Schrödinger, Inc. (SDGR) Earnings Call Transcript & Summary

Okay. Let's go ahead and get started. Welcome, everyone. My name is Vikram Purohit. I'm one of the biotech analysts with the Morgan Stanley research team. Happy to have with me on the stage here, Ramy Farid, Karen Akinsanya, and Geoff Porges from Schrödinger. Thank you all for joining us. Really appreciate it. Before we get started, I need to read a brief disclosure statement. For important disclosures, please see the Morgan Stanley research disclosure website at www.morganstanley.com/researchdisclosures. With that, let's go ahead and get started.

So maybe we can just start at a high level talking about some of the broader considerations for Schrödinger's business and then we can go into specifics from there on both the software business and then also on the pipeline. I think broadly speaking, it seems like there's been a few macro themes that seem to be kind of impacting the narrative on Schrödinger and other life science software companies and companies that serve the biopharma industry. There's been discussions on the broad potential of AI and ML. There's been discussions on changing priorities across pipelines in the biopharma companies. There's been discussions around biotech funding. I would love to kind of get your sense on what are you hearing across all these topics and others that you may be encountering on which of these is kind of impacting your business and kind of your top line.

Yes. So first, we can acknowledge that we have been hearing that, mostly from other companies in the space. And to just be very sort of straightforward about it, while we're hearing about it, we're not actually seeing any impact on our business to date. There's still significant enthusiasm for scaling up usage of the software. We keep getting reports about the impact that the technology is having on programs. And there's generally still an understanding that there's still -- companies are still underutilizing the software. I think they see -- you can imagine, for example, when a company acquires, let's say, one of the biotech companies we co-founded and they see the impact that using the software at scale, which, of course, they were because it was through the collaboration, they see the impact in the assets that were developed, and then they realize, well, they can be doing better. So that's the kind of conversations that we're having is figuring out how they can scale up, what do they have to do, the expertise they have to bring in, the cloud resources, and the know-how sort of transfer that has to happen. So that's really what we're spending most of our time talking about. Even though we're aware that what you just said is happening, it's not really….

Yes. And the direct question there is, there's been other companies in the space talking about guidance revisions, potentially softer outlooks and you haven't in recent quarters, and what's driving that? What's fundamentally helping insulate your business for the time being from some of those dynamics?

I think it's a few things, and maybe Geoff can comment on this, too. I mean, I think one is there's now a clear recognition that the more technology they use, the more software they're using, the cheaper discovery is. So it's actually saving, right? They're reporting that they're making fewer molecules in the lab. So if you're doing that, obviously, you're moving faster and you're spending less money. So that's probably the single biggest reason why it's not really impacting us. Geoff, is there anything else you think?

Yes. Look, I think, first of all, we are a tiny fraction of our customers' R&D spend at the global companies. We calculated, back of the envelope, we are less than 1/10 of 1% of our largest customers' R&D spend per year.

Your entire software revenue base?

Our software content and our largest customer. So that's sort of a drop in the bucket for these very large companies. And given the fact that we are cost-saving compared to what they would otherwise be spending to come up with innovative molecules, it's probably why we aren't part of the discussion about reducing cost. I think that we have commented previously about a small piece of our business being in the emerging small biotech company space. And there's still obviously a lot of churn in that space. But what we're hearing from the other companies, the supply global pharma companies is that they've had some challenges. I think that, that's because companies may be cutting back on clinical trials, cutting back on the number of molecules in development. But we're, as Ramy said, not really experiencing that, I think, because we're relatively small and because we're providing cost benefit to them even in terms of their expenses.

Got it. Got it, okay. I guess another big picture question that really persists for Schrödinger in the space generally is people still think about the theoretical TAM and often get to very large numbers, sometimes to kind of staggering numbers. And then the actual question is, what are the factors that need to happen at an industry level, at a sector level, at a company-specific level, to kind of propel Schrödinger's top line towards that theoretical large TAM? So I don't know if you have any thoughts on that.

