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

Good morning, everyone, and welcome to Morgan Stanley's Global Healthcare Conference. I'm Sean Laaman, Head of U.S. Mid-Cap Biotech Equity Research here at the firm. Before we commence, for important disclosures, please see Morgan Stanley research disclosure website at www.morganstanley.com/researchdisclosures. And if you have any questions, please reach out to your Morgan Stanley sales representative. For this session, we have from Schrödinger, CEO, Ramy Farid; CFO, Richie Jain; and President, Head of Therapeutics R&D and Chief Strategy Officer & Partnerships, Karen Akinsanya. So welcome to the 3 of you.

Thanks for having us.

You're welcome. And maybe just to commence, we have some macro questions that we're asking all of our companies. And the first one is with China's rising biotech innovation, how are you thinking about Schrödinger's competitive position? And will this influence your R&D and business development strategy going forward?

Yes. Yes. It's hard to say that it won't, right? It's a real thing. I'd say, first, our technology, our platform was designed essentially to allow discovery of novel molecules, differentiated molecules. There isn't a reliance on knowledge of existing molecules. That's what -- that's a severe limitation of sort of machine learn based methods that are solely based on machine learning. So to the extent that our platform is built on first principles and our physics-based methods, it allows the discovery of highly differentiated molecules that can solve challenging design problems. And I think in an environment like we're in that you just described, that's particularly attractive to our customers, our partners and even our own internal programs. So that's -- I think do you have anything to add?

I just came back from China. We went there to sort of survey the landscape for ourselves. And I think that there's opportunities. First of all, people view Schrödinger's platform as the gold standard and that's what we kept hearing. And as those companies evolve to working on more novel targets and globalized, I think having relationships with those companies is something that is important for us to do just with our whole customer base.

Absolutely.

Wonderful. Great response. As an AI tech-enabled biotech company, can you describe the key ways your platform is leveraging AI and think about AI's future disruption potential?

Absolutely. So first, a sort of general comment needs to be made. AI and machine learning are only as good as the training set that they're trained on. That's what makes the model effective. So small training set or a training set that's not representative of the problem, machine learning models aren't very effective. But a large training set that represents the problem well and thoroughly, AI is very powerful. And we see many examples of that. That's the case also in chemistry and design of molecules. Now it turns out that if you take all the experimental data that's ever been generated by every company and you combine it, that's still the equivalent of a drop of water in the ocean where the ocean represents chemical space. So we don't have a lot of data to train on for chemistry. We have a lot of data for large language models. We have a lot of data for protein structure prediction, but for chemistry, there's not a lot of data. So we've developed methods using first principles, using physics that can produce experimental data, essentially, the equivalent of experimental data. But on a scale of course, that's many, many orders of magnitude faster and cheaper than experiment. So here's what we're doing. Here's how we're leveraging AI. We are building massive training sets using first principles, using physics. And that is essentially amplifying machine learning, right? Because, again, you need a training set. So we can generate actually in 1 day the equivalent of about 10 years worth of experimental data. I mean that's extraordinary and obviously, a fraction of the cost. So with these massive training sets now, we're seeing how machine learning and AI can have a really big impact by amplifying these physics-based methods. Now there's another area sort of different, which is you hear a lot about Agentic AI, right, agents. That's another really important area because these technologies that we're developing are new. They're complex. They're complicated technologies, and there's a shortage of people that actually -- that are experts, that can be able to run these. But there's no shortage of chemists and biologists on research teams. So we're also developing something completely different kind of application, developing workflows, automation agents to amplify humans to be able to run these kinds of sophisticated technologies more efficiently.

Sure. It's super exciting.

Yes, very.

Yes. Last question on macro side before I'll get Schrödinger specific. But what's been the most impactful on your company from the regulatory side, if anything, is it FDA, MFN, tariffs, like...

Yes. I'll say something and then I'll hand it over to Karen. I mean we're pretty excited as I'm sure you all are too about the FDA's request demand for developing computational tools for reducing animal testing. How is that done? Well, by designing safer molecules, right? So you don't have to test as many -- do as many tests and have as many failures and you put safer molecules into animals. So we're really excited about our, what we call, predictive tox initiative. It's funded by the Gates Foundation, with a rather generous grant. And we've made great progress. We actually released the beta recently. So that whole -- that announcement of the road map from the FDA has really been impactful in generating a lot of interest in computation. And that obviously benefits us tremendously. I don't know if...

