Welcome, everyone. Thanks for joining us. We're kicking off our virtual healthcare conference today and kicking things off on the healthcare technology side. We have Schrödinger here. We have Ramy Farid, the CEO, joining us today. Ramy, I believe that you wanted to present some slides first, and then we'll dive into Q&A. Thank you so much for joining us, Ramy.
Of course. Yeah, I'll just present a few slides, just high-level overview. Should I go ahead and get started?
Yeah, let's kick it off.
Perfect. Yeah, let me start by first describing really what we've developed over the last 35 years and our platform and what we're trying to do, what the goal of the company is. The goal is to essentially develop computational methods that replace experiment, expensive, slow, unreliable experiments. Really, the goal is to predict at the highest level to predict the properties of molecules. Obviously, if you can do that, design molecules and predict their properties, that will have a pretty significant impact on the ability to design molecules. It should improve the performance, get there faster with better molecules, improve probabilities of success. Over the past 35 years, we've been developing methods for accurately predicting the properties of molecules. There are, at the highest level, two ways of doing that. One is using first principles, using physics.
The other is using machine learning or artificial intelligence. It turns out both of these are pretty powerful methods, but they have limitations. The physics-based methods that we've developed over the last 35 years are incredibly accurate. It's pretty extraordinary. This is something that you can imagine. It's not so straightforward to actually simulate, for example, a molecule binding to a protein and all of the atoms associated with that process and compute from first principles, the binding affinity, for example, of a molecule. We've done that. It's highly accurate, but it's computationally expensive. In the context of a drug discovery project where hundreds of billions of molecules need to be explored, there aren't enough computers in the world to be able to do that in a reasonable amount of time.
Now, machine learning has the advantage, as many of you have heard, of being pretty efficient, fast. It has this other limitation that everybody's very well aware of, which is that it requires, and in the case of chemistry, very significant training sets, very large training sets because of the diversity and complexity of chemical space. What are we going to do? We've got this fast method, but kind of by itself, without robust training sets, without large training sets, it has this limitation. You have these physics-based methods that are accurate but slow. Obviously, if you combine them, you can do something pretty extraordinary, which is you can use these physics-based methods, these first principles methods, to produce very, very large training sets, the kinds of training sets that are essentially impossible to produce using experiment.
We can, in one day, using physics, produce the equivalent of 10 years of experimental data. You can imagine how powerful this method is, where you can use physics to generate large training sets. That essentially allows AI and machine learning to actually have an impact in drug discovery by combining both of them. That is the platform we have been developing over the last 35 years. Now, how are we applying, if you will, this platform? As I am sure everybody listening understands and knows, we license our software to life science companies, drug companies, pharma, biotech, globally, and material science companies. Material science, life science applications are, of course, very different. The fundamental physics, of course, is the same. You can see we have quite a number of customers worldwide.
We also use the platform in collaborations, both drug discovery and materials design, with, in some cases, companies we've co-founded and in some cases with large pharma companies. You can see there we have quite a number of collaborators. More recently, we initiated efforts to advance proprietary pipeline. We'll talk about those in a second. Those are in discovery, and there are a few in the clinic. You can see there we have more than eight active programs. You're going to see as we're talking about this, and I'm sure in the Q&A, these are highly synergistic, all three of these. I think we're going to talk about those in the Q&A. Let me just give a quick highlight of the last year. You can see in the fourth quarter of last year, we had quite a good quarter.
You can see there that the software revenue grew by 16%. Total revenue was $88 million. You can see the drug discovery revenue. The full year was also quite a good year. We exceeded expectations. Software revenue growth was 13.3% above the guidance that we gave, which was 8-13%. You can see there the total revenue. Maybe more importantly, our guidance for 2025, you can see there that we're guiding to software revenue growth of 10-15% this year. You can also see that we're guiding to quite a significant increase in the drug discovery revenue, 45%-50%. You can also see there that we're also gaining some operational leverage. You can see there that we're guiding to less than 5% OpEx growth. I think we're going to talk about this in the Q&A as well. This is very important.
We've managed to expand existing collaborations. You can see there with Lilly and Otsuka. I think that's an indication of the success of the existing programs that we're working on together with them. We also have reiterated our guidance that we gave last year to presenting clinical data this year for our three most advanced programs, our three clinical programs, SGR-1505, SGR-2921, and SGR-3515. In particular, SGR-1505, our MALT1 inhibitor, we're going to be presenting data in Q2 of this year. We also are pretty excited about the fact that we continue to advance this platform. We have quite a significant effort in this area. We've announced an initiative which is funded by the Gates Foundation with a rather large grant, nearly $20 million grant to advance technologies for predicting toxicity of molecules associated with binding to off-targets.
