Absci Corporation (ABSI) Earnings Call Transcript & Summary

Great. Thanks, everybody, for joining us. Welcome to UBS Healthcare Conference 2024. My name is Ash Verma, I cover SMID-cap biotech and spec pharma. I'm really excited to have the Absci team here. So I would love to get into the discussion. Just if you can just give us a quick introduction about yourself and then sort of give us an outline of where the story has been, what the evolution has been in the last couple of years and then we can sort of take it there. I'll pass it on to you, Zach.

Sure. Sounds good. Zach Jonasson, I'm the CFO and the Chief Business Officer. I joined the company just under 1.5 years ago, but had a very long tenure with the company prior to that. So I initially was the lead investor. I was a managing partner at a venture firm. So I was a lead investor in the Series A and B rounds when the company was all of 10 people. So it's been a remarkable journey as a Board member and investor. And about 1.5 years ago, the CFO was retiring, and I'm very excited about the progress of the company, and I really believe there's an excellent management team here. Andreas Busch had just left the Board to take over to lead R&D at the company, and the management team said, "Hey, could you come in and join and work with us on a day-to-day?" And I said, "Absolutely." It took me a while to unwind what I was doing on the venture side. But I guess the short story is, I voted with my wallet and then I voted with my feet. And I think I'm in a good position to give you a little bit of the history of the company as well. But why don't I pass it to Alex to introduce himself first.

Yes. Thanks, Zach, and thank you, Ash, for having us here. We appreciate the invitation. I'm Alex Khan, Vice President of Finance and Investor Relations for Absci. I've been with the company about 2 years now at this point, but I've known the team a bit longer than that since the summer of 2021 around the time of the IPO.

In terms of history -- do you want a little bit...

Yes, yes, that will be great.

It's a very interesting history because the company actually didn't start as an AI company. It started before AI really took off. It originally was a synthetic biology company. And so Absci is the first company, and I think the only company, to my knowledge, to ever -- and we still do this on a day-to-day basis, produce full-length antibodies in E.coli. And what they took and did with that technology is built a really high-throughput assay system where you can test up to 3 million unique antibody sequences. We do this in a Fab format against a given antigen of interest. And so -- and we can do that in a 6-week active learning loop cycle. And so as the company developed these assays, they realized that this was the kind of data that you could use as a foundation for AI. And having been an investor in this space for a long time, I would say one of the most important elements in basically developing AI to a place where it's productive as having the appropriate data and that means the quantity of data, the quality of data and usable data. And so for Absci, they met all 3 of those criteria. So we create data at scale on the 6-week active learning loops, and we've been doing it for 4 years. And all of that data goes in, it's functional data. It's married with structural data that we can get within the company or outside of the company. But all of that data goes into training our AI models. And then we use that same assay system to do validation of the designs that the AI model makes. And in that way, we are constantly iterating to improve our models. And that's what's led to what I think is our leadership position in the space for AI design of antibodies.

Great. Excellent. So I guess -- so sort of synthetic biology routes and now AI was not a thing when Absci started, I know that I think now more sort of a resurgence of this, like just where do you think the platform going over the next few years? I think this is sort of like an interesting approach, and a lot of problems can be solved to your point around just like finding like the right optimal molecules, like where do you think is the most value creation that can happen?

Yes, absolutely. I mean if you look out 10 years, I think AI is going to be transformational in a number of places. At Absci, we're laser focused on using our platform to design novel and differentiated antibody-based therapeutics. And it all starts with our model setup, we start by selecting the epitope. So we provide epitope selection to our partners, and we use that for our internal programs. And then the model will design against that epitope, but not just against the epitope. It will computationally and exhaustively sample the interface space for a given epitope. And so what we're finding in-house is by doing that, not only can we select an epitope and if you don't know the epitope to start with but you know the antigen, we can landscape the antigen. We can design against multiple epitopes, test the biology. So if you're looking for agonism, for example, we can do that very rapidly in our lab. But once we identify an epitope, then we're really looking to optimize the interface, so the paratope, and how that interaction is going to drive the biology we're looking for. And so right now, we're finding -- and we've shared some of this data, we'll share some more data at our R&D Day next month on December 12. But by fine-tuning and optimizing that interface with the epitope, we can drive potency. And in some cases, we can drive unique MOA. I think that's really exciting. The second thing we're doing with our models, which is more on the lead optimization side. So the de novo model is what we used to do, the epitope selection, design against the epitope. Once we generate leads from that model, we take it into an optimization model, and it's really a set of models. And so there, we're optimizing for developability, affinity, a number of those parameters. But what we've been recently doing with that sort of model construct and those regenerative models are designing in pharmacological features. So we've had success designing in pH-dependent binding, for example. We've done our own SC mutations using AI that confer longer half-life than what you would see with a YTE, for example. So we're really pushing the platform in different ways but always laser-focused on how do we deliver differentiation.