Yes. I think in the near term, that's one of the most specific things. The FDA is clearly embracing AI and computation across the whole gamut from predicting safety but also in how they operate as an agency with respect to drug development. So those are welcome changes. Beyond that, I think a lot of these other topics don't necessarily impact Schrödinger directly.

Yes. That's right.

Sure, sure. I think we started last week, was it this week? Last week, publishing a regular AI publication as that first one. We've called it looking for my mom, which is maximally optimized molecule. So how does your physics-based platform combined with AI, machine learning, accelerate drug discovery? And what makes it scalable across programs?

Yes. I touched on that, but let me elaborate a little bit more. Again, as I said machine learning requires accurate training sets. So we can use the physics-based methods to generate accurate data sets that are on a scale that actually makes machine learning interesting, and that's accelerating drug discovery. It's resulting in getting to development candidates more rapidly, obviously, more efficiently, but also with higher quality molecules because we're able to explore way more chemical space. Now how is it scaling? The way we license our software is by the number of calculations that you can run. And that's tied directly to the number of molecules that you can explore computationally. So of course, if you have more licenses, you scale it, you can explore more molecules, generate larger training sets, the larger the training set, the better the ML model. The more molecules you explore, the more likely you are to find a molecule that has the properties that are required to be an efficacious, safe drug. So that's how -- I think that's the answer to it.

Yes. Maybe just one other thought here is I think we're looking forward as to how computation and AI-related approaches impacting quality of molecules. I think Schrödinger is in a very interesting place because as a company that's been in this space for 30 years, there are now molecules that are so much more advanced. We've got the Morphic molecule acquired by Lilly for $3.2 billion, the whole company, obviously the TYK2. So there are a lot of proof points about the impact of competition to look back on as well as to look forward to.

Absolutely.

Sure. What are the most compelling examples of platform validation from your proprietary and collaborative pipeline?

Yes. Karen. Go ahead, Karen.

Well, I just started talking about, obviously, the 2 big acquisitions of molecules. But actually, there's 15 molecules that have entered the clinic. Some are not disclosed. Obviously, they're from our collaborations with big pharma. And then we ourselves this year, published the second MALT1 inhibitor to go into the clinic showing that we have a very differentiated profile that was optimized using our platform, took us 10 months to find that compound and we only synthesized 80 molecules. And so those stories of the work we've done with our equity partners, Nimbus, Morphic, Structure, whole host of these companies and now with the programs that we're working on, not just with new collaborators, but our own pipeline as they move forward. I'll just point to the deal that we did with Novartis last year. While the target is not known, we can't tell anyone what the target was, that is, I think, another example of the company putting up $150 million because they thought we could help them win.

Sure. Wonderful. And maybe just to go back and double down a little bit on predictive tox. What feedback have you received from data testers so far?

Yes. We just released the beta and we just started to have customers starting to use it, beta testers, our partners, close partners. We obviously work with companies that we're very, very close to or we can rely on their feedback, and they won't punish us too much if there are little issues there here and there, which of course, there always are with betas, mostly technical things, just mechanics of running it. But we haven't -- we're not in a position yet to talk about the feedback. We're just very happy that it is out there. It is being used. We got the mechanics right. Sometimes these are very sophisticated calculations. They have to run in the cloud. So we've gotten over a lot of the barriers, but now we're just waiting for the actual feedback. That's going to take a little bit of time because you have to make the prediction, then you have to go and test it, right? So -- but we're looking forward to, in the near term, getting that feedback, incorporating it into the technology. That's how we keep improving. That's why all these interactions that we have with so many customers has really helped us build an incredible platform, so this will be yet another example where we will take that feedback, put it -- incorporate the learnings into the software, make another release and keep doing that iteration until we have yet another sort of new breakthrough technology like we've done before. So -- but nothing concrete I can say.