We've also announced the release of a biologics platform that's built on our small molecule informatics platform called LiveDesign. We've released a version of LiveDesign that supports biologics. We're very excited about the third thing that's listed there, the machine learned force fields, that's allowing us to continue to enhance the accuracy of the physics-based methods that we talked about earlier. In particular, there's some very exciting applications in battery design with regard to this machine learned force field that's allowing us to simulate chemistry battery at the electrode-electrolyte interface. I'll just leave us on this slide. I think I've gone long enough. We should get to the Q&A. I wanted to give a sense of how we're advancing the platform in all of these different areas.
As much progress as we've made, and it's been pretty extraordinary over the last 35 years, we continue to advance the platform as a very important part of the business. We're the innovator in this space. You can see here, everywhere from target validation to hit discovery to lead optimization, even to preclinical development, we're advancing technologies that are having quite a big impact, not only for our customers, not only our collaborators, but of course, on our proprietary programs. I think with that, we should probably launch into the Q&A.
Thanks, Ramy. That was super helpful. I think for investors that are less familiar with your story, I think that was a great sort of comprehensive overview. Just as a reminder, investors, if you'd like to submit a question for me to ask, there's a link at the bottom. I'll see it on my end here, and I'll be able to ask Ramy the question. I guess I'll kick things off here, Ramy. Maybe talk about end markets and your customers. What are you seeing in the market currently, both from small biotechs all the way up to large pharma? Obviously, you've put really healthy guidance out both on the software and drug discovery side, but kind of walk us through the customer segmentation. I think we get that question a lot, especially on the biotech side and the potential recovery there.
Sure. Yeah, first, with regard to large customers, I think you know this, that we, of course, all large pharma companies are using our software. That's great and have been for a long time. There is a pretty big difference between the amount of software and therefore the spend that some of our largest customers among, let's say, the top 20 or top 30 pharma companies, there is a very large disparity. That, of course, is a huge opportunity. I think you saw in our KPIs, we increased the number of customers spending over $5 million in 2024 from four, sorry, in 2023 from four- eight in 2024. That's a pretty big increase, but that's only eight. There are quite a few that are still below that.
We see a very nice trend of large companies significantly increasing their usage of the software because that's really where you see an impact. The more molecules you can explore computationally, which requires more licenses, which requires more spend, the bigger the impact and the more likely you are to actually identify a molecule that has all the properties it requires. That's on the large company side. Oh, I should mention one other thing. The spend on our software relative to their R&D spend for our largest customers is around 0.1%-0.15%. It's still a small fraction. I think that's an indication of the significant opportunity on the large company side. On the small company side, this is important. Yes, there are challenges, of course, that are happening. Among our largest customers are biotech companies. They're scaling up their usage as well.
They're spending on our software a larger fraction of their R&D spend, which goes back to the earlier point about the opportunity. Yeah. We still see growth among the small companies. Here's the reason. This is kind of important. I hope this came across in the presentation. The application of our technology, especially as you scale it up, is improving efficiency. It's reducing R&D costs. It's not increasing it.
You are saying that as a smaller biotech in pharma. Yeah.
Yeah, yeah, yeah. I think that's a pretty important point. In an environment where there are these sort of challenges, macro challenges with regard to R&D budgets, that actually results in an increase in the demand for the software because not only does it just accelerate discovery programs, but it improves them. We've talked a lot about this. It increases the probability of success, which also reduces the cost. If you keep having failure after failure, of course, those costs accumulate.
That's super helpful. I guess maybe just breaking down your two revenue streams on the software side, your guidance of 10%-15% for this year. Maybe just walk us through the assumptions on that. For investors that are unfamiliar with the Schrödinger story, traditionally, historically, you've done multi-year on-prem software agreements, which tend obviously to hit in terms of revenue recognition all up front or a large portion up front. We're moving towards sort of annual cloud-based streams. Maybe talk about the moving parts there, what's embedded on the low end of the guidance, what's embedded on the high end of the guidance.
