Thanks for joining us to have a conversation with Sean McClain, CEO, and Zach Jonasson, CFO and CBO of Absci. Absci is a public biotech company focusing on developing novel protein therapeutics by harnessing the power of generative AI. Empowered by its proprietary integrated drug creation platform, Absci has not only established collaborations and partnerships with leading pharmas and institutes, but also has developed its own pipeline focusing on cytokine biology. The company's lead candidate, ABS-101, recently entered IND enabling studies. And to discuss the company's development strategy in 2024 and beyond, we welcome Sean and Zach to this fireside chat. Good morning, gentlemen. Glad to see you both, and appreciate you accepting our invitation to talk to our audience this morning. Yeah, absolutely. Thanks for having us here. So, just to get started, Zach, can you give us an overview of the Integrated Drug Creation Platform and what its abilities are? Yeah, I'm happy to. You know, at Absci, we refer to ourselves as a data-first AI company. So our journey really begins with creating the data sets that drive the AI platform. And so the company's history is really in synthetic biology, where we've developed these very high-throughput systems to produce antibodies in a highly engineered E. coli cell line. And we've leveraged that to make high-throughput assays where we can look at the interactions between antibodies and a given target epitope on an antigen at scale. So we can, in a week's timeframe, look at 3 million unique antibodies and look at the binding profile to a given epitope. So we use these assays to generate large training data sets, which we use to train the AI models that we develop. And then on the back end of that, we look at using those assays to then validate what the AI models have predicted. And so we run this cycle through about a 6-week timeframe where we're generating training data, developing models, and then testing and validating the models. And each time we run that cycle, it's iterative. We're generating more data and improving the models. So feedback from the validation goes into improving model performance. And we create more and more training data as we go. And we create this training data both just as a platform effort, but as well as around every program that we work on, whether it's an internal program or a partnered program. As you can see, as we go through time, we're constantly improving our AI models' capability. We've seen that over the past few years, the efficiency gains that we've generated, and I'm sure we'll talk a little bit about some of our lead program assets, but being able to create best-in-class assets in a 14-month timeframe to get to a drug candidate is, I think, a pretty remarkable milestone for the company. Fantastic. And then, Sean, a couple of years ago, Absci actually published on the Zero-Shot Generative AI model. How does that differentiate itself versus other prediction models which are out there? And in terms of validation, can you talk towards how the Zero-Shot Generative AI model has been validated? Yeah, absolutely. So this Zero-Shot Generative AI model essentially starts with the structure of a target. This can be either AlphaFold or a crystal structure. And then what we're looking to do is then define the epitope that we want it to bind to. And then we're able then to have the model predict the CDR regions in the antibody. Think of these CDR regions as almost like the fingers that reach out to bind to the target. That's the variable region in the antibody. And it's able then to predict the CDR regions that can bind to the target of interest. And so this gives us epitope specificity in a way that you haven't been able to see before. And this is really what's unlocking new biology and allowing us to partner with folks like AstraZeneca, as well as Merck and Almirall. Fantastic. And then the other ability that Absci has is the proprietary E. coli-based synthetic biology platform that you have had for a while. So how does Absci really take advantage of that? And also, can you talk through your wet lab ability as well? Yeah, absolutely. So what the E. coli platform and screening platform is allowing us to do is get data for the training, but then also the validation, as Zach was talking about. So this is the data that's going into our de novo AI models. These are looking at protein-protein interactions. So how the antibody interacts with the target of interest, the epitope, as well as the affinity, kind of the two key attributes for the functionality of an antibody. And we're using that data, again, to train, but that same exact technology then is actually used to validate these models to see how accurate they are. And it helps us refine what model architecture should we be using? What data sets do we need to introduce to increase overall model accuracies? What hyperparameter tuning do we need to do? We can validate over 3 million unique AI-generated designs in a given week. That whole process, from start to finish, from the data generation to the AI training to the validation, is all done in a 6-week time period. This is really what has allowed us to have the success that we've had