Absolutely. So first of all, you're absolutely right and we are so far away from the TAM. That's very clear. We're in the very, very early days of sort of broad adoption of computation. So what does that mean? When you're in the early days, it means there's a culture shift that has to happen. There are still a lot of companies that are kind of used to doing things in a certain way, making a lot of molecules and just testing things. And there's sort of this -- the technology didn't work for so long, this feel that there's still the sort of skepticism that's kind of baked into the whole system. So you have to overcome that. And it happens. It happens very readily at companies that use the technology at scale for essentially one iteration of design and making molecules, and they're convinced that you have to go through that. The other thing is just simply the expertise to actually use this kind of technology. So that's something we're investing really heavily in. We have actually a group at the company called the Education Group that is essentially responsible for training the whole field, from graduate students to researchers and companies to use what is really relatively new technology. And then the last thing, and this is something that's getting a lot better very recently is, of course, to deploy this technology on the scale that we're talking about requires significant compute resource. And that means having either an internal cluster but more often an internal compute cluster, but now more often, cloud -- resources through cloud, which means a relationship with AWS, GCP, Azure. And first of all, the comfort that it can run calculations on the cloud, again, that's something that people have started to get a lot more used to. But there are still some companies that just sort of struggle even just with the IT infrastructure that's required to be able to run calculations on this huge scale. So I think we're making -- those are the areas, but the good news is we're making progress in all of those areas.

Got it, got it. I guess staying on that point then, is there a kind of new and emerging use case you've seen for Schrödinger software services up in the past year or 2 that you think is kind of an interesting example of how use can be expanded and can be used kind of as a template for future growth?

Absolutely. That's a really great question. So drug discovery is a multi-parameter optimization problem. You need a molecule that's soluble, permeable, stable but not too stable, potent and selective, right, can bind to other targets. And traditionally, drug discovery is done by sort of optimizing one of those properties and then finding out that you have a problem and then fixing that problem. And then, of course, it turns into a whack-a-mole problem because now the thing you had working before isn't working anymore, and you've heard us maybe even use that term before. So the optimal way, of course, to use the technology is to optimize all those properties at the same time, all of them. And that's a lot. If you think about selectivity, I mean, you have maybe potentially hundreds of targets that you want to avoid. You've got the target that you're trying to go after, and then you have to fine-tune all these, all the ADME properties and PK properties and so on. So as researchers start to recognize that you can compute all of these things with very high accuracy, now you can -- the scale in which you're using computation really explodes because you're now doing multi-parameter optimization. You're optimizing all of these properties simultaneously. That requires, obviously, a lot of licenses for the software and a lot of hardware, and that's where the growth is coming from. And I'll say one more thing. These physics-based methods that we've developed rely on having a protein structure as an input to the calculation. And now there are companies that are investing much more in structural biology experimental methods. cryo-EM is one of them to actually get structures. You hear a lot more about computational methods. We've been developing a number of them for a protein structure prediction. So then to sort of adopt this idea of actually prioritizing getting structures of proteins will also expand the demand because there'll be more properties you can compute and more structurally-enabled programs that rely on these physics-based methods at scale.

Is there -- sorry?

Well, no. I mean, I think just to add to that, in a way, if you think about it, we've been designing drugs a little bit with blindfolds on, right, for decades. And as Ramy said, now we have 3-dimensional structures, an incredible resolution to be able to both work with the physics-based methods but a lot more understanding of molecular structure function. And we think that this period of time over the next 10, 20, 30 years is really the sort of taking off of what was very difficult to do, which is structure-based drug design in its best form.

Right. And Ramy, you mentioned previously that if Schrödinger were to charge itself for your software service, you put out a number before. Could you remind me kind of directionally where that was?

Yes. That number keeps increasing, by the way, because we keep using more and more software. I think the number we have said before was $70 million, right? So what Vikram's referring to is exactly as you just said it. If we had to -- if we were -- if somehow Karen's group was a separate customer, we'd be spending the equivalent of -- we'd have to pay around $70 million for software. What that really means is that's roughly an order of magnitude higher throughput than our largest customer and yet another order of magnitude of the sort of 20th customer. There is a big difference between the scale with which we're using the software. And it's what we were just talking about. It's using huge numbers of processors, calculating all these properties simultaneous on huge libraries of molecules. And we're -- but it's exciting that it used to be a lot worse. We were much, much higher. So it's encouraging. It's going in the right direction. There are pharma companies that are using the software at the appropriate scale on a number of their programs. And they're more and more trying to figure out how to get there by addressing other things we were talking about before, bringing the expertise, changing the culture, getting the hardware that's necessary. It's all heading in the right direction.

