Hello, everyone. Welcome to our first in quite a few years in-person Global Biopharma Conference here in Miami. I am Mani Foroohar, Senior Analyst covering biotech medicines. I have the good fortune of hosting Schrödinger, a bellwether in the tech-enabled drug discovery space. Though I suspect if I say the words AI, at least someone on this panel is going to throw me off of this platform.
Fighting words.
Geoff, Karen, it's so good to see you guys. I know you guys have a few slides up.
Yeah.
Geoff, did you wanna walk us through those before we dive in?
Yeah, let me just jump up, and go through the slides quickly. Thanks, thanks, Mani. Great to be here, great to see, the Leerink Conference back. I'm in favor. I'd say it was 35 degrees Fahrenheit and 20-mile-an-hour winds when I left LaGuardia this morning. It's much nicer here. So maybe next year we can go to the Florida Keys or something. The trend is in the right direction. So, I'm the CFO at Schrödinger. I'm going to give you a quick overview of the company. Hopefully, I'll be able to translate some of the jargon that sometimes surrounds the computational drug discovery space, and you'll get a better understanding of what we do. Okay, we will make some forward-looking statements. Please, refer to our SEC filings for full disclosure of the risks associated with our stock.
So this, people often say to me: What is it that Schrödinger does? And there's been a 30-year investment in developing a computational technology for modeling the interaction of ligands with proteins. And this is a huge undertaking because it requires very high levels of computation, and it's built originally on first principles of physics, encoded in mathematical formulas that then are commercialized effectively through software that we then license to companies across the industry. And this graphic really kind of shows you the heart of what we're doing. We're modeling the protein itself, which is represented here by the wide bands, and then we're modeling alternative small molecule ligands that will bind to that protein with different properties, different affinities.
And then beyond the affinity, we will look also at a host of other drug-like qualities that I'll talk about, that we try to optimize in those small molecule constructs. So, I'll explain a little bit about how machine learning or AI fits into our technology, but fundamentally, we're about modeling these interactions using very high-powered computation to do that, and then enabling our customers who purchase our software to do that themselves on their own cloud. We have not just applied this in drug discovery, we also have a growing business in materials. You can imagine some of the same predictions that we apply to things like identifying novel binders to, for example, TYK2, could be applied to predicting properties of materials.
For example, biofilms for aerospace applications, consumer packaging materials that have more favorable environmental properties, a host of different industries where wherever there's a focus on innovative molecular design, we're finding great places to introduce our software. So, I mentioned that our technology was built on physics and mathematics and first principles, but it is definitely accelerated by the deployment of AI. So we estimate that, for example, doing one calculation, hypothetically, on one ligand to bind to one target might take a day. So if you wanted to do 1,000 hypothetical molecules, it would take 1,000 days. But of course, if we use AI, we can dramatically accelerate that process by creating a training set from that 1,000, and then using that to identify molecules with more favorable characteristics from billions of hypothetical molecules.
So we definitely have embedded AI at several places in our technology stack, and that's available to our customers, and we use AI at a significant scale internally for our drug discovery efforts. But it's tightly linked with the physics-based methods that underlie the core technology. But the goal, of course, is the same goal that everybody has, which is identifying the very best molecule. So the cloud out to the, to your left represents the entire universe of potential small molecule candidates against a particular target. And we're routinely now examining billions of potential ligand structures against a particular target. And then using these methods, we can relatively quickly get to a small set of molecules for synthesis and then actual physical testing.
In our case, that universe of molecules that we synthesize is usually substantially less than 1,000, whereas conventionally in the industry, it's many thousands, and each of those compounds, of course, cost thousands of dollars themselves to be synthesizing. So what we envisage is that we can get to molecules faster, we can get to them at lower cost, but most importantly, we can identify molecules that have better properties, and we're now able to optimize for all of these different attributes, not just binding affinity or potency, but selectivity, solubility, bioavailability, even a number of off-target toxicity liabilities we can actually computationally model and predict about a molecule well before we actually synthesize it and test it... So we make money out of this platform in three ways. First, historically, we've been in the software business.