Wonderful. And I guess we touched on this before as it relates to predictive tox in the FDA's evolving stance on AI and drug development and -- how is that influencing adoption of your platform?

Yes. Right. I mean, we sort of touched on this earlier. I think everybody in the industry is well aware of the FDA's road map. I think companies are taking that seriously. And I think it's really helped in engagement with companies there. We don't have to spend a lot of time explaining why it's important what we're doing, let's put it that way.

Yes. Wonderful. Maybe to go down into the weeds a little bit, but SGR-1505 and B-cell malignancies. So looking back at initial Phase I dose escalation data presented at EHA and ICML in June of this year. What are the key takeaways from the data readout? How do you envisage its role in the treatment landscape for B-cell malignancies?

Yes. So we were very excited to release that information on our first molecule to go in the clinic. The first thing that I alluded to earlier was that MALT1 is a new mechanism, the prior clinical release from a third party has demonstrated dose-limiting toxicity, and that molecule was discontinued. So the big question was, is MALT1 safe mechanism that can be given to patients in an ongoing fashion. And we're very pleased with the safety profile of the drug. No dose-limiting toxicity, no deaths on trial. So we think that we derisk that whole question of safety. Very happily also we can report that we hit the PD target for this. So we've shown really shutting down NF-kappaB signaling and that translated into early signs of efficacy in a dose escalation trial. Now how can this be used in the treatment landscape? Very briefly, the treatment landscape in B-cell malignancy has been dominated by BTK, $11 billion franchise, also venetoclax and BCL2. There has been almost no other small molecules in the B-cell malignancy space. PI3-kinase came and then sort of went. And so MALT1 represents a brand new, in our hands, very well-tolerated mechanism for B-cell malignancies. And now the question is, how do you go ahead and follow up on that? And that's something we've been discussing with partners because we view this as a mechanism that mid-stage and beyond development is best done in partnership. And so yes, excited about having this new mechanism on the landscape for patients.

Sure. Sure. Still on 1505, How do you interpret the asymptomatic bilirubin elevation observation and how they compare to prior MALT1 inhibitors?

Yes, absolutely. So MALT1 is a relatively new mechanism, as I said, is protease. The orthosteric site, that's where it's a protease for that orthosteric site where the sort of ligands bind is pretty large. Those drugs were not very drug like. So everyone's gone after now an allosteric site. And that allosteric site for some reason does have -- the compounds have activity at what's called UGT1A1. This is an enzyme that also is responsible for clearance of bilirubin. So this class of allosteric inhibitors definitely has this UGT1A1 effect. The prior compound that I mentioned that we discontinued, essentially had grade 3 and grade 4 bilirubin elevations, including some signs and symptoms that would make it very difficult for patients to stay on. Our drug, on the other hand, if you look at the Grade 3 level, much, much lower. While we do tickle it, particularly in people who have mutations in UGT1A1, there's a disease called Gilbert's that people walk around with. So we do see a little bit of this, but we don't believe because there's no signs or symptoms that this is problem for SGR-1505.

Sure, sure. And maybe a question to bring it back to the broader audience. So you think about the development and discovery of 1505. Why could a machine do it and a human couldn't? What's the differentiating factor there that application of your platform? Does that make sense?

Yes, completely. I'll take a crack at that. Such a great question. So drug discovery is a very complex multiparameter optimization problem. When you design a molecule that's potent, which is pretty easy to do, you just add a little bit of carbon atoms to it, you make it a little bit more hydrophobic, it will tend to bind more tightly to binding site. But then it won't be soluble. And then when you try and improve the solubility by making it a little bit more polar, now it's not permeable. And you go around in circles like this, [ whack-a-mole], right? In the middle, while you were trying to improve solubility, you messed up the potency. Now you go back and try and fix the potency, now you have hERG or now you have a SIP or now it's completely insoluble and so on and so on and so on. It's a very, very complicated multiparameter optimization problem. And it turns out that if you just do things by brute force, which is basically just work on it for a few years, make a few thousand molecules, the chances that you find a molecule that balances all those things, where it's potent, selective, soluble, permeable and so on and so on and so on and say is extremely low. That's the statistic that you're all aware of. That's why 5% of molecules make it all the way through. It's because of that. What does computation do? It allows us to explore literally hundreds of billions of molecules. That's what it takes to find that really unique molecule. That's the ZAR molecule that somehow is both potent, soluble, permeable, selective. I mean that's a crazy thing. It shouldn't happen if you think about what I was just saying. So it's the scale. You have to explore huge numbers of molecules accurately. In other words, you have to be able to predict affinity, selectivity, right, all that accurately. Otherwise, of course, it's nonsense. So that's why it's the most complicated, I think, multiparameter optimization problem that we face as -- in humanity. I know that sounds kind of exaggerated, but I really think that's the case. It's really hard. And so you need that help of exploring hundreds of billions of molecules to find that magical molecule.