But Ramy also that number is spread across a dozen or so programs. So that comes back to the TAM question, right? If you take that number and then divide it by a dozen-or-so programs, that's what we think is really required to use the technology to its fullest utility to come up with truly differentiated molecules. So take that and then apply that to the whole industry is where we think the industry can go.

That's where those crazy TAMs come, right? But it's not so crazy. I mean, yes, it's a big number, but it is -- that is the TAM.

Sure. And the reason I asked that question is because precisely what Geoff, you just mentioned. What's required for a customer, a biopharma customer to kind of reach that spend level? And it sounds like it's, at some level, just kind of broadest number of programs they are using.

Well, it's also -- it's the things that Ramy mentioned. And we have customers. We have emerging biotech companies that are, believe it or not, using our technology at that sort of scale on one or two programs. And they're saying, effectively, they're all in on using computation to count with molecules and they are at that scale already. So there are people at the forefront of the industry who've already figured this out. Now kind of the reflex of larger companies is still to try and spread the technology across a bunch of programs. And gradually, they are learning, but that's actually not the best approach that they need to do what these emerging companies are doing, sort of go all in and really go deep with the technology. So that's where the opportunity is.

Now here's the really encouraging thing. These companies now, because they're starting to see impact, they're going to conferences and various other venues, some of them that we actually host and talking about it. So I think we know this very well that the pharma industry is aware of what everybody else is doing. I think there's going -- you're starting to get this sense of a fear of missing out sort of on the companies that are falling behind. And we're -- you can see that there's a real drive to sort of not fall behind and they're getting that kind of validation and so on. They're hearing it not from us now but from customers. That's obviously very effective.

Sure, and I guess I'll ask you a bit maybe admittedly a bit of a nice question on that. As I guess the proliferation of AI/ML technology more broadly in drug development, as it becomes more pervasive, is there a theoretical risk that the barrier to entry for competition to Schrödinger solutions becomes lower, either from in-house biopharma solutions or just the general third parties?

Yes. No, I really don't think so. And the reason is that machine learning and the algorithms that people are using in machine learning and AI are open source. They're already widely available. So where the competitive advantage could come from is, and this is always the case with machine learning applications, is in the training sets. So there are two sources of training sets. There's experimental data, which, by the way, is in the public domain so everybody has access to that. The other, as we've shown, which is pretty exciting, is first principles methods, physics-based methods. That's where the competitive advantage is. It's the ability to generate massive training sets for machine learning using physics, using computation. So that's what differentiates us from the other companies. These other companies cannot catch up magically by just advancing the algorithms. That's not the limitation. The limitation is in the training set. And you are not going to solve this problem by generating a training set using experiment. That's about 100,000x slower than doing it computationally. You will never generate a big enough training set using experiment only. You have to use computation. And that's where we -- that's where our competitive advantage is.

Got it. Okay, great. Maybe I guess turning the focus a bit more on company-specific items then. So we talked a little bit about kind of your general confidence and kind of the cadence of the software business. More specifically on your 2024 outlook now, how do you feel like you're tracking?

Yes. We feel really good about the outlook for the year. We reiterated our guidance. We had a second quarter, we put up $35 million. We beat our guidance for the quarter. Our guidance for Q3 implies reasonable growth compared to Q3 last year. It's still clear that we need to have a big fourth quarter but that's typically how our business is structured. Just for those of you who aren't familiar with our business, the fourth quarter is heavily driven by our largest customers. Those tend to be global pharma customers and they have moved, having used that technology over a period of time, to an annual purchasing cycle. So get to the end of the year, they've been using the technology through the year. They're deciding about a renewal. We negotiate the renewal and then there's a contract and that triggers the revenue. So that's why the revenue was concentrated in the fourth quarter. So we have line of sight to those renewals. We know when the contract expires, we know what the current contract looks like. We've been in contact with the customer throughout the year about those renewals. So it's that confidence -- that information that informs our guidance. Now we don't have certainty about the magnitude of the step-up. So will a particular customer increase by 20%, 25%, 30%? It could be anywhere. Now obviously, we hope it's at the upper end of that range or even above. But that's what is the kind of range in our current guidance. But we have really good information about the trajectory of our revenue outlook.

