Stacie Calad-Thomason is Head of BioPharma Labs and Manufacturing Business Development at NVIDIA. She has a PhD in Chemistry from UC Irvine and a B.A. in Chemistry from UC Berkeley. Topics & Timecodes: Physical AI Tools (Omniverse, Cosmos, Isaac, Groot) [04:46] Lab Automation Bottlenecks [06:20] Simulation vs Physical AI in Biology [09:42] Stacy's PhD Journey & Serendipitous Discovery [13:21] Transition from Academia to GSK Industrial R&D [19:15] Early AI Adoption in Drug Discovery (ADAM Consortium) [23:57] Efficiency as Career Motivation [27:14] Leadership at Atomwise During COVID [29:02] Biosymetrics & Phenomics Work [33:31] Vision for AI Scientists [35:11] AI Impact on Research Workflows [38:09] Career Advice: Perseverance, Adaptability & Skill Breadth [41:10]
Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
Stacie-Calad ThomsonAnybody that has a PhD, I know that they've got some grit and some perseverance and they're capable of learning. Right. That's what I, that's what I see in others when I see, you know, those letters behind the name. I don't use my PhD on a daily basis now. I'm not doing organic chemistry reactions. I have the capacity to learn and understand and ask questions that help me learn something new about what other people are doing, though. And I use that in my job every day.
Nick EdwardsForeign.
Speaker 3This is the Once a Scientist podcast. I'm Nick Edwards. We're back with new episodes, so keep an eye out and subscribe to the podcast if you haven't already.
Nick EdwardsI am super excited to have Stacy co Edson on the podcast. Stacy is an OG in the AI life sciences space. She's the global business development lead for Pharma Labs and Manufacturing at Nvidia, if you've ever heard of Nvidia. And so, Stacy, thanks so much for taking the time to come on. I'm excited to chat with you.
Stacie-Calad ThomsonYeah. Thank you for inviting me.
Nick EdwardsYeah, of course. So what is it that you do at Nvidia?
Stacie-Calad ThomsonYeah, so I've been at Nvidia for two years now. Got to join in this awesome role in business development and life sciences where I've been working across all pharma and biotech and really understanding sort of how people are using AI, how we can help them accelerate some of their models, some of their training, some of their inference, and ultimately, hopefully that drives more compute somewhere wherever they're consuming it. So recently I took on a new role just focusing on labs and manufacturing.
Stacie-Calad ThomsonWe see this as really an emerging area of growth for us. So, so what that means is I get to talk to people that are focused on lab in the loops, whether that's on physical AI and robotics or automation, or the data platforms where data is coming off of all these instruments, or the agentic layer that's helping to orchestrate all of it. Manufacturing has equivalent systems that are, I think, even more complicated. And all of this, I think, has been currently maybe the bottleneck to really unlocking new scientific discover, at least on the lab side.
Stacie-Calad ThomsonSo I'm excited to pull together some of the technologies and research that Nvidia is doing in this space and connect it to a life science audience and hope that the lab won't be a bottleneck going forward.
Nick EdwardsYeah, you have like, probably one of the coolest positions, I mean, because you get to like really work across the entire Ecosystem, lots of different AI companies, probably pharma companies. So must be really interesting kind of perspective. You play a really interesting role. Tell me like how you guys think about kind of building an ecosystem because it seems like that's a really important part of Nvidia. You know, I've been to GTC and some, you know, some of the events that you guys host, you do some like model hosting for Bionemo for example, for life sciences models. But, but then you play this really important intersection and is your goal really mostly focused on is it focused on building that ecosystem so that you can kind of enable more compute?
Stacie-Calad ThomsonYeah, it's. And we've got a complicated, I think go to market. We are doing a lot of active research in building, you know, new models of biology, of chemistry and putting those out as open source. And that's one of the ways that we learn ourselves so that we're better able to understand what are some of the bottlenecks that folks are running into and how can we uniquely accelerate those by maybe at some underlying software kernel, some CUDA kernel.
Stacie-Calad ThomsonAnd so we've got a lot of research focused on that. And then yeah, we're working with a lot of the tech BIOS that are building some of these protein structure prediction models. We're working with that entire ecosystem because we see those as really being underpinning future drug discovery research programs. And pharma are using those programs. Sometimes they're partnering and licensing those models. Like we've seen a lot of deals this year that have happened.
Stacie-Calad ThomsonSometimes they're taking open source versions of those models and fine tuning them on their own data. And so I think that's an example of how working with an ecosystem enables others to then explore and grow from that. So our goal is to not necessarily work with an end customer, but rather enable an ecosystem that then scales and builds on our technology.
