Absci Corporation (ABSI) Earnings Call Transcript & Summary

Good morning, everyone. Let's go ahead and get started. Apologies for the slightly delayed start. We'll make it up on the back end. My name is Vikram Purohit. I'm one of the biotech analysts with the Morgan Stanley research team. Happy to have with me the team from Absci. But first, let me read a research disclosure statement. For important disclosures, please see the Morgan Stanley research disclosure website at www.morganstanley.com/researchdisclosures. And if you have any questions, please reach out to your Morgan Stanley sales representatives. I'll catch my breath. I'm running over here from the wrong room. Sean, thanks for joining us. Appreciate it.

Absolutely. Thanks for having us here.

A lot of interest recently on AI, tech-enabled drug development. Maybe the best place to start is just walk us through the thesis driving the formation of Absci and kind of a description of the key problem that you're looking to solve?

Yes, absolutely. Yes. So as you mentioned, we're on the intersection of generative AI and biology to ultimately be able to design better biologics faster. And if you look at most of the AI drug discovery companies that are out there, they're all focused on small molecules and there's a reason for that. It's because of the access to data. Anybody can go take a small molecule library, train a model and ultimately design drugs using generative AI for small molecules. But the data doesn't exist for biologics at the throughput and quantity that's necessary for training generative AI models. And that's where Absci comes into play. We were really a not AI-first company, originally. We're focused on selling [ development ]. Actually, we engineer the [indiscernible] to produce proteins more effectively. We ended up realizing that you could then use that to effectively screen very large libraries of proteins, looking at protein-protein interactions. And we use that data to train our generative AI models. And ultimately, kind of our big vision is being able to design a biologic at a click of a button. And what does that mean, basically being able to feed in a target structure into your Gen AI model and being able to then predict an antibody that can bind with the attributes that you want. It binds the location of the target or the epitope. It binds with the affinity that you want, so essentially, the functionality and then it has the developability and manufacturability. And what is this ultimately doing? It's ultimately increasing overall success in the clinic because you're able to go after these undruggable targets like GPCRs or ion channels that then have been hard to drug with traditional technologies. And it's dramatically shortening the time it takes to get into the clinic. On average, it takes 5.5 years. With our own programs we're developing internally, we're showing that we can get to an IND in about 18 to 24 months. So roughly decreasing that time by half. So again, focus is on biologics, being able to kind of go after these undruggable targets and then using Gen AI to also shorten the time it takes to get into the clinic.

Great. That's a good segue actually into my next question. So with all of the AI-enabled tech-enabled drug development firms, I think people try to always see how they're fitting into kind of 3 broad use cases, which you just alluded to. One is how can you cut early-stage development costs? Two is, how can you find novel targets or better drug known targets? And then three is, how do you increase POS? So you touched on this, but to put a fine point on it, is there a good example or two you can provide on how Absci has been able to kind of tick through those steps?

Yes, absolutely. So let's just take a GPCR. GPCR is an exciting oncology class of targets that is traditionally very hard to drug with immunization or phage display or use display because there's not much surface exposure of the target on the cell. Now with our platform, what you can do is actually feed that structure into the AI model. And even though there's a small surface exposure, there's enough to be able to identify that and design an antibody to that versus an immune system, it's hard to actually get there because of that surface exposure. But again, with an AI platform like ours, you can actually specify that specific epitope that is surface exposed and generate an antibody towards that specific GPCR, again, kind of unlocking new novel targets that have been known, but just hard to actually drug. Same is true with ion channels. And these are the types of programs that we're focused on, both with partners as well as internally. And this ultimately allows you to develop first-in-class assets. But there's also another use case for this technology, which is designing best-in-class or, let's say, a fast follower approach, being able to actually use the AI to get around existing IP and then being able to explore a larger search space you wouldn't be able to explore with regular approaches like immunization and be able to get patent claims that are new and novel while still having freedom to operate. And then additionally -- so that's the IP side, but then being able to then develop a drug that's more potent. It has higher binding affinity, higher potency and has the ability to be a best-in-class. And so these are kind of 2 examples of how you can use this type of technology to create first-in-class, best-in-class assets. And that's really what we want to turn into is really an IND machine where you're able to rapidly develop first-in-class, best-in-class assets.

Got it. Got it. Maybe then let's talk about your business model. You just mentioned partnerships, proprietary pipeline. Let's talk about partnerships first. How are those set up? And how do the economics from partnerships that you currently have in place flow to Absci?

Yes, absolutely. Zach, do you want to...

