Scroll Top

Ramy Farid on the Power of Computation in Drug Discovery

Episode Summary

Harry interviews Ramy Farid, president and CEO of Schrödinger Pharmaceuticals, about the company’s success using chemical simulation software to help drug makers zero in on promising drug candidates—and about its recent IPO, which brought in more than twice as much cash as the company expected.

Episode Notes

Schrödinger makes software that models the physics of atomic-scale interactions to predict the chemical properties of candidate drug molecules, helping its customers speed up computational drug discovery. A decade ago, Farid tells Harry, the company faced the chicken-and-egg challenge of convincing customers that its computational platform works, so that they would scale up their commitment so that they could gather evidence it was working. Close collaborations with customers like Nimbus Therapeutics helped it improve the software and surmount that challenge.

“In order to really take it to the next level and make a difference, it was necessary to use the software as customers ourselves,” Farid says. “You get real-time feedback, honest feedback. You can imagine how much we learned from that.”

Please rate and review The Harry Glorikian Show on Apple PodcastsHere’s how to do that from an iPhone, iPad, or iPod touch:

1. Open the Podcasts app on your iPhone, iPad, or Mac.

2. Navigate to The Harry Glorikian Show podcast. You can find it by searching for it or selecting it from your library. Just note that you’ll have to go to the series page which shows all the episodes, not just the page for a single episode.

3. Scroll down to find the subhead titled “Ratings & Reviews.”

4. Under one of the highlighted reviews, select “Write a Review.”

5. Next, select a star rating at the top — you have the option of choosing between one and five stars.

6. Using the text box at the top, write a title for your review. Then, in the lower text box, write your review. Your review can be up to 300 words long.

7. Once you’ve finished, select “Send” or “Save” in the top-right corner.

8. If you’ve never left a podcast review before, enter a nickname. Your nickname will be displayed next to any reviews you leave from here on out.

9. After selecting a nickname, tap OK. Your review may not be immediately visible.

That’s it! Thanks so much.

Transcript:

Harry Glorikian: Hello, I’m Harry Glorikian. And this is Moneyball Medicine. The show where we meet executives, entrepreneurs, physicians, and scientists using the power of data to reinvent healthcare from machine learning to genomics, to personalized medicine. We look at the biggest trends in patient care and healthcare management.

And we talk to people behind the trends to find out where data is making the biggest deal.

Imagine teaching a computer, the principles of physics. From the exact coordinates between every atom to how molecules behave in water. My next guest has had this experience and has been quoted saying it is mindbogglingly complicated. I can only imagine the company he runs was founded in 1990 long before the latest generation of startups, its approach simulates.

The physics of how compounds work rather than simply using deep learning to crunch existing data, looking for patterns where traditional drug discovery can be costly and time consuming with scientists, designing, synthesizing, and testing compounds. The company says its platform can predict the chemical properties of molecules speeding up the process of discovering promising compounds.

As they say, the proof is in the pudding. The platform was used in the discovery of two cancer drugs.  at, from Agios and in a pseudonym now marketed by Agios and Celgene that have both been approved by the FDA. He has been quoted by saying our platform has been validated again and again, across hundreds of targets in real world drug development projects.

He joined the company in 2002 as product manager and has advanced through positions of increasing responsibility and incredible leadership. Please welcome Rami Fareed. President chief executive officer of Schrodinger. Rami. Welcome to the show.

Ramy Farid: Thanks very much, Harry. I appreciate it.

Harry Glorikian: First. Congratulations on the IPO. I think I remember reading that the company raised what it was 232 million and the target was about a hundred million dollars. That’s correct. That’s that’s amazing. There must be a lot of excitement around the company and from investors to sort of overshoot that target.

Ramy Farid: I think that’s right. There is a lot of excitement around what we’ve been doing, which as you pointed out in the introduction we’ve been doing for 30 years, so it’s a long time coming. It’s taken a long time to explain to people what a, again, as you said in the introduction is rather common, but I think, uh, after, uh, after having done it for or beers and explaining it, I think we’ve gotten the story out there and it appears that people are interested.