and essentially allow us to almost act like a tech company within biotech on how quickly we're able to iterate on these model designs and architectures. Again, that's what allowed us to or led us to have the breakthrough that we had on this de novo AI model. The validation that we showed on the de novo model was we applied it to HER2, which was a known target. We were able to show the epitope we wanted to bind to. Indeed, we were able to have the model design the CDRs from scratch. We took it one step further this year and applied it to our own pipeline, TL1A, where we were able to, for the first time ever, actually show how generative AI and this de novo foundation model was actually able to not only design an antibody from scratch, but it was actually able to generate a differentiated antibody that was able to be generated in about a 14-month time period. It was 14 months to get to a drug candidate. And then we're on track to have it in the clinic within 24 months. And so this was kind of really taking this hype of generative AI and turning it into reality and actually showing that you can use generative AI, you could use this de novo foundation model to create a differentiated asset. Ultimately, that's what pharma cares about. That's what our investors care about, is how can you use a tool like generative AI to create differentiated assets that have new novel biology? And then how do you leverage it additionally to be able to get into the clinic faster than you would have with traditional methods? So talking about value creation, which we investors are always interested in, Absci has clearly shown value creation, especially over the last 3-6 months, in terms of signing major partnership transactions with Almirall, PrecisionLife, and AstraZeneca. So in general, what are these three partnerships I mean, talk to in terms of your ability? And how do you differentiate between the three partnerships? I'm assuming they're going to telegraph different abilities of Absci between these three different relationships. So if you can enumerate that and highlight that for us, it'll be helpful. Sure. I can take that. I guess at a high level, I would say these partnerships to us are fairly important. They provide non-dilutive capital, which was always welcome. But I think even more than that, they're really synergistic partnerships that display the diversity of our platform. So we can apply our platform to drug candidates, developing drug candidates, in a variety of indications. So if we look at the AstraZeneca deal, that's a roughly $247 million deal if we sort of put together all the fronts, research support component milestones, both development and commercial, but not including royalties. And that is really focused on generating a drug candidate towards an oncology target that has been very difficult or easy to drug. And so this is an example of us deploying our platform against a membrane-bound protein that, again, traditional approaches cannot address in leveraging all of the really strong expertise that AZ brings to the table in oncology and also the biology around the target. Similarly, with Almirall, which is a pharma company that's exclusively focused on dermatology, there we're focusing on, again, difficult-to-drug targets, but leveraging a lot of synergy with Almirall, given their expertise in dermatology and the biology of these targets. So when we look for partners like AZ and Almirall, we're really looking for partners where we see strong synergy that can be leveraged both with our platform and the expertise that the partner brings. The other thing I would just comment on is just picking AZ to highlight for a moment. AstraZeneca has invested heavily in its own AI capability. Their selection to work with us was the culmination of a year and a half of diligence on the entire space. So we were very happy for that stamp of validation. I think we're very excited to embark on that partnership. We officially launched that at the end of last year. Yeah. And additionally, I'll add, if you look at AZ and Almirall, why did they end up selecting Absci? And the ultimate reason was we were able to provide a technology and ultimately a solution that themselves, as well as other companies, could not provide. And this was, again, as Zach highlighted, going after these undruggable targets like GPCRs and Ion channels. And why are those traditionally hard to drug? Well, there's not much surface exposure of these on the cell surface. And so the immune system actually has a hard time generating antibodies in a typical immunization campaign. But again, being able to be epitope-specific with that de novo model, I can feed that structure into the target and specify that very little surface-exposed membrane target and generate an antibody. Now, at that point in time, you actually unlock new biology that you wouldn't have had access to before. That's why pharma and others are so excited about this. It's creating the ability to create new novel biology and create ultimately differentiated assets. Fantastic. So talking about generating new biology, right? So I was just thinking of this. So how we have seen biotech progress in terms of each of them trying to form a