I guess, can you describe the reason for exploring strategic alternatives for further development of 1505. I think I know, but...

Yes. I mean I think, again, new mechanism on the landscape, there's work to be done, right? Any new mechanism enters the landscape requires deep work in the clinic. We talked about combinations. While we do -- we got Fast Track designation off the back of our dose escalation in Waldenström's, 100% response rate. But that requires a large company or a focused dedicated company to continue the development of this asset. In our configuration as a company, we think that that's something that's best done through some sort of partnership and that's why we've elected strategic options. We believe we've done a good job with the discovery and the early development and derisking of the molecule. Now it's time to hand over...

What we're good at.

Yes. It's what we're good at. And we think it's time to hand over to someone else. Now we're not handling everything over, of course. We're going to keep an interest in the program from a financial point of view, right? We've got royalties and milestones on a lot of different programs. It just so happens -- we're doing this on a program that we did the initial Phase I.

Sure, sure. Was that the reason you thought...

More or less. So I've got -- maybe to ask this question is that I wonder just thinking more strategically about your business that what it should be really good to set up to do is to get leads to -- in the hands of those that have the resources to conduct the trials. It's not the value that you had, the value in getting the molecule right.

That's right.

In theory, you've got, I'm not downplaying it, but an appendage of the pipeline, which is really just to advertise to the industry that this is what you can do.

Yes, yes. That's right.

And that's what the strategy is going forward. And therefore, just because you have out-licensed, it doesn't indicate in any way, shape or form that there's low confidence in the program.

Exactly. 100%. I think Karen said it really well.

I mean we've actually partnered pretty much most of the ideas we've come up with. This one just happens to have gone a little bit further than the others.

Sure. Still on 1505, but can you elaborate on the rationale and timing for pursuing combination studies with BTK and BCL2 inhibitors?

Yes. So as I mentioned, in Waldenström's, we're seeing this 100% response rate. It's very exciting. That's the monotherapy opportunity. But that's a small population. If you think about where BTK inhibitors were originally approved, it was an MCL, but people went on to -- sorry CLL and other large indications. We believe that the monotherapy in Waldenström's is the real opportunity, but it's small. And the big opportunity is now combining MALT1 in combination with BTK and BCL2 across all of those indications that have already been established for BTK. So that's the reason for that combination of science experiment that needs to be done, but it leads you very much into the commercial opportunity.

Sure. Wonderful. I've got some additional pipeline questions here, but what led to the decision to discontinue 2921 in AML and MDS? And what lessons were learned from its development?

Yes, great question. Obviously, unfortunately, we did decide to terminate that program. The primary reason for that was not the molecule. The molecule was a very, very nice molecule in terms of potency, selectivity, all the things that we designed into it. This was in relapsed/refractory AML, very difficult patient population, very huge unmet need. Those patients are immunosuppressed generally. Now we had along with academic KOLs identified that CDC7 was phenomenal at shutting down these AML cells, and that's why we went after it. The issue is that just like venetoclax, which is a huge drug, BCL2 inhibitor, this immunosuppression does leave patients susceptible to life-threatening infections. And that's what happened, obviously, during this trial. Now when we looked at the whole thing holistically, while there was activity and the opportunity to kind of do another venetoclax here, we decided that, that was not a good fit for Schrödinger along the lines that we've just been talking about, not a good way for us to be spending our time and money.