Stacie-Calad ThomsonAnd that's sort of the goal here.
Nick EdwardsIt's really cool because you have all the models and kind of, and things that you're open sourcing. And so you see this come out in like the, in a lot of the announcements, partnership announcements where people are using, you know, some of the world models. Remind me the physical AI model that you guys have. I think you had partnership with AIRA at hi Rez where they'd integrated some of the tools.
Stacie-Calad ThomsonWe've got a lot of tools that we're working with in kind of the physical layer and maybe it starts at the agentic layer. So we've Got like our Nematron suite of reasoning models and high res has picked up those. And that's one of the models that's under the hood of their Solario OS orchestration platform. We've also got physical AI models, everything from Omniverse, which is a platform for building digital twins of physical environments.
Stacie-Calad ThomsonWe've got our Cosmos open model platform as well. Cosmos Reason is one model that hi res has also picked up. This is a vision language model that they're using. Basically a camera on an automation system is able to detect errors. Like if a plate is sealed or unsealed, is it in the right position? You're able to reason with it as well and ask it, oh, was there anything weird that happened in that run earlier today that might explain why my results are different than what I expected and be able to go back and analyze results in a really intelligent way by interacting with that physical environment.
Stacie-Calad ThomsonSo that's one option there. And then our physical AI stack extends into robotics, our Isaac platform, our Groot platform. So we've got a lot of cute names for all of our physical AI products.
Nick EdwardsNice. Yeah, they're all science themed. It sounds like Isaac Newton Cosmos. Like, yeah, like the series. I like it. So you talked a little bit about like the bottleneck. What do you think is the bottleneck right now? Like from your perspective? Is it the physical execution of experiments?
Stacie-Calad ThomsonFrom what I hear in talking to our ecosystem partners, I think there's a non compute bottleneck here right now which is just getting instruments connected, whether that's APIs or drivers. And I know there's a huge collection of effort that is moving towards this. Data standards is another one that I know there's a lot happening in the ecosystem as well, but in the physical space I actually think we've got a ton of automation capabilities that have been around for a while and that have been improving since, you know, like the 1990s when high throughput screening became a huge investment for pharma.
Stacie-Calad ThomsonEverybody's been investing in automation, in liquid handlers, in high throughput experimentation and for highly repetitive experiments. I think that we've got great equipment that does it. We don't have maybe great interfaces yet where every Joe Schmo scientist is able to operate that equipment.
Nick EdwardsEquipment.
Stacie-Calad ThomsonAnd that I think is something that we are on the cusp of changing. And I'm excited about the opportunity for agents to come in here and be able to make those connections to instruments, to make those reasoning connections on experiments and to tie together the Hypotheses that AI scientists and AI co scientists can make along with orchestration agents like the Potato platform.
Nick EdwardsThis episode is brought to you by Potato. We're building AI agents for closed loop autonomous science. Check it out at Potato AI or email us at hello Potato AI. If you want to figure out ways.
Stacie-Calad ThomsonTo collaborate and then, you know, every instrument, whether it's an automation instrument or a typical, you know, other bench instrument, I think is going to be at a point where it's going to be connected. Then it's just a matter of like, you know, putting your agents out as a workforce that's able to execute work on your behalf.
Nick EdwardsYeah. So kind of closing that like intent to execution gap and kind of flexible workflows. I think a lot about like, you know, we have all these capabilities for high throughput screening, but even just like getting a simple assay to work sometimes on instrumentation, it requires a significant amount of engineering expertise. So why can't we just simulate biology? Why, why do we need physical AI? What, what's your perspective here, Stacy?
Stacie-Calad ThomsonYeah, there's a lot of, of different layers of biology and I think where we are able to simulate is where we understand physics really well, like free energy perturbation. Now you can run simulation experiments at the molecular level that are basically as good and as predictive as running experimental assays now, which is crazy.
Nick EdwardsThat's insane. Right.
Stacie-Calad ThomsonAnd I think that there will be a point in time where compute costs are coming down and that means that you're going to be able to simulate experiments that are like too expensive to simulate right now. Right. So we will get there to a point where. Yeah, I mean Schrodinger is already doing this right now with their TALKS panel. Like you can do a hundred off target tox predictions, I think on using FEP for predicting that instead of running like your TALKS panel, the cost and economics of this might still be that it's cheaper to run the actual experiment because we've gotten those down to like a, a very highly efficient panel for doing that screening.
Stacie-Calad ThomsonBut those economics could be shifting I think in the near future. Now if you come up a level though, from just the molecular interactions we get to the virtual cell models and there's a lot of effort happening right now to understand all the complexities that happen when you make perturbations in a cell and understand what the downstream effects of that are. But there's significant science I think still happening at that level.