So just a couple of questions based on what you mentioned, like I wanted to follow up. One is that, yes, I know that there are a few other companies that are -- that have like these AI models that are trying to come up with some innovative molecules using different models. Like what makes you believe that your model and your approach is better than the others in the industry that are doing the same thing? And effectively, like is there -- I don't know if there's a like-for-like comparison that can be done on truly identifying like what's the best approach here in terms of R&D find. Like either at like any state like discovery lead optimization or sort of some of the other steps?

Yes. And I'll say it's a noisy space. I was an investor in the space for the 5 years prior to joining Absci. And part of the reason it's noisy is there aren't a lot of ground truth metrics. And so at Absci, we focus on validating everything in the lab. There's a lot of -- in the industry, there's a lot of in silico metrics. We don't find that those often validate when you take them in the lab. So we test every model that we think is interesting that's published on. And we find a lot of times they don't validate the way they do in silico. So when I think about what really differentiates a platform, I think about a couple of things. One, the data platform that supports it. So if you can't generate enough data that's high-enough quality, there really is no way you can get model performance. And so there are some companies that do have a data strategy. Recursion is a good example. Absci is a good example. Recursion is focused on small molecule. We're focused on biologics. But there are a lot of companies that have no real data capacity. And I tend to, I know this is a bit pejorative, refer to them as AI sprinkled on top. We're AI integrated in everything we do. So that data component is integrated with -- our wet lab is integrated with our AI team. So we're generating that data 24/7. And then we're also doing that validation step. So that active learning loop is really, really important. And then the second thing I would say is it's really about the results, right? We've shown that we can take our model to produce our own assets. And so we did that with our lead program, ABS-101, and we've been able to show how we use the platform to engineer in differentiated features. We'll announce some more data around our second program, ABS-201, at our R&D Day, and we'll do the same thing, we'll map through how we use the platform to design the features. And then I would say it's also results that we're having with partners. We had released an 8-K about a month ago on a partnership that we have with AstraZeneca. And here's an example of how we used our de novo model. So we worked against a difficult target that had no known binder, and we were able to deliver high-quality antibody leads in 6 months. And I think that's -- to me, that's a clear sign of leadership in the space. We haven't seen other companies accomplish something like that, nor have we seen other companies with third-party validation.

Excellent. That's great. Yes, I want to talk about like the pipeline that you have and a bunch of partnerships that you have announced. Just before we go there, I guess the other question on what you mentioned. So I think this idea of like trying to identify the data, right, for all the molecules that is publicly available and identifying like the right characteristics for -- and then you try to apply AI to it. I think the one challenge that I've heard from other companies would love to get to your perspective is that there's not a lot of like negative data available. Like for programs that have failed, right, like there is very limited information available. For the successful products like companies like obviously disclose all the information. So is there -- does that become a limitation in a certain way that you don't have like access to what might have caused something to fail and you're only looking at like a positive bias on what has worked effectively?

I think that's a really good question, and it's something we think about a lot. I think my first response would be I think that challenge is a little bit bigger for the small molecule players because they're -- one of the challenges they have, they don't have the computational complexity to deal with designing the molecule. I think the bigger complexity they're dealing with is the off-binding potential. So all the -- is it selective enough? Are you going to have toxicity that's generated by binding other targets that are unwanted? We're, in some ways, leveraging mother nature. We're using the antibody design, which is inherently highly specific. It's not completely specific, but it's highly specific relative to a small molecule. So we have built up internally our own model to just -- to address polyspecificity. So again, we're thinking of how do we make sure we're designing things that are going to be developable, safe, not going to have a lot of upbinding or upbinding that's material, not going to be immunogenic. So we're developing models to address all of those aspects.