Harry Glorikian: Yeah.  So, so, so for those people that, that don’t know Schrodinger, um, you know, 1990 is, is, is quite a long time ago. I, I F I feel like I was at the beginning of my career back then, but I, I can’t imagine the company was, was focused on a drug discovery funnel back then. Was it?

Ramy Farid: Um, not exactly on doing computational drug design, but of course we were focused on developing computational platform that would help our customers do drug discovery.

Harry Glorikian: I mean, you, you started as a product manager and now you’re, you’re leading the charge. Um, that must have been an interesting journey.

Ramy Farid: Yeah. Yeah. I think I can explain that it was, and I think what we discovered. About, I’d say between 10, 15 years ago or so when we started to make a really significant advances in the science, I think we realized that in order to really, as they say, take it to the next level and really.

Uh, make a difference. It was necessary to use the software ourselves, essentially as a customer ourselves, we were getting fantastic feedback from our software customers, but I’m sure you can imagine in any company that, that experiences the same kind of thing. There’s nothing like having people inside the company, using the software, you get real time feedback, honest feedback.

And so that was a really important part of the decision to, you know, approximately 10 years ago, really with Nimbus therapeutics being the first company that we were involved with. And, and you know, that, that, where we started to really use our own, our own software at scale, uh, you can imagine how, how much we learned from that.

Uh, and of course it had a big impact in the other direction as well, but that’s a fantastic way of validating the technology, you know, in this space, there’s a chicken and egg problem, you know, in order to have, in order to convince customers that the platform is really working, they have to really scale up.

The usage of it, they have to really transform the way they’re doing drug discovery. And in order to be convinced to do that, they have to see it working. So, you know, which one’s going to come first. So it was really, uh, uh, I think it worked really well for us to do that validation essentially internally, uh, where we didn’t have those issues.

Uh, and I think that revealed to the industry that, Hey, maybe this platform really finally is know computation. Finally can drive drug, drug discovery tools in a way that we had been dreaming about in 1990. It wasn’t really possible back then.

Harry Glorikian: Well, it sounds like you guys are. You know, I I’ve, I’ve looked at dozens of, of startups in this space and you almost sound like you need to be the case study, that they all need to take a class on to understand what they need to do, as opposed to going through the, you know, metamorphosis, uh, from scratch or, or, or learn it by braille, they could, they could sort of learn from what the, what you guys have gone through.

Ramy Farid: I think that’s right. And we’re seeing a lot of that. There are a lot of companies that are, are approaching us, that we’re in a similar situation to what we were in before we started doing computer aided drug discovery and asking us exactly what you’re describing. Sort of not, not just talking about the science, but even just, how do, how does, how does a software company.

Or let’s say a company that’s providing services start to use those services or software themselves. It’s complicated. Right. You have to establish certain firewalls to make sure you’re protecting everybody’s IP. There there’s. There are a lot of things that go beyond just the science. So I think that’s right.

Harry Glorikian: But it’s interesting that you say software. I mean, I was looking at your back on it and it, it it’s in chemistry. Right. So how did software. Somehow creep into this whole thing.

Ramy Farid: Yeah, it’s a great question. Um, that’s right. I’m going to experimentalists by, by training. And, um, when I was in my postdoc at Penn, I started to work with, uh, Les Dutton, who was my post-doc advisor.

Build a grotto. We were partnering with, uh, uh, on, on designing proteins from scratch to call Denovo proteins. And what we discovered really early on was that was very difficult for humans to do that and we needed to develop software. So when I took on a faculty position at Rutgers, um, where I also worked on the same sort of things or on Denovo protein design, I started to develop software.

And this is sort of interesting, uh, the core of, of the software that I started to develop them was actually something my brother had been developing in collaboration with somebody who happened to eventually join Schrodinger. So this was back that’s crazy. I know this was actually around the time when the company was founded.

