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

All right. Thank you for joining us for our next session. My name is Mike Ryskin. I'm on the life science tools and diagnostics team here at BofA. Joining me here is the senior analyst on the team, Derik De Bruin. And with us for this section, we have Schrodinger. So we're joined by Ramy and Karen. I think -- first of all, thank you for joining us today. I'm sure it's a very busy day. You guys reported first quarter results this morning. Probably could use a break, but we're going to try to squeeze 30 more minutes out of you.

No, no, we're good.

Already got a breather. So I think to start, I think you had some prepared remarks and maybe some slides to go through. So let's kick off with that.

Sure. And I understand everybody has access to those slides. So if you navigate -- read every word in Slide 2, of course. And then when you're done reading everywhere in Slide 2, you can go to Slide 3, and that's where we'll start. All right. I'll give everybody a second to do that. Good. So Mike, I'll get started?

Yes. Yes, go ahead.

Good. Great. Great. So I'll start by stating something very obvious. We all know that drug discovery and materials design are incredibly hard. The large majority of programs failed to deliver a product. And of course, we all know very well that costs are really skyrocketing, due in large part to the very high failure rate of programs, both in drug discovery and materials design. And even when programs do succeed, the compounds often have issues in drug discovery that takes the form of toxicity often. So an obvious solution to this problem really is to develop computational methods that allow exploration of vast amounts of chemical space so that we can identify high-quality molecules. So a way of stating that is what it says on the slide here. So if we can prospectively compute all the relevant molecular properties with sufficiently high accuracy, designing drugs and materials would have a much higher success rate, be much faster and cheaper, of course, and maybe most importantly, would yield much higher quality molecules. So I'll elaborate on that a little bit more. So if you go to Slide 4, this is really now the state-of-the-art. So Schrodinger has made really major breakthroughs in the science. And we've developed, for the first time, technology that can compute key molecular properties with extremely high accuracy and in fact, approach that of experiment. So the technology now makes it possible to explore hundreds of billions of molecules, and it's having a really truly transformative impact on drug discovery and materials design, which you can see on Slide 5, so if you go there. And traditionally run drug discovery, it's typical for about 5,000 molecules to be synthesized. And when successful -- when the programs are actually successful, takes about 5 plus or minus a year, 4 to 6 years or so to get to a development candidate. And again, it's still often the case that those development candidates have issues. So by leveraging our platform at scale, and that's an important point, at scale, we and our partners are able to explore now billions of molecules computationally synthesizing far few, only around 1,000 molecules, sometimes actually now and more recently even fewer than that. And it takes about half the time to get to development candidate. And again, this is really the key point, with extremely high-quality molecules. So if you go to Slide 6, you'll see how we're leveraging this really truly unique platform and how we're doing it in several ways. So we have a software business where we license our software to pharmaceutical companies, biotech companies, materials companies, universities and government labs worldwide. We're also leveraging our platform in a number of drug discovery programs in collaboration with pharma companies and biotech companies. And we also have an internal drug discovery pipeline, and Karen will tell you more about our collaborative drug discovery programs and our internal pipeline in just a moment. I just wanted to highlight here that we have a track record of revenue growth. And as Mike said, I'm not sure if he said it before we came on or not. But anyway, this morning, we reported our Q1 revenue of $32.1 million. That represents 23% growth from Q1 last year. So I'll hand it over to Karen, and then I'll come back. Karen, we're on Slide 7, I believe.