Stacie-Calad ThomsonThen you go up a level and it's like, how are Cells interacting in tissues? How are tissues interacting in organs? How are organs interacting with all of the systems of your biology? And so there's a multi scale biology problem here that like can't be simulated yet because we don't understand enough biology. And so this is where the wet lab is still needed because we are operating in very data poor environments. And this isn't just throw enough physics at it to simulate it's truly an understanding of biological pathways and mechanisms in very complicated ways.
Nick EdwardsYeah, and I think the kind of the physical AI is what really enables that data generation so that you can build you know, more and more foundation models for different types of like simulation.
Stacie-Calad ThomsonFor example, there's physics AI or like physics that is used for generating data and then there is physical AI, which is I think of more as the, you know, operationally in a laboratory, can we bring some reasoning to how we physically execute experiments so that we can generate more of that data that then helps us simulate those biological processes?
Stacie-Calad ThomsonSo that's where I think those two things connect is physics and physical AI being those two different things.
Nick EdwardsThat's interesting. That makes a lot of sense. And yeah, I mean I guess depending on how you think about like world models, it plays a really big role in this in terms of being able to inform of like what that physical experimentation, that physical instrumentation looks like. And really interesting. So you were kind of super early to this space AI for drug discovery, right? I, I want to hear a little bit about your story. Can you tell? So you've had just like this crazy journey. PhD in chemistry at UC Irvine, like a GSK for research, moved into strategy, spent some time at early stage drug discovery.
Nick EdwardsNow you're in Nvidia. Can you just like, you know, maybe from your own words like tell me a little bit about your path and how you got into this.
Stacie-Calad ThomsonSo I, I have a PhD in organic chemistry. I went to get my PhD because I, I did undergrad research at Berkeley and I didn't know what else to do. And everybody I was working with was like, well you gotta go to grad school. Cause that's what they were doing. And you're working in the lab.
Nick EdwardsAnd doing organic chemistry.
Stacie-Calad ThomsonOrganic chemistry at UC Berkeley.
Nick EdwardsOkay. That was my first love in science. It was like, you know, it's like, cuz, cause it's like you get this glimpse of like how the world actually works and like you can make things like creation such a cool, I don't know, I, I, in some ways I'm Jealous of. Like, I should have. I should have done organic chemistry instead of neuroscience.
Stacie-Calad ThomsonOh. I had a. I had a great chemistry high school teacher, which is how I went into chemistry. I sometimes think if. Yeah, I could have gone different ways. I think as well. And actually I started as a chemical engineering major in undergrad and then with my first chemical engineering class because I thought engineers were the people that, like, solved problems. And they are. But when I learned those problems were more like flow rates, I realized those weren't the problems that I was as passionate about solving as.
Stacie-Calad ThomsonAs much as, I don't know, like, building bonds, optimizing synthetic yield. I know that doesn't sound that interesting either, but that's where I realized I. I did have more fun. And, and in undergrad, I got. I got to work for Professor Clayton. He. He was a. A big organic synthetic, you know, professor. He wrote the textbook, got a great exposure in working in his lab. I was basically doing like, scale up for a grad student that was working on the total synthesis of sponge of statin, so a natural product synthesis.
Stacie-Calad ThomsonAnd I was kind of doing early process chemistry because I would. Would scale up the. The routes that he had already. The. The reactions he had already done so that I could make enough starting material for him that was working on, like, the later reactions and trying to figure out those. And then in grad school, I went to grad school at UC Irvine. Had a lot of ups and downs. I think we've talked about the challenges of science.
Nick EdwardsPlease tell me more because this is what people really want to hear.
Stacie-Calad ThomsonThere can be a lot of heartache in science, right? I. Somebody was joking the other. I heard this joke the other day that PhD stands for probably has depression. And.
Nick EdwardsIt's the first. It hits a little too close to home.
Stacie-Calad ThomsonIt. Yeah. Like, my time on the bench was marked by a lot of truly epic failures. And like, my first year of grad school, the project I was working on, we had this idea for this new reaction type, and I literally had nothing that worked. I would. I was making these, like, starting kind of chemicals and then trying to do this new reaction, and it was like nothing, nothing. It was a whole year of like, nothing working.
Stacie-Calad ThomsonAnd it was at this point that. That actually my. My professor, my advisor pulled me in and he was giving me this talk about, like, sometimes you just gotta cut your losses. And I know, like, you've been working on this, but, like, maybe it's actually not gonn. And. And I was like, okay, yeah, but I finally have Something that was like. It was just that day that I got like a reaction that it didn't work, but it didn't not work. It did something weird. And, and it was through that serendipity I had added one extra carbon would end up happening.