Yes. Wonderful. Can you give us an overview for expectations of upcoming Phase I readout with 3515 in advanced solid tumors?

Yes. So for those who don't know, 3515 is a Wee1/Myt1 compound. It benefits from synthetic lethality where these 2 mechanisms working together should open up the therapeutic index. So we've been in the dose escalation trial for just over, I don't know, about a year now, I think, basically studying safety, PK, PD and signs of preliminary efficacy. And so what we're looking for there is obviously very early because you can't compare to a Phase II or III study, but very early signs that we have hit the target and that we have essentially got initial signs of antitumor activity. So that's what we're looking for in this -- just as we did with the MALT1, understand with 2921 where we were, it would be the same with 3515.

Wonderful. Just deviating a bit, but on partnerships and commercialization. So can you give an overview of your most advanced biopharma collaboration partnerships? What validation do these partnerships provide for your platform?

So the most advanced -- I mean, as Ramy has said a couple of times today, our first partnership stemmed back 20 years, and some of those compounds have actually gone all the way to the market. So we were early collaborators with Agios on IDH1, Gilead with Nimbus, obviously, the Nimbus projects, we worked on with the ACC inhibitor that's in Phase IIb, I think, with Gilead and then with Morphic, that collaboration, that's in Phase IIb in the hands of Lilly. So there's quite a few late-stage compounds that I think either because they were acquired or they kept moving or that they were approved to validate that we can make molecules with this platform that enter development, stay in development and really helping patients.

Wonderful.

Yes. And Sean, these programs all have commercial milestones and royalties associated with them. We don't disclose those. We don't guide to that, but in our drug discovery revenue, what you're seeing today is a growing portfolio of collaborations. In the last few months, we've extended our collaborations with Ajax and with Lilly and with Otsuka. That's what's contributing to the growing drug discovery line. But in the future, over the years to come, there's another kind of layer of stream from milestones and royalties that we expect.

Sure. And how much visibility do you have on milestones and royalties, given it's a third party?

Well, I'd say during the active collaborations, there are milestones, obviously, during the discovery phase, we've got a lot to do with that because we're helping to steer those programs. Once things enter the clinic, our work is done, we do have ongoing joint research committees where we meet once or twice a year to understand how those programs are going. But less visibility, clearly, and there's also portfolio and pipeline strategy at these companies that we have nothing to do with.

Okay. Thinking about 1505, what would an ideal partner look like?

I mean I think the most important thing in the very near term is a focus on the mid-stage development, getting those combination studies and potential registration studies designed and executed. But really, I think a commercial powerhouse that has already established a franchise in B-cell malignancies would be ideal, right, to be able to leverage that ultimately, I think, will be the way to maximize the potential of MALT1.

Sure. And I guess with broad of these development programs, like where does the sovereignty of the data set? Who owns it? Can you actually generate a data set from a partnered program, but then you can leverage that data to...

That's a really important question. There's 2 ways to think about data. The first is the IP that then goes on, obviously, with the compound all the way to the market. That IP is exclusive to our partners. And at the moment, that we partner the program, we hand over the IP and all the data that's related to it. But there's another kind of data...

Yes and I can address that, yes. So I can tell you every agreement we have, every one, any improvement to the technology or the platform we own. That's the answer to that question. And no matter what form it comes in, we own that outright.

And then there's a very strict firewall that separates off collaboration data from our platform data.

When we say there are synergies between the drug discovery and the software, that's one of them, that we own -- we bet all that know-how, all the improvements to the platform, everything we learn, that gets all incorporated in the platform and the whole industry benefits from that, including us, by the way, from our own programs.

Wonderful. We touched on Ajax collaboration. Can you share more about the -- more detail about the expanded collaboration and its potential impact on milestones and revenue?

Richie?

Yes, I'm happy to cover it. We have -- we expanded the collaboration a few months ago to add another JAK target in the I&I space. In terms of the economics on that, it mirrors the original agreement, but we've expanded it to include commercial milestones as well as royalties. I would not expect those to contribute meaningfully in the near term, but it creates another long-term opportunity for us.