Yes, I'd be happy to. For discovery partnerships, we use a pretty traditional approach, right? We structure in upfront, research funding, milestones and royalties on the downstream. If you look at the Merck deal that we announced, you can see that play out there with a $610 million deal. So roughly across 3 programs, roughly $200 million per program in upfront royalties and milestones. And then of course, there's -- of course, there are royalties on the back end. I think it's important to note, too, when we set up these partnerships, the research work that we do on the front end on the discovery side is largely funded by the partner. And then if they progress those programs, then we enjoy the downstream payments that are essentially pure margin. So if you roll that strategy all up together, what we're really doing is building a portfolio, a diversified portfolio of these programs with partners where most of the upfront work that we're doing is paid for. But we have shots at -- multiple shots on goal.

Got it. Okay. And I think one natural question that comes up for kind of broad-based platforms like yours is, is there a specific therapeutic area? Or is there a specific molecule type that the platform is best suited for or have you not run into those kinds of limitations at this point?

Yes. The platform is really agnostic to all sorts of indications and therapeutic areas that you can go into. And really what we're focused in on with both our partners as well as our own internal pipeline is really focused on how do we use this technology to unlock new novel biology and not being able to kind of do something faster and cheaper, but actually unlock new biology, again, kind of going back to the GPCRs, the ion channels. being able to develop best-in-class assets. And so really kind of focused on solving these hard, challenging problems and unlocking new biology. That's really where the -- we have our focus, both with partners and our own internal pipeline.

And actually, I'd add to that, too. It gives us the ability to really think about it from a portfolio standpoint, so we can diversify with partners and our own internal programs across indication, including across target validation.

Sure. And maybe walk us through the cadence of interactions in a typical partnership between yourself and your partner. How does the collaboration work maybe from start to finish on a specific project? And in addition to what you're providing to them, is there a knowledge exchange? And is there -- or even like a data exchange? And are there capabilities you think you've had the chance to develop through your partnership that you wouldn't have been able to do otherwise?

Look, I mean, in any partnership, we're collaborating very closely with a pharma or a biotech company that has deep expertise in the indication and typically in the target. So there's a lot of learning back and forth between what we provide and what they can provide. So there's a very good synergy there.

Yes. And I'd really say that pharma is really set up for late-stage clinical development as well as the commercialization and manufacturing. And so our domain expertise is getting assets into IND. And with our own pipeline, we are taking it to Phase I, Phase II clinical proof of concept. We don't plan on taking it past that because again, that's not our domain expertise. And we also think that that's where you can have big value inflections from a monetary standpoint as well. And partnering at those certain phases makes a lot of sense instead of building out those capabilities internally. And when you build those out, you kind of start to focus in on a certain therapeutic area. Again, we want to be able to stay broad, work with a lot of different partners. And I think partnering makes a whole lot of sense to stay broad. So you don't have to build out domain expertise across all the different therapeutic areas, leverage what our partners have and we can bring in the domain expertise on getting assets to IND that have the potential to be first-in-class, best-in-class.

You mentioned data, too. I think one of the exciting things here is we sign a partnership with a given partner and there's research funding. And we use a fair amount of that to develop the data to train the models. So all that data we generate from every single partnership is reusable in building a platform. So it's the flywheel effect.

I think that's a really good point. We've made it very clear to our partners that the data that we generate in our own wet lab, even though it's being used for the partner program, we can use it with other partner programs because they're not giving us the data, we're actually creating the data in the partnership. So we own that data, and we are able to take that data and apply it to other partner programs. Now, if the partner gives us data, we don't use that data in other partner programs. That's kind of a -- we wall that off. It's a particular model that's only used in that partnership. But again, all data that's generated in partnerships or internally is used to continue to increase the overall accuracy of the model.

Got it. And that was actually going to be my next question for you, which you, I think, just [ have might ] answered, but is it challenging -- given it's a partnership-based model, is it challenging to be able to socialize progress for future business development? Because I'm assuming a lot of the progress you make with partners, they're on programs that are of high priority to those partners so they don't want to externally disclose the work you've done with them, especially having gotten them to a certain point, so then how do you make the case to these other partners that we're highly value-add and then we can get you from A to B, if you can't explain to them what B was?