We have close to 2,000 customers spending over $1,000 a year. I'll give you a little bit, few more parameters. Every company that we know of in the industry engaged in small molecule drug discovery is a customer. Now, they're not all scale customers, but you can see, you'll see in some of the KPIs a number of the largest companies in the industry are spending $ millions on the technology, and we envisage the rest of the industry also ramping up into that range. So software business is really foundational for the company.
The second thing we're engaged in is collaborations, where our software partners, who have been impressed by the technology but also perhaps a little intimidated by its complexity, come to us and say, "We have a really high-value target that we would like to work with you on. You have better experience with the technology, more expertise with it than we do. Can we form a collaboration?" In return for that, we get downstream participation in the molecule: royalties, milestones, et cetera, and in some cases, equity. And then lastly, and most recently, we're developing a proprietary portfolio of our own, where we see opportunities that are just too tempting, and we can deploy the technology ourselves, use our balance sheet, invest in the development of novel molecules against those targets, and Karen will give you a bit more information about them.
Just some KPIs on the software business. Last year, we reported revenue of $159 million, growing at 17.5%. We had 54 customers who were spending over $500,000, 27 over $1 million a year, 4 customers over $5 million a year. The ACV, which is the annual contract value of the top 5 customers, was 6.7. So you'll immediately infer that there are a number of customers significantly above that $6.7 million in annual software spend. So if you think about the lifetime value of that customer, that actually becomes a significant source of value for us, and we have more and more customers moving into those top tiers.
Still, however, there is a huge spread across the industry between the level of use of the most advanced companies in the industry. We think of them as advanced 'cause they spend a lot of money on our technology, but computationally advanced companies and their least advanced rivals, who are equally large global pharma companies. That's a 20x difference. We're frequently having conversations with global companies who are saying, "Yeah, we're, we really believe in Schrödinger technology. We're using it a lot." And then we sort of say to them, "Well, actually, you're using it more or less than a specialty biotech company." We have a number of biotech companies who have committed to the technology, who are in the $ millions in terms of their annual use on just one or two programs. So our benchmark for the scale of the opportunity is that sort of range per program, to come up with a truly novel molecule. So this is the collaboration. Maybe I'll hand over to Karen. You want to jump in and talk about collaborations in pipeline?
That was supposed to happen. Okay, I think this is working. You can hear me, right? So, as Geoff said, we actually deploy the platform not just in terms of the software sales, but we've also, over the last 15, maybe even 17 years, been collaborating with a number of companies. In fact, very early on, Schrödinger co-founded Nimbus, we're co-founders of Morphic, and in addition to a number of biotechs that we've worked with on their pipelines from the very beginning using this platform, we also have a number of collaborations with large pharma. One might ask: Why are they collaborating with us if they have access to the platform?
The reason for that is that deploying this is complicated, and while they're buying the software, they also want what we call front row seat on how we actually use the platform. Many of these programs are now advanced. They've exited discovery, and they're in the clinic. I think the most famous, perhaps over the last couple of years, has been the Nimbus TYK2 program, which was a very significant transaction. In addition, we have programs that are in the approved status, working with Agios actually very early on when the platform was brand new, but a number of collaborations still ongoing in the discovery space. In addition to collaborations, as Geoff said, over the last few years, we've decided to now deploy the platform internally. We've built a group, the Schrödinger Therapeutics Group, which is deploying the technology at full scale.
We have a number of programs now in the clinic. Our MALT1 inhibitor is in phase 1. We completed a healthy volunteer study, and we're in an oncology trial in B-cell malignancies. We have a CDC7 inhibitor for AML and MDS that's also in the clinic, and this year we expect to put our WEE1/Myt1 synthetic lethal program into phase 1 as well. And behind that are a number of other targets that we're working on, using both the platform and our new structural biology capability, where we have first-in-the-world structures for some of these targets. And there are more programs behind those that we haven't disclosed yet.
You've got another slide.