Wonderful. I might start asking a little bit about the financials and the longer-term outlook. But how are you managing operating expenses with scale and clinical development? And what impact does the recent restructuring that I think was in May had on your financials?

Yes, I'll start there. I think I wouldn't say that we're scaling clinical development. We've spent a lot of time talking about 1505 and seeking a partner there or seeking out strategic partners. Our 3515 program is early in the Phase I stage. But we aren't guiding and we're not talking about adding additional programs into the clinic. So just to address the clinical spend piece, from an overall company expense profile, we announced in May a reduction of $30 million of operating expenses, most of that will be realized this year. Some of that will still balance out in the first half of next year. But overall, if you look at our results for Q2, we have a great profile, which is growing revenue, mid- to high teens, but also year-on-year expense decreases, mostly driven by the R&D line item.

Sure. And the software revenue growth, I think it's 10% to 15% for...

That's what you guided to this.

Just sort of the drivers behind that. And I guess sort of closing the adoption gaps amongst the mid-tier customers.

Yes. I can.

Go ahead and start.

Yes. The primary source of growth at the moment, to the extent that biotech companies aren't doing so well right now. It's pharma companies, our existing customers. So every pharma company is using our software. But there's a pretty big range and how much they're using it. We have a handful that are using it at a very significant scale, and then another handful that are quite a bit lower scale. So a huge opportunity and we think it's inevitable. Of course, every pharma company is going to be using it at same scale, but it's a process. Some pharma companies are a little bit slower than others. So growing those large pharma companies that are sort of utilizing the software at a smaller scale is an opportunity. It also is an opportunity -- even the customers, this is important. Even our largest customers are actually still underutilizing the software relative to what we're doing internally. So that even in and of itself is an opportunity as well. But that's the -- that's a big part of the focus.

Sure. And maybe help me model question. But how do you measure success in your software business beyond revenue? What metrics best reflect customer engagement and platform adoption?

Yes. ACV is a metric that we focus on for customers greater than $5 million in ACV. In 2023, we had 4. In 2024, we had 8. So that's grown substantially. Where we keep future growth is the divide between customers greater than $5 million and customers greater than $1 million, 8 customers greater than $5 million, about 30 customers greater than $1 million. So that's where we seek the opportunities to step up relationships, deploy our technology at scale and increase the throughput of the access that they have.

I think the retention rate for customers that are spending over $500,000, over $0.5 million, is essentially 100. I mean that's a pretty important metric. That says a lot about the efficacy of the technology. I don't think you are renewing the software over and over again, right, if it isn't actually working. Another thing I looked at recently, I don't know if this is interesting enough, but I'll just throw it out there anyway is we looked recently at the number of patents that pharma companies and biotech companies submit where there's mention of the use of our software, like over 2,000 or something in recent times. I think that's another metric. So it's not exactly a financial. It doesn't show up in the SEC fund. But I'm trying to give you a sense, right, to answer your question about just indicators, right? As you said, metrics that demonstrate something about the efficacy of the platform, the potential for it to grow, the impact it's having that are just beyond the sort of revenue.

Sure. you. And your long-term vision for balancing profitability with investment in the platform and innovation. How do you think about that?

Yes. Well, profitability is a goal. Obviously, we're not guiding to exactly when, but I think it's a strong statement that it's a goal. And so -- but to do it in a way that allows us to continue to innovate the platform. We are the leaders in this space. We are innovating. We define the field. All the breakthroughs here about in this space are -- happen inside Schrödinger, and then other companies come and try and replicate it to not very successfully because of the effort that we're putting into it. So we think that's really important. Where our goal is to achieve profitability, continue to grow the software business, but do it in a way that allows us to continue to innovate in the platform. It's such an important part of our business. And then I think you heard how we're helping the situation by our plans around partnering the clinical programs at the point where they become very expensive, which is obviously the mid-stage and later-stage clinical programs. So that's -- well, that's pretty good timing.

Perfect. But is there any message you would like to leave the audience with before we close today. Any message?

We haven't already covered. I think this conversation has been quite good at covering everything. So yes, no, I don't think so. Great job.

Fantastic. Well, thank you to 3 of you for participating. Wonderful to host you.