experimental capabilities.

Great. Well, with that, thank you so much.