So it was around 1991. Um, when I started my faculty position at Rutgers started to build a, a research group there and that’s where it started. And it’s funny because a lot of the principles that I was sort of becoming aware of at the time, uh, were, were relevant to what we’re doing now. Um, recognizing that pattern recognition.

And sort of traditional so-called QSR methods weren’t working and that you had to have a deeper understanding of the underlying physics associated with the things that you’re trying to, to predict, uh, in order to really get the right answer. So it started started way back then, then, then that’s actually how I met, uh, uh, you know, or, or started to have interactions with people at Schrodinger was through that, through that work.

Harry Glorikian: So, I mean, you know, you gotta hand it to the investors for. This was not a, a, a, uh, short term, uh, sort of turnaround, uh, to get to where you want it to go to took 30 years to sort of get to where you are now. Let’s say with, with, you know, twists and turns along the way, would you say that that timeline because of technology’s advancements, maybe is getting shorter.

Ramy Farid: Well, first you, you, you raise a really important point, which is the patient investors that were involved in Schrodinger, or none of this would have been possible without David Shaw and Bill Gates who are really extraordinary people who I think are the type of people who can see exactly as you’re saying, you know, who can see 20 years into the future. That’s not too many people can do that and have a track record of being able to do that. So. So they’re they’re the ones that provided the sort of extraordinarily patient capital that was required to solve these really hard problems.

Now you’re asking another question, which is, you know, does, you know, going forward will advances sort of take less time? I think the answer is in a lot of ways. Yes. Because of course computers are advancing as well. So advances in the hardware are really critical, too. Um, being able to make these types of advances.

And of course, once you have, uh, A lot of the basic science, I don’t want it. I’m hesitant to use the word solved because of course, you know, in some ways you’d never solve these problems, but you know what I mean? Right. Once you have strong foundation, you can build on that and you generally see progress accelerating, right.

Once you have that sort of baseline science and in place. So yeah, I really think that we’re going to see a shorter time between sort of major breakthroughs in this space because of what we and other, other groups have done. And again, because of advances in hardware.

Harry Glorikian: Yeah. I mean, it was, it was interesting.

I was talking to somebody earlier in there they’re like, well, when do you think we’re going to have a vaccine for coronavirus? And I was like, you know, I said, well, I probably not till the fall. And they’re like, Oh my, that long I’m like, are you kidding me? I mean, compared to. The past, we’re moving at the speed of light, relatively speaking, um, compared to how we used to do it. Uh, so-

Ramy Farid: That’s a great point, Erin, and we’re, we’re definitely seeing that on just drug discovery itself. I think there is now it’s now possible to develop medicines for targets that, uh, would have been very difficult in the past. And certainly we’re seeing an acceleration. And the time that it takes from when you have an idea, like a target, you know, to something that’s ready to go into, for example, GLP tox studies in advance of going into the clinic.

I think that’s right, but also on the underlying science itself, we’ll see more rapid advances there. As well.

Harry Glorikian:  So there are different pieces of software Schrodinger where, you know, they’re, they’re developed to solve different pieces of the puzzle. And so why is this approach more favorable than, than say the way chemists have approached, you know, the design of molecules historically.

Ramy Farid: Yeah. So that goes back to what we were talking about before, which is essentially what you said in your introduction about the complexity of these. Of the properties that, that are required to design into, into a drug molecule or into a material. So let me give you an example, and I think this will really drive home the point.

So a traditional way, as you say, a chemist, trying to let’s say gas or predict what the solubility of a molecule will be, what can they do so they can just look at the molecule. Maybe they can look at it in 3D and see what it might look like in solution. But usually what they’re doing is looking at a 2D representation of the molecule.