Got it. And then just, I guess focusing in on that, the bookends of guidance, kind of talk us through what drives either end?

Yes. So the high end of the guidance range is a larger -- implies a larger number because obviously, the tight range in Q3 is a larger number in the fourth quarter. And that assumes that we then have customers who renew with significant step-ups so not 10% but more -- higher step-ups and that they elect to renew with multi-year deals, right? So our largest customers over time, when they renew something that's essential to their business process, they tend to want price security. And so they say we'd like to renew this with a two-year or a three-year contract. So if that occurs and some of the annual contract customers switch over to multi-year contracts, then that will drive us to the upper end of the range. Conversely, if the level of step-up on the renewal is at the lower end of expectations and/or they stay at annual contracts, then we'll be at the lower end of that guidance range. But those are kind of the key variables really.

Got it. Understood. And then similarly for the drug discovery guidance for the year, just kind of your general thoughts on where you're tracking there. I know you've mentioned that typically, it's a lower visibility business for you than software, but any color you have would be helpful.

Yes. So look, the drug discovery revenue is definitely bumpy quarter-to-quarter. We had a strong Q2 where we were able to recognize some milestones that we've expected later in the year. We've made progress on those programs and have to recognize the revenue in Q2. That was good. Now in terms of the balance of the year, we do have some additional milestone opportunities that contribute to us reaching the guidance. Right now, our expectation of that plan is that we'll hit those milestones and we'll trigger that revenue. But it is drug discovery and drug development so we have to wait until we hit them to be certain about it. But that's -- and it's not a huge lift as you know in the numbers for this -- for the balance of the year either, so we're pretty confident about that, too.

Got it. Great. Maybe then we could talk a little bit about the internal pipeline. I know you tightened the guidance range for MALT1 with your past earnings -- most recent earnings update. Just kind of remind us, starting at the start, what led you to MALT1 as a good target for the Schrödinger platform? And more generally when you design therapies for targets like MALT1, what is the aim that you work towards when you're starting out? Is it better safety, better potency, better specificity, all of the above?

Yes, absolutely. So the reason we picked MALT1 speaks to our general philosophy that we like to work on targets where there's good evidence that there's potential for efficacy in humans. We don't tend to work on things that are being published in nature, for example. So MALT1 is in a very important pathway. It's in the BTK NF-kB pathway. It's a protease. The orthosteric site of that protease is very large and had peptidic molecules historically that hit that target, not very drug-like. And so we felt we could use that platform to design really very potent molecules at an allosteric site that really in of themselves bring selectivity. We also wanted to optimize all the drug-like properties. Potency, we felt was very important because in B-cell malignancies, NF-kB signaling is really a driver of this B-cell proliferation. And so having a molecule that can essentially shut down NF-kB signaling, which is genetically programmed into some subtypes of B-cell malignancies, we felt, was important. Secondly, the opportunity to combine a MALT1 inhibitor with a huge franchise in BTK inhibitor, we felt, was very, very important. And so early on in the program, we combined the early MALT1 inhibitors with BTK inhibitors. And so essentially those patient-derived tumors go into a complete regression. So we thought there was good sort of rationale for the target from a biological perspective. But very early on, we identified highly potent MALT1 inhibitors and it was actually an incredibly fast program from the discovery point of view. Within 10 months, we'd identified the development candidate, probably the most potent inhibitor that has been identified to that time point. And now we're in the clinic. We're just, as you pointed out, taking next steps with that program in dose escalation, studied a number of dose levels. As you know, we did a healthy volunteer study. Now we're in patients and the reason we sort of focused our guidance around data is that by the summer of next year, we expect to have completed a good portion of that dose escalation, gathered PK, PD. We're obviously gathering that now as well as clinical activity. And so we're looking towards the middle of next year at a scientific conference to share more information on the clinical program.

Got it. What is the right way to look at that data from your perspective? How should people analyze it from a safety perspective, from an efficacy perspective? What's the right benchmark?