Okay. Oh, right! So, as I said, the 1505 program, we have completed a healthy volunteer study. This was in over 70 subjects, and we're very pleased with the profile of the drug so far. And, we're now in this malignant, B-cell malignancy study, with data expected, late 2024, 2025. As I described, our CDC7 inhibitor has a very interesting translational package in AML, where we're actually showing activity in all backgrounds of AML, regardless of genetic mutation, which we think is a very exciting opportunity. And again, here, this phase one study, we expect to be reporting information from the clinical trial in 2024 and 2025. And then we are filing our IND for the WEE1/ Myt1 program. We expect that to happen in the first half of this year. As everyone knows, WEE1 is a validated target from a clinical POC perspective. In terms of targeting specific patient populations, this synthetic lethal pairing with Myt1, we think is very interesting, and we think will open up the opportunity for WEE1 more broadly and in a more precise fashion in these patients. Okay, so back to Geoff.
Go.
I think I've broken this.
Thanks, Karen. I'll quickly go through the numbers. We reported Q4 last week. Total revenue grew by 30%, $74 million. Software grew nicely to a record quarter for us, of $69 million in the Q4. Drug discovery went from $9 million down to $5.5 million. The $9 million, there were a number of milestones that were non-recurring between the Q4 of 2022 and the Q4 of 2023. Operating expenses did increase, principally R&D, and as a result of that, our cash burn was slightly higher for the Q4 compared to the prior year, and our cash position at the end of the year was $469 million.
For the full year, revenue grew approximately 20%, 17.5% growth in software and 27% in drug discovery. The gross margin is hovering around 80% for the software business. Operating expenses did increase, as I mentioned, principally due to R&D. We did report a positive GAAP net income. That was due to the distribution of $147 million that came to us from Nimbus in the first part of the year. And this is just a sort of longer-term trend on our business. You can see there's a nice CAGR in both the software and the drug discovery business, and it's probably hard to see from the back of the room, but in addition to this, we are creating substantial value through our equity positions.
The distributions over the last five years from our ownership stakes in the companies that we've co-founded is around $180 million. So we're sort of building value both through the revenue, but also through the balance sheet. Our financial guidance for 2024, initial guidance at the first of the year, we expect software revenue growth to be in the range of 6%-13%. Drug discovery, we're forecasting revenue of $30 million-$35 million. We expect our gross margin to be similar to last year, and then our operating expense growth, we expect to be 8%-12%, compared to 28% last year. Our cash use, of course, will be slightly higher than last year. So just to summarize the milestones, Karen's alluded to all of these development milestones.
We're really excited about the progress we're making with the portfolio. We'll also continue to invest in the platform. We think that there will be significant upgrades to the force field, which is the underlying model that we use to calculate the pattern of charge across target molecules and then calculate the binding affinity. We have a significant research collaboration with the Gates Foundation to design novel materials for batteries. We think we'll be able to publish some of that research in the next 12 months as well. So, Mani, a little bit of time for Q&A.
Geoff, we'll start with you. On the last earnings call, there were some questions around, you know, the nuances and variability around guidance. Some of that's unavoidable, because the underlying business model is not as linear as a traditional SaaS business. Walk us through a little bit of kind of the basis of that guidance, how it takes into account sort of the lumpiness of the sort of underlying business for the software, the software component.
Yeah, this, this is working. Great question, and certainly the, the number one question coming out of the earnings call. I think it's important to go back and look at 2023. We indicated through 2023 that we were engaged in multiple $ multi-million-dollar, multi-year renewals. And for most of our customers who have so-called on-prem licenses, meaning we distribute the license, effectively, the license server is on their premises, we report between 65%-85% or 90% of the revenue for the full period of the contract in the period in which we sign the contract. So if we close a multi-year, $ multi-million-dollar contract, and it's a three-year contract, then we're effectively recognizing two whole years of revenue in that quarter.
So those contracts were coming to fruition through 2023, and we closed a number of them successfully, and we're really happy about that, but of course, it caused that terrific revenue result we had in the Q4. Now, in 2024, we have to grow beyond that surge of revenue, which was significant. That was revenue from 2024 and 2025 that came in in the Q4 of 2023 because of the multi-year contribution. So what we've said is the revenue guidance is 6%-13%. The operational growth of the business, mathematically and structurally, is gonna be substantially higher than that, which is the ACV growth. So what we expect to happen is the annual contract value will be growing more like a significantly higher number than the revenue guide.
But because we're having to catch up on that Q4 revenue bolus, the revenue result will be lower than the ACV growth. So that's just a little bit about the sort of structural things that we're dealing with. In terms of opportunity, though, we, many of the conversations that we started last year, we still have opportunities to continue them, so there are still a host of other global companies that we can close a contract with.