Yes. Got it. Okay. That's perfect. So yes, let's just talk about the partnerships that you have had like over the years, several partnerships. I guess like what has driven that? Is it like more sort of like incoming inquiries to you? Or do you -- have you been active like trying to reach out these big pharma? And just related to that, sort of like what is the commercial model effectively from these partnerships? When do you start to see some like revenue generation happening?

Yes. I would think about what we're doing on a commercial front is we're building a portfolio. And there's sort of 3 elements of the portfolio. The first element are the partnerships we do with companies like AstraZeneca and Merck and Almirall. Those are structured as more traditional drug discovery partnerships. They bring a target and they bring -- the synergy they bring is deep knowledge of that target, that indication, the ability to commercialize and develop the molecule. And those are structured very similar to what you've seen in the industry. So there are upfront payments, R&D funding and then there are milestones, development milestones as you go through the clinic and then commercial milestones and royalties. So that's sort of a traditional kind of anchoring structure, and that's one element of the portfolio. The second element is something that we've started doing relatively recently, which is to do co-development partnerships. So we announced a partnership this year in August with Memorial Sloan Kettering. And there, we're codeveloping 6 assets with them. So again, there's a great synergy. They bring great target biology in oncology. We build the asset and then we work together to develop it through Phase I proof of concept. And the other nice thing with MSK, is they have access to patients and can help us execute on a Phase I trial. And then we would turn around and license -- out-license that asset to pharma. So we like that model because, again, we can leverage synergies, but it's also half the cost. And then the third element of the portfolio are the assets that we're developing internally that are wholly owned. And so that would be, for example, like ABS-101, our lead program, the TL1A antibody. We've announced 3 of those programs, and we'll announce another one here at our R&D Day. And the objective of those is to bring them to a proof point where we would then out-license or partner with pharma. So these 3 elements have -- all have different kind of risk-return profile. But if you think about it holistically as a portfolio, what we're trying to build is something that's balanced, diversified that can give us the best ROI.

Okay. So for these partnerships, right, that you have with big pharma or some of these like 3 players that you mentioned, I'm assuming they have their own internal capabilities like when they are reaching out to you or striking a partnership with them, the idea that -- is it like exclusive to certain know-how that you might have that they wouldn't? Or is it more of like sort of complementing each other's skill set to try to best solve the problem?

Yes. We view the world as we need to find collaborators where 1 plus 1 is greater than 2. So we're looking for partners that bring, in particular, a lot of knowledge around the target. And I think from their side, and it's a little different with every partner. But in general, one of the things that I think our discovery partners are most excited about is that de novo model I was talking about. And that's where we can work against difficult targets even if there's no known binder. So we don't even know what the parenteral framework is. We can find success that way. And we can find success where other traditional approaches can't find success. And then beyond that, being able to dial in the parameters so that you have a truly differentiated asset. So it's gotten very high potency or against -- you're targeting the epitope that's the right epitope for the biology and that you're delivering some of these pharmacology features that I mentioned.

So some of the things that you mentioned, right, like data generation or screening capabilities and like rapid cycle times, like what is sort of the advantage of that from a competitive standpoint, like what can you solve for that others cannot?

Yes. I mean we've been refining that sort of -- we call it an active learning loop, right? That's creating data, you're training models, you're testing the output of the models, improving the model that just you're iterating that loop. We've been refining that for 4 years. And I can tell you that's the reason why we've had success with the de novo model, for example. You need to have that large quantity of data, but you also need to be able to fine-tune the model and test external models, too, to see if there's something interesting that's out in the literature that you should incorporate into what you're doing. So if you don't have the laboratory that can do that at scale and that can do that quickly, you're at a very big disadvantage in this field. I think it's a core requirement.

So when you like started to develop your own molecules, I think that's an interesting pivot. Like was that always a plan from a strategy perspective? Or did you sort of realize along the way that you're already testing some of these molecules, might as well start to develop them in-house? And it's sort of like a two-pronged approach now, right, as I understand it, that you're sort of like doing R&D to a certain extent. So what drove that? And how does the platform evolve from here?