Yeah. As you mentioned, and I mentioned too, we're guiding to 10%-15%. We're very confident in that range. It's broad-based. There are many ways of achieving that. Let me now address your specific question about on-prem and hosted. It's true that the majority of our revenue is on-prem, which means that most of the revenue from the deal is recognized in the quarter that it closes. Because our contracts are tied to budgets and pharma companies, and their budgets are typically tied to the fiscal, to the calendar year, we have a rather large Q4. I think if you look at our quarterly revenue, you can see that it's quite large in Q4, smaller in Q2 and Q3 because there just aren't that many renewals there.
If you have most of the revenue being recognized in the quarter of the deals closed, you will tend to have larger Q4s. You also have larger Q1s, and then Q2 and Q3 are lower. We have explained that over and over again, but I understand that it is a little disturbing and a little bit hard to follow because of revenue sort of going up and down. We actually present a trailing Q4, 12-month trailing average of the quarter, and you can see that smooths it out. Another way to smooth it out in a GAAP way is this, as you said, this sort of transition to hosted. I think it is very, very important for everybody to understand that when we say hosted, it is not the traditional sense hosted.
That is, we're not all of a sudden hosting all of our customers' software and thereby sort of increasing the cost to ours. All that's being hosted is what's called the license server, which just monitors the licenses that are being used, not the actual underlying code. That's still on-prem, so to speak. It doesn't matter. If you're hosting any part of the software, the whole entire thing is considered hosted. Okay. Now, what does that mean? That means that now, of course, the revenue is recognized radically over the year, and that tends to smooth it out. You saw that we've increased the percentage of revenue that's hosted in 2023 from 13%- 20%. It's increasing. We're doing it slowly because you can imagine if you do it too fast, you create another problem, which is in the year that you're undergoing the transition.
A number of well-known companies have done this. The revenue drops. We are doing something pretty incredible, which is we are shifting the revenue to hosted, but we are doing it in a sort of controlled way where we continue to see growth in the business while we transition to hosted. That will have a tendency, of course, to smooth out the revenue over time, but it is a gradual process.
That's helpful. Maybe provide some color there, Ramy. How are those contracts or how are the negotiations or behavior of customers in that progression? How is that all? What are the logistics behind it? Are you pushing it? Are they sort of requesting it? Maybe walk through those customer dynamics.
Yeah. To the customer and really to us, the two are nearly identical, as I was sort of hinting at. We're just hosting the license server. The experience for the customer is identical. The only difference is that it removes a little bit of burden on the first day that the licenses start up from their IT departments because they don't have to spin up the license server. We just take care of it. When we approach a customer to say, "Hey, maybe we can help you with this, and this might help you because now we're going to understand what licenses you're using, and we can help you make sure you have the right licenses." That's sort of helpful. We've had very little pushback when we talk to customers about that transition. As I said, we're undergoing that transition sort of slowly.
It is essentially under our control. When we ask a customer to switch to hosted, generally, that is well received. That is more the direction that we are doing it. It is not like there is sort of demand coming from customers. Again, they do not really know the difference except for maybe a few hours of IT work at the beginning of the contract.
Great. I guess I'm going to shift the conversation over to, I think, the most important sort of milestones that you guys have seen recently. These are what we call combination deals, these large collaborations with Novartis and Eli Lilly. Can we talk about that more? What that means for your proof of concept for your platform, really, and where you expect-do you expect more of these large combinations with large pharma over time?
Yeah. That's a great question. First of all, let me explain. One of the big challenges that we have, I hope that this is something that's coming through, is that the application of our platform at scale is a completely new way of doing drug discovery versus sort of traditional trial and error, essentially, relying solely on experiment. We know how productive that is, not terribly. We know about all the failure rates. That is, as is the case in every other field that's been transformed by computer-aided design, there's a period of disruption, and there's a period where nobody knows how to really undergo that change. You do not have the expertise to do it. You do not have the culture. Again, you can look at other industries, even animated movies and airplane design and all these industries that have been completely transformed by CAD.
If you look at the history of those companies, you'll see the same thing, nothing different here. What does that mean? That means that there's a challenge in actually training companies to use the software at this scale. There are two ways of doing that. One is just training. We do that all the time. We have a very, really incredibly talented education team at Schrödinger that does workshops and all sorts of training sessions. We have online courses, which are very, very well attended. There is another very effective way to do it, which is just work together with a company on a project where they can see they have a front row seat, as we like to say, to the collaboration. They can see how we're doing it directly.
They can see what it means to run a very provisioned 100,000 nodes on the cloud or GPUs on the cloud and run things at this sort of large scale. When they see that firsthand, it's a very, very effective way of that knowledge transfer that's required to get companies to adopt the technology at a large scale. We really like these combined collaborations. It results in not only a very exciting and important drug discovery collaboration. You can see that we talked about the revenue from those and the value that's been generated from those has been quite significant, not only from revenue, but also from even the equity stakes we have in the companies we've co-founded and so on.