Thanks, Ramy. So as Ramy just highlighted, we have deployed this platform across a large number of collaborative programs, and I'll just spend a minute to characterize the different types of collaborations we have. You can see here that these have been applied to programs that are now advanced and in the clinic and indeed, in our early collaboration with RGS, those programs are now FDA-approved. In addition to those, you can see that early on, about 10, 15 years ago, we had been participants in launching companies who are exclusively using this technology to develop their pipelines. So Morphic, for example, and Nimbus, those programs that you're familiar with, came through using this approach to drug discovery. And those programs are now in advanced stages of clinical development, in some cases, or in Phase I. And [ new sets ] are moving into IND-enabling studies. And I think what's important to note about those collaborations is that these are across all therapeutic areas, all target classes. The technology can be used on many, many different types of drug discovery programs, protein-protein interaction inhibitors, kinases, proteases and so on. In addition to Morphic and Nimbus, there are a number of new companies listed, Ajax, Faxian, Bright Angel, Petra and ShouTi. These are also biotechs that we work with on a number of programs. In addition, you can see that we have collaborations with pharmaceutical companies, ONO, Sanofi and Takeda, where, again, while those companies buy our software, they're also working directly with us to realize the full power of the technology. We deploy these methods across the whole drug discovery program from identifying new starting points all the way through to a development candidate and many of these programs we're able to do that at very high scale, and that's what these collaborators are benefiting from. In the case of Takeda, we run those programs in a cross-functional manner so we actually deliver our development candidate to Takeda. So moving on to Slide 8. Over the last 3 years, we've actually embarked on running our own wholly-owned programs. This is why we've selected oncology programs for Schrodinger's internal drug discovery team to first come up with the ideas as well as push them all the way through to development candidate. So you can see here, CDC7, WEE1, wholly-owned programs. Last year, you may be aware, we did a transformational business development deal with BMS, where some of our earliest wholly-owned assets were actually included in that deal, 2 oncology programs, that we had previously disclosed a third, and then an immunology and neurology target that our teams had come up with. In terms of the progress on the pipeline, I will say that the 3 wholly-owned programs are in advanced stages now. We are moving these forward into IND-enabling studies and expect to be in the clinic in the first half of next year. In terms of the update we provided this morning on Slide 9, in our MALT1 inhibitor program, which we presented at ASH last year, we have actually now named our development candidate, and that is moving forward into IND-enabling studies, as we said, and the other 2 programs are following. We're also building out our capabilities to support our clinical development activities, including regulatory CMC, DMPK, ADME and all that you'd expect to see in a fully integrated program moving into the clinic. And so we expect to actually submit up to 3 INDs next year, with the first being in the first half of '22. In addition, we are planning to replace those targets, so expanding into additional disease areas. We've spoken about immunology previously. And so we'll be initiating new programs this year. And with that, I'll hand back to Ramy.

Thanks, Karen. So on Slide 10, it's a little bit of a busy slide, but let me walk you through it. It's pretty important actually. So we're obviously really excited about what the future holds. So to give you a sense of what the next decade is very likely to look like. It's useful to actually see how we got to where we are now, and I think that will give you a sense about where we're going. So of course, you all know that computer performance has been increasing exponentially for a long time, continues to do that. Now but what may be less well-known is that the number of proteins in the human genome, for which we know the 3-dimensional structure, is also increasing exponentially. The number of compounds that we can explore computationally is also increasing exponentially as is the number of properties we can compute with near experimental accuracy. So what does all this mean? It means that we can reasonably predict within the next decade or so that certainly computers will be probably more than 100x faster than they are today. I mean that's really pretty extraordinary. High-resolution protein structures of probably around half of the human genome, I mean, high-resolution protein structures of approximately half of the human genome will be available. We're probably around 10 or so molecular properties that we can compute with experimental accuracy. And I think this all leads in the next decade or we believe this really will lead in the next decade to being able to reliably discover extremely high-quality development candidates in about a year from program launch. So it's clearly a really exciting time for computationally driven drug discovery and of course, in general, molecular design. So on the final slide, on Slide 11, I'll just close by highlighting our key growth strategies. So we will, of course, continue to advance our computational platform to fully realize the vision for the future that I described earlier. And at the moment, we are the clear leader. And we intend to maintain our leadership position by continuing to aggressively invest in the science that underlies our platform. It's a very important part of our strategy. We'll continue to license our platform to pharma and biotech companies. We'll continue as well to advance our collaborative drug discovery programs and initiate new ones, as Karen talked about. We'll continue to license our platform to material science companies. We'll continue to advance our internal drug discovery programs, and again, as Karen said, initiate new ones. And we expect in the future to initiate material science collaborations, as we did a number of years ago with drug discovery collaborations. And I'll just leave you with this final point, which is a very important point, is that these multiple business areas and growth opportunities is the extraordinary synergies between all of these areas of our business. I know it may not seem very obvious, but there are -- maybe we can talk about that later in the Q&A. These extraordinary synergies between these business units and that we're continuously leveraging. That's a very important part of the growth strategy. So thanks, everybody, very much, and happy to answer any questions.