Yes. I hope investors appreciate this. I was on the Board at the time when we were having these discussions. Are we going to build our own pipeline? And one thing that Absci did extremely well, but you really don't see very often, I think, in the industry. Typically, companies rush towards the clinic as fast as they can. And I think oftentimes, suboptimal decisions can be made around target, asset, all of that. We took our time to recruit in a world-class discovery team that we could have worked with the AI platform as a tool. And that discovery team is led by Andreas Busch. He was the former Head of R&D Global for Shire and for Bayer before that. He has over 15 approved drugs under his belt, and our discovery team in general has over 20. And when he joined the company, he brought in people that had worked for him in Shire and at Bayer and sort of the cream of the crop. And so we put that infrastructure, that human infrastructure in place before we embarked on creating our own assets. And I think that's a really important point. And to your question, why did we do it as a strategic element, because the ROI is so significant. If we can choose targets well and we have the human capital to do that, and we can use AI tool to design the assets to be differentiated, the return on investment there is substantial. And Alex is very good at reminding us of this. But to develop our lead program to DC was roughly $5 million spend on our part. So we can be very efficient and we can leverage the platform to design and differentiation that we think is meaningful for patients.

So the idea of this sort of like developing your own proprietary pipeline, right, like it would always come from your own AI engine and just like sort of like optimal molecule that you would have identified that others might not be pursuing, let's say, in the right way and you're trying to extract the best out of it. Is that a fair synopsis?

Yes, in part. So if you look at our internal portfolio today, internal pipeline portfolio, 2/3 of it is fast follower. So that it's more the model you were describing, where we've identified a target that is validation, but the innovator molecules have weaknesses. And we think we can address those and maybe even go beyond designing features that would be truly enabling for that market and for those patients. There's -- 1/3 of the portfolio is focused on first-in-class, and we have a target discovery platform that we acquired into the company prior to the IPO, that allows us to assess biopsy samples from patients, particularly oncology patients where we then look at what the tertiary lymphoid structure repertoire of antibodies is, find the antibody that the TLS is evolving to and find its cognate target. And so we have a target discovery program that -- a platform that allows us to pull out novel targets with their human cognate antibody, which we then would AI optimize. So we have an ability to do kind of novel targets as well. But for diversification, we want to have a mix or a balance that means a little bit towards fast follower where we layer in maybe 1/3 of the pipeline is first-in-class.

Okay. Perfect. So yes, I mean I wanted to talk about TL1A, like I was explaining to you is sort of like a topic near and dear to my heart. There's a lot going on. This is quite a big of a few different big catalysts coming up in this space. Just wanted to understand like how you're approaching this program. What do you think is the differentiation that you have? There are a few different competitors out there. Like ultimately, what makes you -- what gives you conviction that you have like the right best molecule?

Yes, absolutely. Maybe just start with how we design the molecule. So we really wanted to set out to design a molecule that would be best-in-class and differentiated. And we started by selecting the epitope. So we work from the same where Merck binds, so the same epitope Merck bound. And we chose that epitope because it has a much cleaner ADA profile. If you look at what the ADA profile for Roivant is, for example, we believe that is epitope driven. We've done T-cell activation studies on all of the first-generation TL1As. Everything looks clean, and there's a good literature in the TNF space showing that oftentimes, you can find complex formation that's driven by a specific epitope interaction. So we believe that's what's going on. That's our hypothesis. So we designed specifically around that Merck epitope. And then we use the platform to greatly enhance potency and we -- and that's affinity and that's translated into potency, which we've shared that data. So we believe we're going to come to market with a much higher potent molecule with a clean or a much cleaner ADA profile and then we've added in half-life extension as well. So we've added in an Fc mutation that should give us 2 to 3x at least the half-life of what you're seeing with the first-gen molecule. Beyond that, we use the developability model. So in lead optimization, we optimize this molecule or our molecule to be highly developable. So we believe that it's a very clean profile from many standpoints. And just a proof point there, we've been able to formulate this so it's highly stable at up to 200 mgs per ml. And so we're really well set up for a subcu dose all the way through the clinic. Yes. So I think we're very well positioned to differentiate against the innovators.

And so just as a reminder, so where are you in the sort of like the development process? Like when do you start to do some clinical trials in it?