And then they’re making guesses use based on past experience about how adding a polar or a group or making the molecule more, three-dimensional less flat might improve the solubility. And the problem is that that’s missing a huge part of the problem. Solubility is not just about how stable the molecule is, or let’s say how happy it is and water, but it’s how unstable or stable the solid form is.

And no human, no person can look at or chemists. It doesn’t matter how experienced they are. Can look at a molecule and make any kind of determination on what the structure is of the solid and let alone the stability of the, of the solid. So it’s an enormously complicated problem. Solubility, that’s not amenable to pattern recognition, especially pattern recognition that is only looking at one half of the problem, which is the only half you can look at the molecule itself.

You can’t look at the solid, you don’t know what it looks like, and you can’t guess that. So what we’re doing or trying to do with the so-called physics-based methods is model both the molecule in solution, as well as the molecule and solid same thing as for affinity or potency, you know, looking at the molecule and trying to determine whether it binds to a protein is missing a huge amount of the problem.

You have to understand the bound state as well as the Unbounced state. That’s just the nature of the problem. So does that make sense?

Harry Glorikian: The first thing that crossed my mind is cause I remember, you know, chemistry a long time ago. Right. But are we teaching for, for this next generation? Are we teaching chemistry the right way or are we, Oh, I don’t know.

You know, getting people stuck in a rut, uh, cause they learn. A certain set of information pathway, um, you know, the machine sort of doesn’t actually do this, but it’s free to sort of move about as it sees the most efficient path.

Ramy Farid: Yeah. Here’s the remarkable thing about that. We are actually teaching in freshmen chemistry classes, the right thing.

We are a freshmen chemistry student. Understands. What I just described to you understands the concept of free energy and the concept of enthalpy and entropy and the concept of, of the final state, the energy of the final state minus the energy of the, of the initial state is what defines free energy. We all know that that’s not actually the problem.

The problem is that we started talking about this idea of using computers. To drive design of molecules in the mid eighties. And it was premature, the science wasn’t there as far as being able to actually run these calculations and certainly compute power wasn’t was not there. So what happened is that people, even though they understood these basic principles that I just described basically said, well, the problem is just too hard.

So we have to cut corners and we have to keep cutting corners and keep cutting corners. And then what happened is it took too long. To basically get the science to the point where it’s working the way it is right now. So you’ll have a whole generation, maybe even two generations of kids, Janice, that are just used to this idea that, well, we can’t get the physics, right.

So we’re just going to have to cut corners and do things by intuition. By trial and error by pattern recognition. So now what’s happening is that there’s a recognition that there have been actual advances in the last 30 years or 35 years. And the new generation of chemists are, uh, are, are much more open to essentially re-evaluating the ID, the role of computers and drug discovery, because of course they’re not, they didn’t go through the 25, 30, 30, 35 years.

Of over promising and under delivering that created, I think this, uh, you know, skepticism about the role of computers, even though we understand the underlying principles and the physics.

Harry Glorikian: That’s interesting. Right. We talk about computer bias and now we’re talking about human bias, right.

Ramy Farid: Exactly

Harry Glorikian: Um, so you, you, you know, looking at the company you guys have, you know, sort of, when I look at all the products, it starts at a, and it ends at Z.

Right. Um, and, and so how did, how can you give me an example? Well, maybe of, of how they might be used in combination to either solve a. A problem in either drug discovery or material science.

Ramy Farid: Absolutely. Yeah. It’s a really great question. So let’s consider drug discovery and let’s think about a problem.

That’s very typical for a pharma company or biotech company worry about, which is they have a very interesting target and would like to develop a small molecule inhibitor for that protein target. And the first thing that you have to do is you to, to deploy these physics-based methods, you need to know what the structure, the starting structure let’s call it of the protein molecule bound together.

So we’ve developed software for determining the structures of proteins by either refining x-ray structures, cryo-EM structures, or even building. Uh, structures from, for example, using homology modeling. So there were, we have a lot of software around the very beginning of the process, determining the structures of proteins to high enough accuracy, to be able to use them as input for these physics-based methods.