I want to remind investors really quickly that if you've got questions, feel free to submit them via the portal. You should see a question field right there or even just Bloomberg chat us or send an email to Derik and I, and we'll try to get them in. I think just to get the ball rolling a little bit. I want to ask sort of like a high level one that we've gotten a lot over the past year, 1.5 years since the IPO is if you just think about computationally driven drug biology -- drug discovery, there are a number of these platforms coming up, right? There's a number of machine learning, fast computation platforms that are all working to improve R&D efficiency and sort of disrupt the traditional paradigm. So one of the questions we get a lot is why can't Google or Microsoft just throw -- or Amazon just throw couple of hundred million dollars to this problem, come up with the best AI and machine learning system out there? So what's your secret sauce? What's in the core of the Schrodinger platform that gives you a leg up that can't readily be replicated by just -- by anyone else right now?

Yes, yes. So look, we're very proud of the leadership role that we played in this. So this is a field that actually goes back more than 30 years. I think people -- a lot of companies have been trying and academic groups have been trying to solve this fundamental problem of predicting properties of molecules prospectively. And we are the first ones to be able to achieve that. And here's how we did it, and now this will answer your question, then I'll talk a little bit about the other approaches. So this idea -- there are basically 2 approaches at a very high level. One is what we often refer to as knowledge-based. That's now people are calling that machine learning and AI and using all kinds of fancy words, but really what it is knowledge-based. And what does that mean? That means that you -- and this is what you were -- I think what you're alluding to what Google might be able to do is say, well, we have a lot of information right now about compounds, about targets and so on. So we can feed that into some kind of model, and now we have much more sophisticated deep learning and convolutional neural networks and so on. And maybe that will somehow change the potential for these kinds of technologies and of course, the availability of more data. But here's the thing that's really important to understand. The size of chemical space is almost infinite. It's estimated to be around 10 to 50, 10 to 60. What does that mean? That means that the size of the training set that's required is enormous. In fact, it's larger than can possibly be ever produced using sort of traditional experimental methods. So even though it's exciting to say we have a lot more data and we have these sophisticated machine learning methods, it's actually not possible to develop a global model for any of these properties. And with the limitation then of these methods, these machine learning or knowledge-based methods, is that you only know about -- you can only know about the data that you already have. And then my point is that data is incredibly small. It's a very, very small amount of data relative to the size of chemical space. The other approach and what we've done is we've developed first principles methods using what we call physics-based methods, which capture the actual physics of the problem. So if you want to calculate the affinity of a small molecule to protein, you have to actually simulate the molecular interaction. You have to simulate at an animistic level of detail, all of the terms that are required to be able to compute something like Affinity. And that allows us to go into new chemical space. You're not limited. This is very important. It's not a knowledge-based method, this is the first principles method. There's no training set. There's no -- you don't need knowledge, you just need an input structure, the input structure of the protein, which is why I highlighted that as one of the exciting things that's happening now and into the future, and you can actually compute these things using rigorous methods. Now here's the really interesting thing. These methods, of course, you don't get anything for free. These methods, as they're incredibly accurate, but they take -- they're computationally expensive. They take about a day to calculate on one GPU for one molecule. And so what we're able to do now, and this is a really extraordinary way of kind of combining these 2 methods, is we can develop enormous training sets using these physics-based methods because it's exactly like doing an experiment, but we can do hundreds of years of experiments in a day by, of course, running these things in parallel. So just think about that. You can collect an enormous amount of data using these physics-based methods, which you can then use to develop a local machine learning model which can be used to process, and that's how we can get through hundreds of billions of molecules. You can't run physics-based methods on that many molecules. And they're still not very accurate models, but they're good enough to essentially enrich the top, say, 10,000 molecules out of the 1 billion where there's some good molecules in that top 10,000, then you put those through the physics-based methods. So the secret sauce, and I'm sorry, the answer is kind of long, but sort of important. The secret sauce, if you will, is these physics-based methods. But I think it's also the combination of machine learning with physics-based methods that's allowing us to do the sorts of things that we talked about today.