Yes. We're completing IND enabling right now. We released data on our NHP studies, demonstrating the half-life extension as well as really nice tissue distribution as compared to some of the lead clinical molecules. Our plan, and we're on track to enter the clinic here in the first half for a Phase I work and we'll have an interim readout in the second half of next year.

Healthy volunteer study?

Yes.

And like do you ultimately -- I know several companies have talked about this that, like how much of a utility is in like ulcerative colitis versus Crohn's. There are a few different other indications where you might start to see value. Like do you have your mind set on like where you might ultimately pursue it?

Yes, it's a great question. We think TL1A is a very exciting target. This is one of the reasons why we've decided to work on it a couple of years ago. We're actively exploring some fibrotic indications in-house. We haven't made any announcements about that, but that's something that's under active investigation.

Got it. Okay. So as you think about like the ADA differentiation point that you were talking about, like in clinical trial setting, that's the type of thing that can take a while to prove sort of more the antibodies are forming or not, sort of like more -- like even just like beyond like the short-term induction studies that you start to see the value of it. Is that a fair assessment? Like would get to know that maybe in like a Phase II long-term follow-up or some OLE studies?

In general, I think that's true, but I would say there were signals in this landscape earlier on. And certainly, you saw a very high ADA rate in the Phase II work from Roivant. And not all of that was neutralizing, but I think the expectation is as you move into Phase III in the longer dosing period, that's a 6-month dosing period, you're going to see a lot more of those ADAs transition to neutralizing. And that's not uncommon in the TNF space as well. So our expectation is having a cleaner ADA profile, we'll start to see some early signals early in development. We have to prove that, and we're focused on doing that.

Right. Yes. I think there is definitely like a lot of focus on this and big market opportunity just with IBD being just like $30 billion, and you're talking about fibrotic indications as well, which can be another set of its own thing. You do have like a few competitors in this space, like so yes, with the Teva, Sanofi and like this Roivant molecule, and then Merck, Inspire is another one, right? And I guess different companies have talked about like different level of like in vitro strength and like rationale for why they think that the molecule is differentiated. Like do you think ultimately like these type of things would matter? Like if you -- if I've heard some arguments around like not going -- keeping the decoy receptor free and like this DR3 versus DcR3 selectivity? Or if you think about like pre-TL1A being available for a certain period of time, just what would like these type of attributes mean for the profile of the drug? And if you believe that some of these hypotheses are correct, what people are pursuing?

Yes. I mean I don't -- we were talking about this before the panel. We don't believe there's a lot of merit in the story around the decoy receptor. We do think there could be some merit around the monomer, being able to target the monomer. And we do it the monomer, we chose an epitope that allows us to hit the monomer, and we hit it with very high affinity. There's evidence suggesting that the monomer is active, certainly in the cell-based assay setups and systems. The question will be what the prevalence is in the patient populations. Nobody has done that work or shown that work yet. But our view is there could be some additional efficacy, particularly for certain patient subgroups that have prevalence of monomer, and we'd be well positioned to take advantage of that. That's the one area where we think there is -- we have to prove it in the clinic, but that's the one area where we think there could be some additional differentiation.

So try to identify some sort of like hyperresponders sort of like the Prometheus approach, but then like more of a monomer selective patient population, is that...

I don't know that we'll get to an assay where we can assess the prevalence of the monomer, but just pointing out there could be -- the monomer is active biologically. It's not inert. And so if it has a high prevalence in certain patients or certain patient subgroups, a molecule that it can also target the monomer, I think, would have a higher efficacy. But again, we need to prove that out in the clinic, and it's not clear what the prevalence is in different patient groups at this point.

Okay. All right. That's great. So yes, I wanted to just like ask about the other proprietary molecules that you are working on. Just anything that you can share?

Yes, why don't I bring Alex in? You want to comment on that, Alex?

Yes. I think next month at our R&D Day will be an opportunity for us to showcase our other internal programs a bit more. In particular, we do expect to unveil the target and some preclinical data for ABS-201, which is our potential best-in-class dermatology program. To this point, we haven't said much about it publicly for strategic reasons. Other than that, we think it's a very underappreciated target, and we see vast potential for that one. And then for ABS-301, that's our potential first-in-class novel IO target that we discovered through our reverse immunology platform. We do expect to have some data to share in the first part of next year. But at our R&D Day next month, we do also expect to unveil 1 new program that we're adding to our pipeline on top of the other 3 that we have in there already.