Stacie-Calad ThomsonI was trying to do this reaction and it ended up that the chain was long enough that it like reacted with itself. And that then became what my entire dissertation in the next, you know, four and a half years of research ended up being around was by serendipity I ended up discovering a way to make a, a chiral quaternary carbon center. So that type of reaction is super hard in organic chemistry. Those types of like quaternary centers are common in nature or natural products or and in potential drugs that you might want to make.
Stacie-Calad ThomsonAnd I found a way to make it really efficient. Right. And so that was great. Then it was like off to the races and we got a great publication and a great journal on that. I used that to then synthesize some molecules of biological importance to prove that the methodology had real world use. And that's great, right? That's why we do science is for those moments of like, aha. And those like, amazing discoveries similar to my experience.
Nick EdwardsLike, same thing. It's like spent, you know, years just like grokking and like nothing's working and getting so like down. And then, and then eventually one thing, it's like this tiny little signal that I. This blip in the data and you know. Yeah, that's science for you. What did that teach you about science or about like yourself?
Stacie-Calad ThomsonYeah, that's a great question. Anybody that has a PhD, I know that they've got some grit and some perseverance and they're capable of learning. Right? And capable of persevering through some challenging stuff. And so I think that's what I see in others when I see those letters behind a name. But I think that's what it taught me. I don't use my PhD on a daily basis now.
Stacie-Calad ThomsonI'm not doing organic chemistry reactions. I have the capacity to learn and understand and ask questions that help me learn something new about what other people are doing, though. And I use that in my job every day.
Nick EdwardsTotally.
Stacie-Calad ThomsonSo I think that's the thing that's really pulled through for me.
Nick EdwardsYou're doing the research and then you kind of made it to gsk, like.
Stacie-Calad ThomsonYeah.
Nick EdwardsYou were doing R and D process chemistry there as well, is that right?
Stacie-Calad ThomsonYeah, I started in process chemistry and that leap from, like, academia to industry was one of the biggest, like, culture shock learning shocks, I think, of my career still. I was just talking with somebody the other day about this because, you know, in grad school, you hear a lot today about in tech, people working 996 and. And that kind of was the, that was the approach maybe in grad school.
Stacie-Calad ThomsonAnd it's not because I was working that hard. It's because science just takes a long time. Sometimes reactions take a long time. Data analysis takes a long time. There's a lot of waiting that happens in there as well.
Nick EdwardsIt's not predictable.
Stacie-Calad ThomsonYeah, it's not predictable. Yeah. But then being thrown in at gsk, people are working eight to six now and five days a week, but are so much more productive and the stakes are so much higher and the expectation and actually kind of some of the competitiveness was like, a lot more intense. So it was a good period. I learned a ton. And that's where like, actually, you know, everything that you'd been preparing for was coming to fruition.
Stacie-Calad ThomsonLike, I was actually working on drug programs that were scaling up that were going into clinical trials, and were trying to make these syntheses as efficient as possible. The thing I really liked working on there were the things that were coming straight out of Med Chem that had, you know, like 17% total yield over, like, you know, 15 steps or something like that. And being able to redesign that into, okay, I think we can do this in five steps now because we can optimize the starting materials and optimize the reactions and do this overall in, you know, maybe that first one out of the bat is 60 or 70% overall yield across those five steps.
Stacie-Calad ThomsonThat's pretty good. That's like, really satisfactory. Being able to take it from, I don't know, 17 to 70 or 80%.
Nick EdwardsYeah.
Stacie-Calad ThomsonThe hard thing then is taking it from that 80% to the, you know, like 99%. But there's so many parameters that you need to optimize, and I really had an appreciation for that. But then I guess the next stage of my career then became wanting to know more about what was happening before it got to me and what happens after it leaves me. And like, there's so many other work packages that happen in a drug discovery program.
Stacie-Calad ThomsonAnd that's why I really enjoyed the time that I got to. When I moved off the bench into operations planning and strategy, I got to do a bit of project management and learned all about the resource planning and scheduling and all of the different work packages that have dependency on all the other work packages that are coming from different departments and how all of this, you know, is streamlined into some critical path to get a drug like forward to the next milestone.
Stacie-Calad ThomsonAnd so that gave me a lot of perspective. And then I got to work on some strategic initiatives of which one was, you know, this thing called artificial intelligence. And that was basically about a decade ago. Okay, 2016.
Nick EdwardsOkay, cool.
Stacie-Calad ThomsonYeah, 2016 is. GSK was early in this game too. We had a really visionary.