little silo of themselves, whether it is oncology, neuroscience, or whatever. So do you see TechBio or Absci sort of companies also go down that route in the sense, just as you talked about GPCRs, would there be a silo of companies that would be just going after very hard-to-get targets or target areas? Would that—do you see that happening? Or because this is kind of a TechBio and it only depends upon how powerful your tech is going to be, is where this is going to go, just like the Intels and the Microsofts of the world? Yeah. Look, you can have a technology that unlocks all this new biology and all these different therapeutic areas. But at the end of the day, to scale the domain expertise on the translational side and the clinical side in each of these areas for one company at our stage just isn't feasible. And I think going back to partnerships and why we want to partner is actually being able to access that domain expertise. If you look at Almirall, I'd argue that they have a very, very strong presence in dermatology. They know, they understand the biology. They understand the targets that are interesting. AstraZeneca with oncology. And so we're able, with these partnerships, actually to access domain expertise that we don't have internally. And so therefore, it actually allows you to create a portfolio of assets that you couldn't have generated any other way. And so that's why this focus on both internal pipeline as well as partnerships is so key to us because we can use the technology to unlock these new differentiated biologics without having to scale the expertise internally. We can partner with that. And over time, you continue to gain some of this expertise yourself. But that's why I think these partnerships are so critical in helping us create that diversified portfolio of drug assets. Yeah. And if I could just add to that, I mean, the really exciting thing about the platform we're building is it's really indication-agnostic. It's really target-agnostic. So we can work on targets for a whole variety of indications. And that's what we like to do through partnerships to give us that diversification. But when we want to build our own internal program, then we leverage the experience that we have deep in the biology and the target, as well as the indication area. So in a way, we're really maximizing how we leverage this platform by doing partnerships with companies like AZ who are experts in oncology, Almirall experts who are experts in dermatology. And then we focus in on the areas for our internal pipeline that are around cytokine biology where we have in-house expertise. Yep. Yeah. Speaking of your own internal pipeline, it was last year. Can you talk about your business model of how this internal pipeline can create value for Absci? Yeah, absolutely. What our plan with our own internal pipeline is to take our assets to value inflection points in the clinic, whether that's entering into the clinic with an IND, phase 1, or phase 2. Ultimately, at any step along the way from a drug candidate through phase 2, we will be open to partnering those assets. But we do believe that with this AI technology, since we're able to create differentiated assets, be able to get to these value inflection points in time faster and cheaper than ever before, it makes a lot of sense for us to maximize the asset by taking it deeper into the clinic ourselves. And with TL1A or any of these other assets, we're fully prepared to take it deep into the clinic, but we will not take it into late-stage clinical development. We do believe that I mean, there is an ecosystem that's out there. Let's leverage that ecosystem. Let's leverage partnerships where large pharma, they are great at late-stage clinical development, commercialization, marketing. So if we can take it to a point in time where we hit value inflection points from a proof of concept standpoint and then partner or license it out, we can really maximize the value of the asset versus partnering super early on. Again, I think you're able to do this now in a very cost-effective manner because of how rapidly you're able to get to these value inflection points and how cost-effectively you can get there as well. Yeah. I just sort of come back to just looking at the economics of it and why there's such a tremendous ROI to do this. For us to get to an IND, and we're demonstrating this now with our ABS-101 program, is roughly 2 years. It's going to come in at about $12-$13 million. If you look at what pharma does to get to an IND, that's usually $50-$70 million. It's 5 and a half years. So for the same budget that pharma has, we can run 4 different programs to an IND. So really highly efficient asset generation using this platform. So to develop these assets and take them into the clinic is very inexpensive and very capital-efficient for us. The payoff is very significant. As we develop those clinical proof points, the partnering and licensing terms go up dramatically. Yeah. Got it. Thanks for that. So at JPMorgan Conference this year, you showed highly encouraging data on ABS-101. For investors who are not familiar with those data, can you give us a high-level summary of the drug's differentiated drug profile