Okay. That's a very helpful answer, Ramy. I actually appreciate that. Following up on some of those comments, and this is something you kind of highlighted in Slide 10 of the presentation, I thought that was a great over slide. But if you could expand on that a little bit in terms of what's changed in your algorithm? In your system? What's changed on the data sets over the last couple of years? Whether it's compute power, whether it's the cryo-EM structures themselves, sort of -- and what I'm getting at is one of the things we've always talked about is sort of like quantifying the value add, if you could just put a dollar or some other metric on it. Where is it today versus where it was a couple of years ago? Because this helps us think about how quickly pharma will ramp up spend over time in utilization of the platform. If you think that 10 years ago, there wasn't very much value add, 5 years ago, there's some. Now there's a lot going forward, there's a little more.

Right. So there are 3 things, basically. One, is the fundamental accuracy of the methods and the number of properties that we can compute. And that's something where, as you know, nearly half of the company means a couple of hundred people in the company are devoted to working on that. That's the basic physics, the science to improve the accuracy. The more accurate the predictions are, obviously, the more impactful it is. But the interesting thing is -- and we can go back to that. But computer speed has played a huge role because of what we were talking about before. How do you solve this incredibly hard problem that is drug discovery? And we're convinced of this that because -- remember what drug discovery really is, it's this incredibly complicated multiparameter optimization problem. You're trying to design a molecule that's potent and selective and soluble and permeable and all these other things. But those properties are fighting each other. I'm not sure people really fully appreciate that, that when you make a molecule potent, usually, that's not very soluble. And when you try and make it soluble, it's not very permeable. So you get the idea. So it's this incredibly it's very fine tuning. Now how do you solve that? Well, you calculate the properties accurately, but the other thing is you have to explore huge amounts of chemical space, because you're sort of threading this fine needle, right? Just to find that perfect molecule that balances all those. All these properties that are fighting each other. And the way to do that is to explore hundreds of billions, trillions even more, and I think that will continue. And that's the point of that slide is that with compute power -- with expanding compute power, we can now get to a point where we're exploring enough molecules. And that's what it takes, hundreds of billions of molecules, to find those perfect molecules that have the potency and the selectivity in all the other drug-like properties. Now the other thing that's really exciting, and you touched on this, is the input to these physics-based methods is a structure of the protein. And the explosion in availability of high-resolution structures through advances in cryo-EM, but also in [indiscernible] and, by the way, in computational methods for refining those structures because that's important. It's expanding the number of targets we can work on. So even just in the last few years, just the number of targets that are within the domain of applicability of technology is really increasing. And that has implications not only for our own internal programs, but for the whole industry, right? The more targets that are amenable to this kind of design, the faster we'll get to development candidates on a larger class of targets, which has obviously incredible implications.

That's very helpful. Karen, maybe I'll pose one to you. I had a couple of questions coming from clients. So one is for the wholly-owned internal drug discovery candidates, if successful, does Schrodinger intend to develop these on their own or a self partner? And if you could sort of expand on how you think about different stages, pre-IND, post-IND Phase I? Sort of how that factored into some of the Bristol conversations last year, just sort of cover those bases.