Now, once you have done that, it’s time to start designing molecules. So where are the designs going to come from? So we’ve developed software for a numerated chemical space for. Uh, identifying molecules that are synthesizable that have some possibility of being a decent inhibitor for the target. And then we’ve developed physics-based methods for calculating the properties, some of the properties of molecules like solubility.

We talked about like affinity, like selectivity, but we also have had to develop methods for dealing with the fact that these physics-based methods are uh, computationally expensive. They take about a day to calculate one property of a molecule and that’s, um, you know, that’s a lot faster than it used to be, which was maybe it would have taken years, you know, in the past, right.

Given what compute power is now compared to in the past, but it’s still. Slow slow enough that we have to come up with ways to, to, um, to explore more chemical space. So we’ve actually developed methods of using machine learning, trained on the, using these physics-based methods to develop the amount of data that’s required to, to, to advance, uh, a machine learning model or what some people call AI to, um, to be able to filter large numbers of molecules.

And then. As the last step, you take them through these physics-based methods. So you essentially get the advantages of the performance of machine learning, even though they’re not very accurate. But if you combine that with the accuracy of, of these physics based methods, uh, you have something pretty unique.

So I hope that gives you a sense of sort of from beginning target selection, all the way to actually predicting the property of a molecule.

Harry Glorikian: Yeah, no, and it sort of validates some of my, you know, my ideas of like, it’s a combination of some of these tools that I, you know, I can see in many of these companies being the next breakthroughs, as opposed to, you know, people talk about one of these tools say being a magic bullet, I don’t.

I have a hard time believing that. So it’s always, uh, that next level order, you know, improvement comes from how you stack these, um,

Ramy Farid: That’s right. And I think, you know, the field got a little distracted a couple of years ago when AI was. Being very, you know, quite severely overhyped and, and that exact word I think is what people were having in their minds.

This idea that somehow this is magic, you know, that the way it can beat humans at playing go and chess right. In the way it can process images. Right. So people who don’t understand the underlying algorithms, it’s sort of extraordinary. Right. Super human.

Harry Glorikian: Right,

Ramy Farid: right. Um, and, um, but, but, but it’s not, it’s, it’s actually quite simple and It, has a domain of applicability and has, has a role AI it’s not magic. And if you try and apply it to problems that it doesn’t make sense to apply it to it. Isn’t going to work. It’s not that complicated. And I think people are finally realizing that. And that’s nice to see, you know, this field often suffers from sort of over-hyping right.

And it sets us back. You know, you, you, you know, the usual overcompensation right. Of, uh, you know, w you know, over over-hyping things. And then there’s this sort of moment where everybody’s saying, okay, this is all. Garbage. And then eventually you, you calibrate at the, uh, where, where, where the truth lies.

And I think we’re approaching that, that period now.

Harry Glorikian: I just, I, you know, people keep saying, Oh, we’re going to hit another, you know, Ice Age for this stuff. And I’m, I just do not believe it anymore. I mean, I, I try to keep up with, you know, The new chip sets that are coming out, the new software methods that are coming out, the combination that, and I’m like, you know, there’s too much going on.

I mean, it might slow down, but it is, it’s not going, you know, it’s not going to hit an Ice Age in my mind.

Ramy Farid: I don’t think so. I think it’s now very clear that we, it may be a bumpy road, but it’s very clear where we’re headed the. The validation that we’ve seen internally and that our customers have seen, I think is real.

And I don’t think we’ll, you know, as you said, I don’t think we’re headed for some sort of surprise that this stuff doesn’t actually work. I think it’s a very exciting time, um, where, where the advances in the computers and the science have, have come together and sort of extraordinary things are are happening.

Harry Glorikian: Yeah. I think it’s the implementation. That’s going to be the bumpy road more than the technology itself.

Ramy Farid: Yeah, maybe. Yeah.