compared to its competitors? Yeah, absolutely. We do believe that TL1A has the potential to be a best-in-class. And what we've done is we have engineered the Fc for extended half-life so we can go from potentially once monthly to once quarterly dosing. Additionally, we have hit an epitope that we believe is actually going to bring in or show lower immunogenicity or have lower ADA response. It's a hypothesis of ours based on the clinical data that's out there. If you look at the Roivant molecule, you have lower or there was pretty high ADA response for the Roivant molecule. And with the Merck molecule, there's a lot lower ADA response showing lower immunogenicity. And we believe that this is actually complex-driven or B-cell mediated versus T-cell. And so the epitope that you hit is really important for ensuring that low immunogenicity. And so we ended up hitting an epitope that was similar to the Merck epitope. But additionally, though, we wanted a molecule that had higher potency or superior potency. And we did indeed have designed a molecule that hit that epitope but also had superior potency when compared to the two competitor molecules. And so all in all, we are showing superior potency. We'll have that extended half-life for increased dosing intervals and the ability to potentially have lower immunogenicity when compared to one of the other competitor molecules. And I think in a chronic illness like this, like IBD, that's going to be really important. And again, that highlights the capabilities of our de novo foundation model, being able to design those antibodies from scratch that have that epitope specificity. And again, being able to create that differentiated biology is a really important aspect. I think this is a really great proof point for this, but also is a target that when we're developing ourselves, I think, has a lot of upside with it as well. So we're really excited to be taking that into the clinic early next year. Our kind of roadmap right now is that we've entered in IND-enabling studies. We'll have those wrapped up by the end of the year. We'll be presenting NHP data on that this year, along with our clinical trial roadmap and plan. We'll enter the clinic early 2025. The plan is to have an interim phase one data readout in the second half of 2025. Got it. Now you have achieved this significant milestone for the company. Any other pipeline updates investors should be looking out for in the next 12 months, maybe? Yeah. So the other two data readouts that we'll have come from ABS-201, which is a potential best-in-class derm target. This is a target that we think is underappreciated. And we will have a drug candidate on that this year, with data being shared shortly thereafter, very similar to how we did with TL1A. And we then have ABS-301, which is a novel I/O target. It's a first-in-class asset. And we have done in vitro validation on that. And we're in in vivo validation currently, and that in vivo validation will be done this year with a drug candidate shortly thereafter. And so we'll have data on both ABS-201 and ABS-301, as I had outlined there. In the last 40 seconds or so that we have here, Sean and Zach, just a quick commentary from you of how you're managing your resources between generating models, creating partnerships, as well as developing your own pipeline. In terms of cash position, what sort of a runway you have? Sure. I can take the last question first. So as you know, we just raised the financing last week. And so we raised $86.4 million gross proceeds. That gives us a cash runway into the first half of 2027. So if you back out kind of what Sean was talking about in our development roadmap, we'll be able to achieve phase one completion and data readout on our ABS-101 program. We'll also be able to are focused on delivering two additional INDs as well as some additional drug candidates. So we have, I think, a pretty rich pipeline catalyst calendar coming up. And then in addition to that, we'd be looking to add on more significant partnerships with pharmaceutical companies. Fantastic. Any last-second comments from you, Sean, as we close out? Yeah. This is an AI drug discovery fireside chat and an event that you all put on. I would say that I'm extremely bullish on the power of AI and what it's going to do in this industry. I think that we have demonstrated for the first time, especially in biologics, that we've gone from this AI hype to reality. What is AI actually going to do? It is. It is going to create differentiated assets faster and cheaper than we've ever seen before. We're really excited to be pioneering this. I think that you're going to see a lot of exciting outcomes from generative AI when applied to drug discovery and development. Yeah, we're very excited to be paving the way forward here. Yeah. We are excited as well. It looks like 2024 is, I guess, the start of the new AI era in drug discovery, at least as far as the amount of noise that's being generated this year, which is really, really good for the space. Absolutely. Good luck. We certainly will be talking to you soon. Thank you very much for joining us in this conversation today. Yeah. Thank you so much for having us here, RK. Thank you.