Certainly. So I think for each program, it really is sort of program dependent, to some extent, but I'll just make some general remarks that we are looking at lots of targets all the time. And some of them, I think, are appropriate for early partnering, and that's, to some extent, what happened last year with BMS. Now for the advanced molecules, where we've identified late-stage molecules or development candidates, we've decided that actually taking them into the clinic and generating a Phase I package is certainly within our scope of what we'd like to be able to do that we can fund and support from a cross-functional team perspective. We have an early development team now, which is gearing up to do that. In terms of the outcomes of each of these studies, whether it's a preclinical late stage, like our ASH presentation, or it's a Phase I package, we think that gives us a lot of optionality for some mechanisms that are really gaining a lot of momentum in the industry. We're not [ averse ] to partnering those because in some cases, actually, the combination agents that you require to fully maximize potential of the mechanism are available at another partner. And so whether we do a co-development, clinical collaboration or a preclinical collaboration, I think it really depends on the target. But at the moment, our view is that taking these programs into the clinic ourselves, expanding the data package, showing proof of biology and in some cases, were needed because some of them have proof of concept already. Taking those further, I think will lead to more interesting business development opportunities for Schrodinger where we're able to capture more value from the assets, and in some cases, have a long-term potential revenue coming from royalties on what we believe are some pretty interesting and derisked targets.

Yes. And it certainly feels like it's a flywheel effect where you've got the software, you've got the collaborations, you've got the internal pipeline and all of that kind of helps fuel the rest of the business and it all kind of works together. So...

Correct. As Ramy said, there's a sort of virtuous cycle here. And I think as we've pointed out, as these molecules move into the clinic or get partnered, our goal is to replace those with more exciting programs.

Okay. One other one came in from an investor while we were talking. It's about -- it's something we've discussed a number of times with clients over the last couple of weeks and months, sort of the pacing of revenue over the course of 2021 in particular on the software side of things because we understand that drug discovery can be a little bit more volatile. So it's a question on sort of why is most of the revenue coming in the fourth quarter? I understand that it's based on when contracts were signed last year, the comps get tougher. So maybe if you could talk through, was there really a step function last year? During the middle of the year? That people really start spending more to 3Q, 4Q, and that's when therefore, you're getting the same seasonality this year? Was there any -- how much of a role did COVID play in this sort of timing or pull forward of things?

Yes. That seasonality, by the way, really goes back a very long way. I mean that's been historical for as long as the data I've been looking at, 20 years or so. It's seasonal, and it has to do, of course, being tied to pharma budgets, and that's just when they're doing their contracts. And look, as you know, quite a number of our customers have been with us for decades. And so again, this can go back. Just that happens to be when the contracts were signed and that's when the renewals are. So that explains the seasonality. With regard to the other question about what happened in 2020, it was an incredible year, absolutely incredible. And it might have been -- what happened may have been catalyzed by COVID. All of a sudden, everybody found themselves not being able to get in the lab and not being able to make as many compounds as they were used to using. So obviously, there's a sort of demand, well, we need to do something. And obviously, essentially doing synthesis and computational -- and assay, sorry, on a computer using computational assays looked quite attractive. And a large number of companies adopted the platform. Now I think that had to do also -- I don't think it's a coincidence that, that was happening at the time when our internal programs are progressing. We were talking about those. There was a lot of validation, right? The progress that Morphic was making, the progress that Nimbus was making and a number of other companies we're involved in. I think that obviously played a role. And so we're excited about that. I mean I think that's a really good sign. And we're seeing now -- this is really exciting. For the first time, pharma companies are now talking in public about the impact that our software, our platform is having on their programs. And I think you know how encouraging that could be. I mean, there's no better marketing, if you will, I almost hate to call it that. But obviously, that's very helpful to have other companies talking about the impact that the software is having. I think other pharma companies really pay particular attention to those sorts of things. So that's what we have to look forward to this year. And of course, the following year and the year after that.