Harry Glorikian: Well, I see, I see so many people try to take it and put it into their process and what I think they’re not considering is they need to redesign the whole process.

Ramy Farid: Oh, I see what you mean by implementation.

Yeah, absolutely. I think you’re absolutely right. And that is now the next. Uh, the next sort of, um, challenge we have to overcome because, and this is, this is always the case, you know, this is why, whenever there’s a transformative technology, and look, we talked about this, right? What are the chemists? Who, who are.

I’m really determined to continue to do things and sort of a traditional way. And I don’t think there are that many chemists left by the way that think that way. But in order to fully deploy computation at the kind of scale that we’re talking about, definitely, definitely there’s a barrier to doing that.

Um, you’ll have to hire more computational chemists. You have to get on the cloud and get used to this idea that you’re going to be putting data on the cloud. And I think the cloud providers have all done a great job of talking about the security associated with their clouds. And, and I think companies are getting used to that and, and you, you may have to buy computers, you know, internally as well.

So yeah, those are the sort of barriers, but that’s engineering now not, and it’s something that you can sort of, uh, you can, you can see a path to that getting done. It’s not like it was 20 years ago when we were talking about the science and we weren’t, we didn’t know if we were 20 years away or 50 years or a hundred years away.

Right. So I think you’re absolutely right.

Harry Glorikian:Yeah. I like engineering problems.

Rami Farid: Yeah, exactly. Yeah. That’s right. You can put them on a schedule and you can get them done. Just a matter of just getting it done.

Harry Glorikian: So begs the question of, you know, what are the, what are the goals of the company? I mean, I, I, you know, I can, I can read on the website, but, but you know, where do you see the company going?

I’m sure that, you know, Mr. Shaw and Mr. Gates have have visions, um, of where they would like this to ultimately go. And I’m sure it wasn’t just a, you know, an IPO it’s, it’s, it’s gotta be more than that. Um, w where do you see the company going?

Ramy Farid: Yeah, yeah, no, that’s absolutely right. That’s definitely not what they, why they invested in the company early on.

There were a couple of areas, one, and the really the most important we’ve touched on this. Is continuing to make advances in the science, as exciting as this time is I think 10 years we’ll look back and say, wow, that was sort of primitive what we were doing back then. There’s, there’s a lot to do. And I’m excited about the, um, what we just touched on, which is that as you start to solve a lot of these basic science problems, It does start to become more of an engineering problem.

And that’s obviously an exciting transition than any field can undergo. Now, of course, there’s still a lot of basic science to be done. So that’s really a big, big focus of the company is to continue to make these, these scientific advances. And keep getting better. I keep expanding the domain of applicability of the technology and keep improving the accuracy.

And then now, now. What does that translate into, hopefully that translates into, uh, more people, what we’ve been talking about or more people, um, deploying the technology at a higher and higher scale. And of course our own, uh, projects, the ones we work on collaboratively with our partners, as well as our own internal programs.

Benefiting from these advances in the technology and themselves advancing to sort of value creation points, you know, compounds that are ready to go into the clinic, uh, compounds that make it through the clinic and actually help patients, you know, ultimately that’s why Bill Gates and David Shaw. Excited about this, that this could really help people, not only in advancing medicines that help patients, but of course, developing materials that, that improve the quality of life.

Harry Glorikian: Yeah. Um, uh, Mr. Gates seems to have that emo,

Ramy Farid: uh, no question about it in both spaces, you know, in materials he’s and, and, and, and medicines, it’s pretty extraordinary.

Harry Glorikian: Well, look, I can only wish you and the company tremendous luck and tremendous success. Um, and, uh, it was, it was really great talking to you on the show today.

Ramy Farid: Uh, same here. I really enjoyed it, Harry. Thank you very much. Excellent.

Harry Glorikian: And that’s it for this episode, if you enjoyed Moneyball medicine. Please head over to iTunes to subscribe, rate, and leave a review. It is greatly appreciated. Hope you join us next time until then farewell.