Yes, it's a great question. Partly number one because the only other MALT1 inhibitor that's been published on in the clinic did have some safety findings that perhaps gave people a little bit of pause around the opportunity to get efficacy in a safe manner. So one of the key things that for us has been to demonstrate that our MALT1 inhibitor is well tolerated at doses that are hitting the target well. And we shared some of that data last year at our Pipeline Day in healthy volunteers. I think what we can say is that in patients, we're seeing a lot of similarity with respect to how hard we're hitting the target, and the safety profile looks very similar to what we saw in healthy volunteers. So we're pleased with that. I think it's very important that we show that this is a well-tolerated drug, particularly because small molecule combination regimens, you don't really want to see overlapping toxicity. You want to have a nice clean profile. So safety is very important. We'll obviously be reporting on that. And from an activity point of view, what's important here is that while MALT1 does have monotherapy activity, and that's been shown obviously preclinically, J&J showed very nicely and got proof of concept for the compounds, that you can have an impact in many different B-cell malignancies. We also want to show that there's monotherapy activity. So that's one of the things we're looking for in this initial dose escalation study. I will also say, though, that in terms of an apples-to-apples comparison, our first disclosure on MALT1 will be in a smaller population, obviously, than the many years study that Janssen ran. But nonetheless, we are looking for that clinical activity in our smaller trial.

Got it. So on monotherapy activity, it sounds like you're not putting out a specific number or a threshold just yet but you just want to see some activity?

Exactly. In the small number of patients, obviously, relative to a larger complete Phase I study. But yes, we've sort of got an internal benchmark for the threshold that we'd like to see to be sure that we're actually engaging the target fully.

Got it, okay. And I know you've mentioned the aspiration to potentially partner out these programs over time and then earn partnership economics as they move forward. Is following this next data update, could that be a reasonable time to look into a partnership?

Yes. I mean, I think one of the things that partners want to see, and I was in licensing at Merck so I would want to see this to safety, right, PK, is the drug behaving well, those types of things. I think completing the Phase I study is going to give people the information that they're looking for to sort of decide, are we going to jump in now on those combination trials with a BTK inhibitor or a BCL2 inhibitor? So yes, we think that, that will drive deeper conversations with partners around joining us or taking over for those combination studies.

Got it. I guess a question related to that, maybe for Geoff then. If you think maybe 5, 7 years out, if MALT1, other programs in your pipeline, they get partnered, you're earning economics on those. What could, I guess, the mix of the top line end up looking like from a software revenue perspective, drug discovery revenues, the new partnership economics? I know a lot of moving parts in there. But how could, I guess, the top line complexion change directly?

Yes, it's a great question. And we are sort of going through our sort of updating our 5-year outlook now. I think that in most scenarios, if we assume reasonable success rates for our programs, that we have an opportunity to have a much more balanced revenue profile as we get out to that time period. Five years doesn't quite get us to the point of having royalty economics on programs, but it does give us the opportunity to have lots of kind of intervening milestones. Of course, Karen's team is not sitting still. Either the kind of the platform and the capability and the group of people that came up with the 3 that are in the clinic are still working hard on further iterations and further opportunities. So we do think that we'll have an opportunity to advance more molecules. The caveat that I would say to that is that there are lots of opportunities that we may be able to partner even before we get into the clinic. And so one of the things that we've done in the past that probably isn't well understood is in some of those collaborations, we had proprietary insights. They may not have been development candidates, but they might have been leads or hits or something along the way that we put into the collaboration, in effect, partnering a molecule prior to going to the clinic. And we do think that we have the opportunity to do that. And as we get more and more validation, we hope to get more and more partner-like economics on those programs that we partner preclinically. So continue to come up with new molecules. Some of them go into the clinic, but some of them partner preclinically for economics that might get better in terms of resembling clinical milestones. I think that's the kind of outlook.

Yes. And therapy to carry a diversity plays into that. Obviously, in oncology, Phase I data is important but we think for some precedented mechanisms with either clinical or commercial proof of concept and very well-established mechanisms, a differentiated molecule in discovery could be highly attractive to a large pharma partner. For example, a small molecule version of a peptide or a biologic, we think those could definitely be partnered earlier on.

And in others, immunology...

Immunology.

Immunology, neurology, et cetera.

Immunology, neurology, cardiometabolic disease, for example. So some of our earlier programs are in those other spaces.

Got it. Do you think it's -- do you have the aspiration of ever running a full-fledged Phase I, II, III development program by yourself? Do you see any molecule or any situation where you feel like that would be more suitable than pursuing a partner?