Yes. No, absolutely. And I think that segues into what we're going to be announcing at our R&D Day coming up on October 4, is actually our own internal pipeline. So we're going to be talking about the therapeutic areas and the pipeline we've been building internally. And so we're kind of shifting to this hybrid business model where it's both a platform company, but also developing our own internal assets. And we really did this for a couple of reasons. One, we really believe in the platform. We realize that there's a lot of value that you can generate by developing your own asset. But then it's also -- to kind of your point, it's a validation. And we can get to clinical proof of concept faster ourselves than partnering it out. One of the frustrations we had was just how long partnerships take to progress assets. And if you can do it yourself, you can get to a clinical proof of concept much faster. And I think that, that's really a validation point for investors as well as partners. It's another kind of flywheel moment where you can drive more folks onto the platform. I think investors see that as kind of a critical piece, not just for Absci, but I think for all AI drug discovery companies. And we've taken a very strategic approach on how we selected the targets that we're building out internally. It's really how quickly can we get potential best-in-class, first-in-class, what is the market size and then ensuring ultimately, we can get there as quickly as possible. And so again, I think we have some exciting updates coming on October 4 with our own internal pipeline.

Got it. So you may not want to talk about this in a great level of detail now. But at a high level, when you think about your internal pipeline, which areas of work are exciting? Which -- are there specific therapeutic areas that you feel like you're just more inclined towards? Are there specific kinds of, I guess, constructs of biologic antibody that you think are just well suited to your platform? How do you narrow down the world of what you could work on to actually what you will prioritize for the first couple of programs?

Yes, absolutely. So I will say that a couple of the programs are in immunology. That's a big focus. And then additionally, looking at [indiscernible] some immuno-oncology as well based off of the Totient platform using the reverse immunology we've looked at. We've discovered some new exciting targets and pathways in immuno-oncology. And I think that, that really shows some really nice validation of our Totient platform for target discovery. And so those are -- and I think one of the things that's interesting is the immuno-oncology is still based on an immunology approach. And so there is a definite theme to our pipeline based on the technologies that we do have.

Got it. Got it. And is there any limitation on your freedom to operate? Are there specific therapeutic areas or specific areas of work that you're walled off from just because of the work you do on partnerships?

Not really. I mean, when we sign a partnership, we certainly -- there is some exclusivity around target, but we don't wall ourselves out of any indication.

Okay. So it's always -- it's typically target-specific work. So it's never for a whole category of indications when you work with a partner?

I never say never, right? We'll see what the deal terms are, but we'd be very careful about not boxing ourselves in.

Sure. Okay. Understood. And just taking a step back, one broader investor debate, as the focus has grown on AI-enabled biotech over the last several months, is how do these platforms stay proprietary, how do they stay guarded? So my first question for you on that topic is, if someone wants to try to recreate what you've built for your platform to date, what would they need to be able to do that? And how long do you think it would take them to be able to catch up to where you are now?

Yes, absolutely. It really comes down to the data, but it's really the integration between the AI and then the wet lab and how quickly you can iterate. And when I talk about an iteration, it's about how quickly can you go into the wet lab, generate data for training and then being able to then use the model and then validate it in the wet lab. And so basically, being able to use that same kind of wet lab technology to validate how accurate your models are. So you're using it both for training as well as validation. And we can do that cycle in about a 6-week time period. So we're able to rapidly iterate on our model designs and architectures. And that's what's allowed us to recruit some of the best AI talent in the world because I think one of the struggles that AI scientists have coming into biotech and coming into life sciences is just how slow it takes to generate the data and to validate the models. And so you could have a cycle time of a year or 2 years just to test 1 iteration of your model. And that's just not exciting enough for these AI scientists who can literally go anywhere and make a boat load of money. And so being able to convince them that, no, look, you can almost like work at Absci like you would work at a tech company because you can rapidly iterate with the cycle times because of that really tight AI wet lab integration, that's what's allowed us to recruit the best AI talent. And they see the upside. It's a -- they see it as binary like, if we're successful on this, this will be huge for the industry. And so we've been building that integration out for over the last 2.5 years and -- 2.5, 3 years, and that's really what's led to the success that we've had so far.

Got it. And are there novel capabilities you're developing as you just fundamentally get more data to be able to train your models on? And looking 5 years forward from now, when you presumably amass a much larger data set, I guess what kind of capability do you think that unlocks? And what kinds of -- I guess I'm asking a little bit about the unknown here, but where do you think this could take your models as you develop more data to train those models on?