So I think some of the, I'm gonna put my hedge fund hat back on, and I think one of the questions you get back, you-- that we hear, I presume you're hearing, well, approximately 15 times as often, is if you're recognizing revenue across a couple year period, disproportionately, in a linear sense, in the, in the period during which it's signed, should you not, as ACV grinds up, be a beneficiary of that dynamic again this year? I.e., given what you talked about, the depth of un-- the universe of unpenetrated or under-penetrated contracts, shouldn't there be novel, multi-year, multi-million dollar contracts popping up this year? And if that's true, is there a seasonality? Should we expect that in 4Q again? Like, to what extent does the guidance not capture one-off events to the upside, but does compare versus one-off events that make a tougher comp last year?
Absolutely, we have opportunities for new multi-year contracts this year. That's certainly not assumed that we will close those deals this year in that guidance, but we have those opportunities. It does come down... What we're gradually seeing is a shift over to what's called hosted, and hosted is where we actually host the license server, and we can recognize the revenue ratably over the period of the contract. So I did highlight in the prepared remarks of the conference call, that a couple of customers switched to hosted, which takes revenue away, but dials it in in the future. So estimating what revenue is going to actually be in 2024, to a certain extent, depends on the nature of the renewals that we get. If we have customers showing up, for example, like Lilly did, and saying, "We're a long-standing customer, we're already spending millions of dollars on software," and they renew, then yes, we will get that bolus, but we're not counting on that for this year.
Okay. So from there, let's talk a little about the underlying economics for hosted versus on-prem, which when most biotech investors hear those things, they make certain assumptions about where physical-
Yes.
Where physical infrastructure is and what that means for OpEx and CapEx. Can you carve out to the extent there are and are not differences in sort of the financial implications of a dollar of on-prem versus a dollar of hosted?
It makes absolutely no difference. All we're talking about is the hosting of the license server, and that's in effect... Imagine the toll booth on the Jersey Turnpike, right? It's the place where we keep track of how much of our software the customers are using because we sell a token-based license. So a customer who has a 1,000 token license is paying us 10 times what a customer with a 100 token license would pay us. So we need to keep track of the number of tokens, which in effect is the number of gates going over the turnpike. So what the variance for the economics for us is simply in that license server hosting, not the actual use of the technology, which they're running on their cloud anyway. So tiny little difference. In fact, I've looked at it, maybe there's half a point of gross margin difference between a, in the lifetime of a contract between a hosted and on-prem.
Okay. And then moving up from the gross margin point through revenue, almost to TAM or backlog.
Yep.
One of the comments you made, and you've made versions of this comment in the past, around every single counterparty client customer that is doing small molecule development is a customer, but not necessarily a major customer. That captured a little bit of the opportunity in the customer lifetime value growth.
Yep.
Talk to us a little about what you're seeing in end markets in terms of growth in the creation of new customers, and to what extent do you guys have leverage to expansion and loosening of venture capital funding and new company creation, and how is that captured or not?
Yeah.
In your guidance and potential? How do you talk about long-term growth?
Yeah, this is a really important point. So we did have some turnover in our smallest customers in 2023. That was increased compared to 2022 and 2021. And that turnover, this was in the customers between 100,000 and 500,000 ACV, which is where a typical small biotech will sit. And that turnover was because customers lost funding, you know, stopped doing discovery and were just doing development, were acquired, a multitude of things. But the net of it was more of those customers went away than showed up, and so our retention went from 96% to 92%, and all of that drop was in that smaller customer group.
Now, we are seeing a lot of inquiries from small companies who are looking at what we're doing on the venture side and saying: "Is there a way to do something creative with you taking equity positions?" Our preference would be to get cash, all things being equal, but we've clearly created a huge amount of value from the companies we've co-founded. Our total cost for that $180 million distribution is probably one-tenth of that. So we're contemplating a number of those types of opportunities. I think you'll see more of that kind of filtering through the bottom end. And I don't think that we'll see another big leg down in our customers this year, but we're not counting on a recovery to meet that guidance.