At the same time, it's also a very effective way of transferring the knowledge, which is resulting in the companies really adopting the technology on a very large scale. I think you asked, are we going to continue to do these? Yes, absolutely.
Moving on here, maybe talk about the you had a slide on your new platform. I want to drill in more on the predictive toxicology platform.
Might as well go back to the slide, right?
Yeah, let's go back to it.
Right here.
Yeah, there you go. I think when was this introduced last year, Ramy? Is it fully?
The grant was announced last year. We've been working on that. That's when we started to work on the project. It hasn't been released, but we can talk about that. Please go ahead and ask the.
Yeah. So I mean, I guess what excites you most about this platform? Have clients sort of been, was there a need from clients that precipitated this? Maybe just talk about the sort of origins of this platform and why it's so exciting.
Yeah, yeah, most definitely. It wasn't just from our customers, but it was also from our own drug discovery collaborations and our own proprietary programs. In fact, every single project in the whole industry encounters off-target toxicity. That is, the molecule that you're developing and trying to improve its affinity to the target that you're going after, there are always, without exception, other proteins that are either through high-sequence homology or just by chance that have similar pockets that the molecule also binds to, and that almost always causes some kind of undesired toxicity. Yes, it's a very significant need. We've known about this for a long time, but the technology just wasn't at the right state to even think about a project like this. As things go, we kept advancing the technology. We kept improving this platform of being able to combine physics-based methods with machine learning.
We got to a point where we said, "Wow, we can actually attack this incredibly hard problem." Through funding from the Gates Foundation, we really ramped up the effort. What we've done so far is we've enabled, and by enabled, I'll explain what enabled means, but we now have a virtual toxicity panel of around 50 targets where we can say, "Give us your molecule, put it into this, send it over to this virtual assay, virtual toxicity, off-target toxicity assay, and predict, compute whether the molecule binds to any of these 50 targets." It's a diverse set of targets. We've tested this. We've actually started to, we published on a few of them, but we'll be publishing soon. We'll be releasing that product later this year.
So far, the results are impressive, where we can, earlier in a project now, much more efficiently, of course, relative to sending the molecule out for experimental assays where you actually test it against each of the proteins, we can do it computationally. More importantly, and this is really critical, not only do we get a sort of a panel of yes/no's, right? You hit this target, you didn't hit this target. The ones that we hit, you actually can do something about it. You know how to because you have a model. You have actually a physics-based model that tells you essentially how to fix the molecule, adjust it, change its chemical structure to not hit that target. There is a lot of excitement around this project. We think the demand for it's going to be pretty significant.
We're continuing to advance it, obviously, and add more and more targets to it.
That's amazing. You said 50, right? You have 50 targets right now?
Around 50 at the moment. That's right. Yeah.
While we're on this page, let's talk about the next two, I think, more exciting releases here, maybe on the biologics optimization and protein engineering first. Maybe we'll start there. Where are we in that development, and what excites you about that?
Yeah. Biologics design, there are sort of two areas of opportunity to advance the technology. One is in the underlying science. We've developed physics-based methods that allow us to predict binding affinity of an antibody to an antigen. That can be used for affinity maturation. It can be used also to optimize affinity as a function of pH. We can also compute thermal stability. These physics-based methods are allowing us to actually design, or our customers to design, better antibodies. There is another challenge, another opportunity, which is managing the enormous amounts of data that get produced by the experimental methods that produce antibodies. That requires an informatics platform that can store all that information and deliver it to the decision-maker, the people that are making decisions about what the challenges are with the antibody that they're working with and how to improve it.
We have extended our small molecule informatics platform to support biologics. That was released last year. We are seeing there is very significant demand for this because there are not any good solutions now. People are actually using things like Excel to do that. That is crazy. Or their own homegrown things that, of course, are very hard to maintain. There appears to be a real pent-up demand for a way to manage this massive amounts of data that get produced by the experimental method. It is called LiveDesign Biologics. Again, that is something that we think is going to result in people really adopting this platform, as they have done very nicely with the small molecule platform, LiveDesign itself.
Great. Maybe let's shift over to your pipeline. I guess walk us through the molecules, the therapeutic areas. I know you kind of did this a little bit initially, but maybe dig deeper when the clinical readouts and preclinical data comes out for each one. Really, I guess, Ramy, in your opinion, what excites you the most on the internal pipeline side?