Yes, absolutely. I think you go from one problem to the next. I mean, right now, coming out of the model, we have roughly a 10% hit rate to an epitope of interest, but that hit rate has a wide range of affinities coming out of the antibodies that are generated. And so, as we train our models, essentially, we're kind of in the stage of going from -- I think a good analogy is, it's going from GPT-3 to GPT-4 and really using Open AI, use reinforcement learning to increase the overall accuracy. We're not necessarily using reinforcement learning but something very similar. And as you start to train with more data, you increase accuracy, which accuracy also then means more capabilities that you have. So instead of having a 10% hit rate to a particular epitope, you're up to 50%, 60%, and you're able to actually hone-in on the exact affinity that you want. And then you can start shifting into developability, manufacturability and really having a model that can predict the attributes you want for all the important aspects in antibody development. And right now, the models are really focused on 2 things. It's the epitope specificity and then additionally, the affinity. But ultimately, we want to progress into developability, manufacturability. And so that's why we ultimately see things progressing over time. And again, you have to have wet lab technology that generates the data. So we're always going to be kind of focused on the next technology to then scale the data to solve that next problem.

Got it. And I think you touched on kind of the longer-term vision. I mean 3 to 5 years out, from like a business model standpoint, where do you ideally hope to be in terms of how much of the firm's economics are coming from partnerships versus proprietary programs. And on the proprietary side, do you imagine that being solely proprietary to the point where you envision yourself kind of building out a full drug development portion of your business? Or do you think you're going to be leveraging partnerships to progress those beyond a certain kind of value creation point?

Yes. I mean I think it's a really exciting future. And I think we'll look at all the options. But right now, I think, we think more about developing our own internal assets potentially early into the clinic, but really looking for strong partners either at IND or with Phase I data.

Got it. Got it.

Look, I mean, I think I talked about 3 to 5 years out, I mean, really, what we want to become is the next Regeneron. What Regeneron has built is really incredible. I mean they were the ones that had the first humanized mouse. And if you look at the drugs that have been approved both in partnerships as well as their own pipeline, it's been quite a bit of the overall biologics in the space. And I think that this is the next iteration of Regeneron, and I think we have big ambitions here. And I think that is both our internal pipeline, but continuing to partner.

Got it. And then, I guess, going from long-term vision to focus on very much on the now, are there specific use cases for your offerings that you think would provide and present like a low-hanging fruit to additional kind of potential partners where biopharma is just not there yet and they're just not seeing the value yet? And if so, how do you kind of drive that awareness? And I guess, what role can a firm like Absci play in flushing out those use cases to potential partners?

Yes, absolutely. I mean, I think Sean touched on it before. It's doing a lot of work with our internal programs on first-in-class, best-in-class. There's sort of 2 different strategies there. And we have a discovery platform we can leverage on the target side as well. So I think it's generating more data and starting to share more about our internal pipeline, which is coming in the near future, and we're excited about that. And I think fundamentally, it's really demonstrating the speed and efficiency of the platform and also the new target biology we can unlock.

Yes. And I will say one thing that where we're excited about with doing this kind of hybrid approach with our internal pipeline as well as the platform is it creates more opportunities for catalysts, for investors. So you have partnership announcement types of catalysts with large pharma and others, but then also having data readouts and we do expect to have exciting updates again at our R&D Day at -- on October 4, talking about our own internal pipeline.

Great. Great. And before we close out, could we quickly touch on just your capital position, cash, your expected runway? Kind of what's contemplated in that runway? What level of platform pipeline development is kind of all baked in there?

Sure, I can kick it off. Yes, I mean, look, as we released in Q2, we had about approximately $125 million in cash, cash equivalents, short-term investments. And we recently reiterated our guidance that we've got cash runway into late 2025 based on the current balance sheet. And the other thing I'd like to highlight, and I'm really excited about this, is we're really starting to see lots of efficiency from the AI platform. So this year, we're estimating we'll spend on a gross basis, approximately $80 million in cash as compared to last year, which is $105 million. But keep in mind, we're doing that with more active programs and also by driving forward our internal program. So we really are getting a lot of efficiency. And I think we look forward to the future, sort of leveraging the AI platform to continue to get efficiency gains.

Got it. And a final question for you then. As people think about the next 6 to 12 months, are there partnership-related milestones or proprietary pipeline [indiscernible] I know you mentioned an event coming up. But outside of that, anything people could just keep on their radars?

Yes, I would definitely say, again, we're going to -- I think we've said that we've had announced 2 active programs. We anticipate or we stuck to our guidance of 10 through the rest of the year. And so I definitely think partnership catalysts coming up in the next 3 to 6 months. You'll definitely see that. And then I think a really nice potential data readouts in that 3- to 6-month time period as well.

Great. Great. Okay. I think with that, we're out of time. So we'll go ahead and close out. Thank you both for your time. I appreciate it. Thanks for the discussion.

Yes, thank you, Vikram.