Yes. Yes. Derik, I think you had a question?

Yes, actually. So obviously, there's been a number of companies that have shown up in the biology space, a lot of them are focused on material sciences. You have a material science program. Could you talk a little bit more about that? And when we're going to start to see some collaborations? I mean I always thought that was a very intriguing part of the platform and to something you just really don't talk about as much.

Yes. No, that's true. It's a younger business. It's a smaller part of our revenue. What is really important to point out here, though, is that because we have developed these fundamental physics-based methods that are, of course, completely agnostic, not only to the modality and drug discovery projects, for example, or target class, but also, of course, to the systems themselves. And the physics that governs for example, the behavior of proteins and water, protein being a polymer, is fundamentally the same as an organic polymer that might code an airplane wing. And so we've been leveraging a lot of that same technology in a number of different verticals and material science. We talked a little bit about that, OLEDs, polymers and now most recently, batteries, and we're really excited about that. Here, we have to, again, building on the existing science, develop new technology, for example, simulating chemical reactions at the interface between the solid and electric -- and illiquid. That is the electrode of a battery and electrolyte. And so that's a basic research project. We're excited about the progress we're making. And we're excited to start -- maybe later this year, start to talk about some of this technology. But I'll tell you, we have quite a number of customers. They are to publish papers actually. So if you look, there are quite a number of patents actually and papers that are coming out of these companies in a number of different verticals using our technology. So I think the word is starting to get out, but it is a younger business. On operations, I didn't answer. So we obviously -- I think you can see, we benefited tremendously from starting to do that. I mean that was really enormous, and it really started, for the most part, with Agios and then with Nimbus in a really significant way. And so that's not lost on us. And we're looking to do something similar on the material science side as well.

Okay. I think we're almost out of time.

Yes. I just thought we'd go to 0. Sorry.

I want to ask one last question, sort of our concluding question is Derik always likes to ask sort of what's most misunderstood. I'll put a slightly different twist on it. Sort of what's the biggest debate point? What question do you get from investors the most? Where do you think -- I don't know if controversy is the right word, but sort of what do you think the focus is? And sort of what would you like to address with investors?

No, it's a great question. And I think there's a really obvious one, and I sort of touched on it, which is we have a software business, and we're providing software to the whole pharmaceutical industry now more and more to materials companies. And at the same time, we have a biotech business, both collaborate, right? Collaboration programs and internal programs. And that's a little bit difficult for people to immediately get their heads wrapped around, right? How does that work? And what you're competing with your customers? And why are you doing that? And are these 2 separate things? And I think it's actually an incredibly important feature of the company. And I think it's actually one of the key ingredients of the success of the company. The fact that we have a software business where we need to produce software that we give to somebody else, and they have to use it and get value out of it, that's a lot different than developing software that's a black box that you just need to use yourself. I mean it really -- that's a different proposition. Now I think that's led to the advances. And of course, think about that, thousands of customers we've interacted with, how much we've learned from that? And that's all been put into the software that's helped us drive these projects. And then I think on the other direction, it's more obvious, which is, of course, the success of the drug discovery programs has really, I think, woken up the industry and gotten people excited. And without that, how does a software company convince pharma companies to completely change how they do drug discovery? That's not going to -- that's a very difficult thing. And believe me, we tried that before we had things like Nimbus. So I think it's something that's a little bit complicated, right, and a little bit hard to explain. But I think once you hear, hopefully, what I just said, it convinces -- I think it's -- hopefully, it's convincing that this is actually quite an important feature of the company. And I think another -- in some sense, a competitive advantage, I think.

Yes. Yes. All right. Thank you so much. We're over time. I appreciate you joining us. Thanks so much. Thanks for listening in. [ I season ] is opening up shortly, so Derik is making me ask. So keep us in mind when the vote opens up. Thank you, everyone.

Great. Thanks a lot. Bye.

Thank you.