Yeah, yeah, yeah. We're excited about all of them. This is the three, and I mentioned them earlier, oncology programs that are all in the clinic, and we've said that we will be releasing at medical meetings for the first time clinical data. We're obviously very excited about SGR-1505, our MALT1 inhibitor, and B-cell malignancies. That's because that's the one that's coming up. That's the nearest term one in Q2. We're also excited about SGR-2921, CDC7 inhibitor for in AML. We'll be presenting data later this year in the second half of the year. Of course, sorry, SGR-3515, our Wee1/Myt1 inhibitor in solid tumors. Again, presenting data at a medical conference later this year. I know I didn't pick our favorite child, but I hope it's clear. I hope it's clear there's reason to be excited about all three of them.
Yeah. Maybe walk through the competitive environment around the MALT1 molecule.
Yeah. Look, I think everybody understands that J&J demonstrated in their release of data that they saw responses, and the target is obviously validated, but they also saw toxicity. That was the thesis from the beginning. I think you can see that from the platform. Can we design a better molecule with a larger therapeutic window? We, in the preclinical data that we released, demonstrated that we designed using the platform a molecule that's 50 times more potent. Of course, in a situation where you have higher potency, obviously, in principle, you should be able to dose lower and avoid certain toxicity. That's something we're going to have to demonstrate. That's what we're hoping to show: preliminary safety data, PKPD as well, and even preliminary efficacy. Yeah, that's where we are with regard to SGR-1505.
Great. I guess maybe, Ramy, the last couple of questions here. It's more about strategy. How do you think about balancing your internal programs versus collaborative programs? Maybe I'll start there first.
Yeah. No, that's a really good question. You can see we have a whole range. Obviously, even the software licensing, that's one part of it. With regard to the internal versus proprietary, sorry, the collaborative versus internal/proprietary, we like this. This is what we have. We have a balanced model where we're balancing sort of risk-reward profile of the programs. We really like that. We have some programs where maybe in the form of partnerships we have with companies we co-founded where we own some small fraction, single-digit fraction of the company. The risk is relatively low in those cases. That's generating, well, it's not revenue. It's generating income every once in a while from the equity stakes we have. That's been quite successful: Nimbus, Morphic, Relay, Structure, and so on. That's great. Okay, that's one form.
We have another type that's maybe a little bit more risk and a little bit more reward, which is the collaborations with pharma companies where we're doing a little bit more of the work, where we're not only doing the design of the molecules using computation, but we're doing the chemistry, some of the biology, some of the biochemistry, biophysics, and so on. Those generate revenue in the form of upfront, in the form of milestones. In the future, we believe in the form of royalties on sales. We have quite a few programs where we have royalties on sales. That's another sort of form of collaboration. Of course, the ultimate where we take on more risk, but of course, the rewards are higher where we own 100% of the programs.
We like that sort of balance, and we will continue to have that balance. We think the collaborations are important not only for the reasons we just gave, which is the revenues generated from, but remember what we talked about with those combination programs. Working with pharma, working with the customers is pretty important. Those collaborations are important. You can tell we're pretty excited about the opportunities of owning a much larger portion at the moment, 100% of programs that we're taking on ourselves. Yeah.
Okay. That leads me to my sort of last question here as we're almost up on time here. You're excited about your internal pipeline. Walk us through the strategy there of monetization optionality, right? Do you bring these things to phase two? I know it's going to be different for each internal asset and each molecule, but kind of walk us through your strategy and how you think of monetization of your internal pipeline.
Yeah. Obviously, it's very important to be guided by the science and to be guided by what's the best thing for the asset. In a situation where, for example, combinations are critical, that's pretty challenging for a relatively small company to do on their own. Those programs make sense to partner. It is our intent. I think you can see from the programs that we've talked about that these are definitely programs where we're expecting to see efficacy in monotherapy. The real opportunity is in combination with other agents. In that situation, it makes sense to partner these programs. That's our intent generally with these programs. Now, that doesn't mean that we will never, ever take a program further into phase two.
For the right kind of program, the right profile, the right target, the right therapeutic area, that could make sense in the future. This is our plan with these programs, is to partner them. Yeah.
Great. I think that brings us up close to the time here. Thank you so much, Ramy, for this fireside chat. Thank you, everyone, for joining us and kicking off our virtual healthcare conference here.
Thanks a lot, Scott.