I see. Got it. So I think you're going into a few different spaces there, like the IO you mentioned, right. So this is sort of like -- is there any sort of thinking behind like trying to focus on like different therapy areas, really, you're trying to go wherever the opportunities? Or ultimately, like do you think that it would boil down to like trying to focus on one therapeutic area?

Yes. I mean we can't be defocused. So we're focused on an area of biology that's cytokine biology. And that really takes us into I&I and some areas of oncology, and that's really where we're focused. And you'll see part of doing the MSK partnership, the partnership with Memorial Sloan Kettering was strategic too because we're moving in some ways into oncology, given our reverse immunology platform tends to provide oncology target. It also can provide I&I targets as well. So we're building more and more of those relationships to help supplement when we build internally. And then we access the diversity, the platform is agnostic to the indication. So we get diversity and indications through our drug discovery partnerships where we have partnership -- partners that work in different indication spaces. But for our own platform, I think it's important for us to build out the infrastructure to be able to take those molecules, understand the disease biology, do the translational work and move those into early clinical.

Got it. Okay. That's great. So yes, you recently just like announced the third quarter earnings as well. Just like if you can comment on the financial position of the company. Like how do you see your cash use in the coming year and beyond?

Yes. I can -- we just announced we're sitting today with -- at the end of the quarter with roughly $127 million in cash, cash equivalent, short-term investments on the balance sheet. We are retaining guidance that, that should take us into the first part of 2027. And that includes our work that we're planning to do advancing ABS-101. Again, going back to earlier conversation, we expect to have an interim readout on that Phase I work in the second half of next year.

Great. Does this include like the milestones that you might be getting from these partnerships?

Our forecast has some modest assumptions around partnerships that are under contract today as well as some modest assumptions around adding additional ones on the cadence we've done previously, but it doesn't include upside scenarios of doing a large thematic deal or out-licensing one of the internal programs, that would be all upside to the forecast.

Got it. Okay. That's great. All right. So just to close, if you can just like highlight a few different catalysts that are coming up for the story in the next 6 to 12 months and beyond that, that would be great.

Sure. I will bring Alex in.

I mean that's another thing we're really excited about is as we sort of complete our transformation next year into a clinical-stage company with our Phase I -- with our ABS-101 entering the Phase I in the first part of next year and having the interim Phase I readout in the second half of next year. But before that, we'll have our R&D Day next month. And like I said, we do expect to unveil the target and some market data and indications for ABS-201, the potential best-in-class derm program and unveil a new program added to our pipeline on top of the 3 we have in there already. And then looking into the first half of next year, in addition to what I mentioned on ABS-101, we do expect to share some data on ABS-301, the potential first-in-class IO target, our IO program for a novel target we discovered. And of course, always looking to do additional programs added to the pipeline that our team is working on. So just stay tuned for those. Beyond that, always speaking with potential partners around deals we could do either on the drug creation side, whether it be at the discovery phase, for example, or even around our internal programs that we have in the pipeline right now. And I think at the end of the day, it really comes back to the platform that we have. And I think what we've been demonstrating with these partnered programs and our internal programs, is the vast potential that we see for the platform to create these differentiated -- potentially differentiated antibody programs using engineering principles in a very rapid and cost-efficient way. Like as Zach mentioned with our ABS-101 program was about $5 million from the start of the program to get to the drug candidate phase in a matter of just 14 months, which we see as very rapid. With the current iteration of the model now, we think we can get that down to about 12 months or so. And I think with the programs that we have in our pipeline and the partner programs we have, like with AstraZeneca, for example, we're seeing some good validation of the platform. And I think some other external parties are starting to pay a little more attention to the platform and see the potential value we can offer for those. So I think there's a lot of potential, too, to be working with other outside parties, biotech or pharma, on certain programs going forward, too.

Yes. Great. All right. Thank you so much for this very insightful session. Good luck with all these different milestones that you have coming up in the business. And yes, I'm looking forward to staying in touch. Thanks for coming to our conference.

Thank you.

Thanks, Ash.