Okay, here, I'm gonna hop over. Let's talk a little about WEE1. You talked about the opportunity to expand the universe of patients that could benefit. Talk a little bit about what data points you've seen internally and what your team has seen that lead you to think about that opportunity, and then beyond that, how that data might play out in the clinic to sort of prove out that you're expanding the application of this approach.
Yeah, great question. So as you know, DDR targets have been difficult to essentially figure out who is going to be sensitive to these mechanisms. So one of the things we did in our collaboration with MD Anderson was to really focus on identifying sensitive tumor types. That's one component of the expansion of the opportunity. Those sensitive tumor types, some of them have been published previously. We know that ovarian, for example, is highly sensitive to WEE1 inhibition. You're seeing monotherapy activity, uterine serous carcinoma also, but we found tumor types beyond those in the MD Anderson PDX panels that we think are going to be really interesting opportunities, and some of these are larger indications, actually, where we think we've identified sensitive populations. Secondly, one of the limitations of WEE1 has been this narrow therapeutic index.
People actually not staying on drug. We know the drug works, but if you don't stay on it, obviously, you're not going to benefit. So one of the things we've done is design a molecule which we believe allows one to hit the target hard, drive those cells into a catastrophic apoptosis, and then back off, leave those patients alone for an extended period of time so that their heme compartment can recover, where you've driving efficacy in that short period of three days of dosing. This is what we see preclinically, and we've published this last year at AACR.
You can dose for 3 days, get out of there for 11 days, and you see exactly the same efficacy, and we think that the properties of our molecule, this residence time in the tumor, the exquisite potency, this is a picomolar drug, that allows us, we believe, to open up the therapeutic index, which is, we hope, going to actually benefit more patients. Now, in the first-in-human study and the dose escalation, we'll be looking very closely at these types of parameters: safety, time on target, target engagement, which we hope will then correlate with what we've seen preclinically and open up the opportunity for this exciting mechanism. And then thirdly, our compound hits Myt1, which is the synthetic lethality opportunity, which again, in Myt1 expressing tumors, we think this is an opportunity to actually go after those as well.
So let's talk a little bit about exactly that nuance, the interaction between developing a molecule that has differentiated characteristics, in this case, on therapeutic window, and taking a differentiated approach to dosing, clinical trial structure, et cetera, i.e., sort of reducing it to practice m aking it work in sort of the crassest and most simple terms. To what extent is that is that a obvious signal internally from the experience you have with your technical platform, where you can tell very quickly, if I see this type of profile coming out of the platform, I know that I can make these changes to dosing? And to what extent is it still kind of as bespoke and as analog as all drug development on some level is?
Yeah. I, I actually like your question because I think it ties back to the very beginnings of a drug discovery project. Often, when people are going after a target, they're sort of locked in by the chemical matter they first identify, and most companies are in a hurry to kinda get to that point where they have a development candidate. We actually had a very broad set of completely unique leads for WEE1, and we were able to actually screen those leads very early on for the very properties that I just described. We weren't locked into chemical space. I think you know that most of the WEE1 inhibitors are actually very, very closely related to AstraZeneca's molecule.
We were in totally novel chemical space, thanks to the platform, and we actually then profiled those molecules through these in vivo models that we've been generating over the last few years, that allowed us to only select the compounds that had these unique properties, not just on binding to the target, but on safety, at least in preclinical studies, you can look at some safety endpoints, and also on efficacy with these different doses and schedules. Now, doing those doses and schedules need to actually occur in the clinic, but we've written a protocol that we believe gives us a lot of flexibility to be able to answer those questions very quickly with pharmacodynamic endpoints to see if our preclinical actually matches the clinical profile.
Let's talk about what, again, the practical opportunities to reduce the practice. Because in a wholly owned program, you have tremendous flexibility as you continue to gather more insight into your candidate. Actually, candidates is a better way to think about it, your candidate cloud. Which is not necessarily true in a partnered program or collaborative program. To what extent are your partners, collaborators, positioned to take advantage of exactly this flexibility? I'm gonna stop before I turn this into a three-question question.