I mean, I think it depends very much on the mechanism and the data package. I think as a general rule, that is not our strategy. However, if you're sitting on extremely compelling Phase I data, where the very next study could be pivotal or registrational, that's a decision we're going to have to make at that time. But generally speaking, if you think about all of our advanced programs in the clinic, they all have opportunity in very many indications, right? And so we think partnering with a large pharma that's going to be able to realize the full potential of those assets in combination probably makes sense. But that may not be true for some of the other assets where you have an opportunity to rapidly get to an approval. Now we'll make that decision when the data package materializes.

Got it.

We're going to be really pragmatic and realistic. We aren't likely to embark on very large complex combination development studies or huge pivotal trial programs or anything like that. But if we have an opportunity where we can take a program a long way forward even to commercialization that doesn't require that, then we'll certainly contemplate that. But those actually are quite different to the kind of programs that we can partner really for attractive economics, right? So as we focus more on those types of programs, the likelihood that they are programs that we'll keep ourselves actually goes down.

Got it. And Karen, on the point you mentioned around therapeutic area expertise and diversity, you do notice some tech-enabled platforms have pipeline focus areas kind of more heavily on oncology. Do you have any thoughts as to what's driving that? And then for Schrödinger's platform, is there a therapeutic area that the platform is just better geared to for some reason?

I would say that the platform is agnostic to target-type therapeutic area because as long as you have a structure for the protein and you have a good hypothesis as to how you're going to test activity, you can pretty much work on anything. That said, I think you're right that a lot of people, over the last few decades, have gravitated towards oncology because the preclinical models, the access to cell lines that generally speaking are useful in sort of making decisions around, not designed per se, but do I have activity in this cancer cell line, that allows a lot of people, obviously, to work in the space of oncology. We've got an amazing genetic data behind oncology and patient-derived samples. In contrast that with neuroscience, where there aren't a lot of good models. I think for us, we believe that actually collaborating in neuroscience makes a lot of sense because without those advanced assets or even approved assets where you know what good looks like and you can benchmark yourself against those, you're really kind of taking a lot of risk in neuroscience. So we love working on those programs. We like to do it in partnership, though. And so that mix is really around how can we get to proof points in discovery, proof points in the clinic that allow us to show the power of the molecules that we're designing without sort of ballooning the discovery team to have lots of therapeutic area expertise, but perhaps in some cases, lean on partners who already have that expertise.

Got it. In our last minute remaining, I'll just ask you one final question maybe on MALT1 and the readout for mid next year. What do you think that data set is going to help you answer and help everyone on the outside looking in and glean about kind of proof of platform, and what the appropriate read-through could be to future data readouts? What do you think is fair to glean from MALT1 to the other molecules in your broader platform and what do you think is not fair to glean?

I can start. I mean, I think we've already presented data that I think informs on our ability to design an extraordinarily potent molecule. So we've presented that data at ASH last year. The clinical readout is really around, does that molecule perform well with respect to the properties that we optimized? That includes potency, how hard we're hitting the target; PD, obviously a surrogate readout of how hard we're hitting the target; and also just how well the molecule behaves in humans, and that includes safety. We optimized the molecule taking into account some of our off-target screening that we've been talking about recently. Those are all very important aspects with regard to the compound design. I think what's not fair to kind of sort of read into the platform is, did the mechanism work, right? And not specifically for MALT1 but just in general. I think we all know that biology is very complex. And if a molecule's first-in-class target fails, it's not necessarily because of the design of the molecule, it's because the target was invalidated. But in our case, we're very pleased to say that MALT1 has been validated in the clinic already. And the bar to sort of match that or beat is what we're obviously very focused on.

Yes. There's already extraordinary validation of the platform. I mean, certainly it would help, but it's not -- we don't need that to validate the platform. I think that's all very clear, right?

Yes. I mean, if you think about Morphic, Nimbus, right, Tyk2, alpha4beta7, those are all programs we worked on, and I think they give a lot of evidence that the platform has performed very well.

And the thousands and thousands of users of the software, right, they keep renewing it every year.

Fair point. With that, we're actually at time and that's probably a good place to close out. Thank you for joining us. Really appreciate it. Thank you all for joining in.

Thank you.

Thank you.

Thank you so much.