It depends which collaborators you're referring to. I mean, we have a variety of different collaboration types, but I would say, for the most part, when you collaborate with Schrödinger, you get this complete team that understands not just the biology of the target, but also this sort of breadth of chemical opportunities. And so if you think about our collaborations with Nimbus and Morphic and Ajax, going after some really difficult design challenges, right? So with Ajax, it's JAK selectivity. With Nimbus, it was getting to that exquisitely selective TYK2 compound. Our collaborators are fully in this with us, right?
They have access to all of the opportunity that this chemical space provides us, and we actually spend a lot of time actually honing that target product profile and the differentiation we want to see, not just in terms of the chemical structures, but how that translates, to your point, to meaningful differentiation. I think the proof points that we have from those compounds that have moved on into development would suggest that this approach we have of refining and honing that target product profile alongside the chemistry really works.
So as I, as I've slowly been breadcrumbing my way across this conversation, I wanna zoom out to a larger question, I think. Across the world of tech-enabled drug discovery, including the companies that are explicitly AI-driven, companies that use AI to complement a core tech, a core technology platform that has more established and durable benefit, as you guys have shown in terms of drug development. There's an open debate not just on can these technologies do anything, and specifically what they can and cannot do, but what the actual impact is, i.e., to what extent are you accelerating drug development, and to what extent are you moving probability of success up? And then to what extent are you allowing targets to be approached that could not be approached otherwise? When you think about these three separate sort of value propositions, where do you spend most of your time, and where do you see the most opportunity and value for your platform specifically?
Well, first of all, is there evidence of impact? I think was your first question. I would say that having 10 programs that have transitioned from discovery into the clinic, stay-- that stay in the clinic and are leading to these meaningful transactions, which is third-party endorsement from pharma, I think that that's strong evidence that getting these molecules through discovery and into the clinic and having them stay there and actually be the subject of very significant transactions is a sort of, you know, telltale of whether this platform works. But I would say that, you know, where we spend our time, as you can imagine, there's a lot of inbound interest in working with us.
Where we spend our time is on which design challenges can our platform really produce highly differentiated drug target, drug target compounds that marry with really significant unmet medical need. That's where certainly my team spend a lot of their time. How can we transcend what a traditional medicinal chemistry group, how can we transcend those unsolved challenges? I wanna point out that this is partly because of the chemistry platform, but in addition to that, the fact that we are now getting to the point where we have first-in-the-world protein structures for some of these intractable targets, I think is moving us into a really interesting space for the next decade. We were the first group to get a full-length structure of LRRK2. That's opened up brand-new insights that allow us, we think, to have a different approach to LRRK2. We also have the first-in-the-world hERG structure for the cardiac channel, right? Those are going to have, we think, very significant impacts on how we're deploying the platform for drug discovery from both an efficacy and safety point of view in the future.
So I think one of the other dynamics that comes up when talking about Schrödinger in terms of the universe of assets that are being developed on your platform, both your own and also your collaborative pipeline, which is not just those slides, but it's a vast universe of non-disclosed assets that live inside all your customers and just shows up as ACV on your, on your financial statements. But in terms of data, has obviously tremendous value. One of the questions that people ask is, to what extent is the not necessarily highest, but certainly most rapidly monetizable use of this technology to take a known and competitive target and make a uniquely differentiated molecule, i.e., you know, the pejoratively called a fast follower, but more honestly, just called a better drug. Like, to what extent is that the key driver for your partners in large, especially in large-cap pharma world, and to what extent is that actually sort of a idea of the company that's five years old?
I mean, I could start and say that, I think we see both interest in the moment a target is validated by some endpoint, whether that's genetics or whether it's clinical data or even preclinical data sometimes. There's a lot of interest in how can we both, be both best and almost first in class, right? The companies that sometimes validate a target often don't have the best-in-class molecule and may fall by the wayside. And so we get a lot of interest in-- from companies who want to be both first and best in class, moving fast and solving the design challenge. And then on the other hand, there's emerging biology coming out of brand-new spaces, like CRISPR functional screens, where people frankly wanna make the first drug against that target. And so we get all of that sort of interest coming at us, and it shows up in biotech.
Yeah
Collaborations, pharma collaborations, as well as in our own pipeline.
Great. And I'm getting the "Mani, you're running late" face.
Getting the hook.
Thank you. I'm sure we'll continue this conversation again soon. Great to have you guys.
Thanks, Mani.
Thanks.