Nick EdwardsThat was the year that OpenAI was founded I think as well. So it's like, that's interesting. You're like working on very early. Sorry.
Stacie-Calad ThomsonYeah. With a lot of the, you know, the tech bios that are now pushing into the clinic, the recursions, the. In silicone medicines. It was an exciting time where were finding our way, trying to find the data, trying to figure out what models were best, what model architectures were best. So much has changed in the last decade, but yeah, what ended up coming out of this strategic initiative was a collaboration with the U.S. department of Energy, the National Cancer Institute, UCSF and GSK.
Stacie-Calad ThomsonIt was called the Adam Consortia. And I helped do the business development for that deal. Get it stood up with all of the IP restrictions and the consortium goals. It was the hardest deal that I've done in my career now in bd and it was the first one that I had worked on. It was exciting because were all willing to really lean in on this. And that's when I moved out to San Francisco to then lead the consortium that initially was focused on kind of chemistry, small molecule models of ADME and TOX and generative molecular design.
Stacie-Calad ThomsonSo really thinking about how you could do multi parameter optimization and generate new structures that optimize not only for the right binding of a molecule, but what do you not want it to bind for? So like let's take into account selectivity, let's optimize up at the beginning for all of these physical properties, solubility, the kinetics that we want and the toxicity that we don't want. So that was a really exciting project where we did a lot of research, we learned a lot.
Stacie-Calad ThomsonAnd, and you know what? The, the wet lab was still the bottleneck that we had.
Nick EdwardsThinking back on that, was it another moment of serendipity that you found this like ADAM consortia project at gsk or was it like, was there something that you saw that you were kind of, that drew you at an early stage to AI and Different kind of approaches for this.
Stacie-Calad ThomsonYeah. The through thread of my career has really been efficiency. The goal of process chemistry is to make a route more efficient. You want to get the highest yield in the shortest amount of steps. And it's all about efficiency. And then when I was in operations and strategy, that was all about making the business run more efficiently and more streamlined. And so even then I was starting to see the opportunities of computational approaches and just recognizing how much 90% of clinical trials fail.
Stacie-Calad ThomsonYeah, right. It takes six years to get from target to candidate. It takes 12 years on average to get a drug to market and $2 billion. And like the majority of our research never sees the light of day because it fails before we even get to a clinical trial. Right. So there's so much waste and failure that happens in R and D and there's gotta be a better way to do it. And I think has always been my approach.
Stacie-Calad ThomsonAnd while I was early, I was maybe early in the believing that AI could find us a better way. But I'd like to think that I was hopeful that we're sitting on this plethora of data and our human brains are only capable of interpreting and learning a certain aspect of that. And computers can learn a lot faster and have the capacity to take in a lot more data.
Stacie-Calad ThomsonSo I was placing this bet early on that there would be efficiencies here that would find a better way.
Nick EdwardsThat's really cool. That's interesting. I like. You took that kind of process chemistry approach and just like abstracted it more and more, you know, sounds like that's, that's fascinating. So you spent some time doing the atom consortium and then you also did. You were working in, then AI drug discovery for a little bit. Was that right after.
Stacie-Calad ThomsonYeah.
Nick EdwardsOkay.
Stacie-Calad ThomsonYeah.
Nick EdwardsSo what were the early days like for that? Can you tell me a little bit about like what you've seen as like the progression of this field over the last however many years you've been working in it?
Stacie-Calad ThomsonYeah, that's a great aspect. So like in the atom consortia, we built these beautiful data processing pipelines. We built these model pipelines. We were using all kinds of architectures I thought were really at the cutting edge. And were doing this proof of concept project where were doing this generative molecular design, demonstrating that we could predict different structures that optimized for all these parameters in parallel.
Stacie-Calad ThomsonThen went to experimentally validate. It's like, well, it's going to take us eight weeks to synthesize compounds and then it's going to take, you know, another month to get them into all of these other assays. And by the way, it was going to be like, I can't even remember the budget now, but a lot of money. And ultimately, like, it didn't work out of the box. Right. We needed to then take that data and say, oh, well, we got to learn, we got to stick it back in and like, figure out what to do with this.
Stacie-Calad ThomsonBut that cycle time then is enormous, right?
Nick EdwardsYeah.
Stacie-Calad ThomsonThat's when I was really clued into. There's gotta be a better way of moving either more of this in silico or streamlining the experimental feedback loop. The next job that I had after the atom consortia, I went to atomwise, which was another one of these early AI drug discovery pioneers. They are known for the virtual screening technology that they invented and pioneered and patented. Had some great, amazing pharma partnerships where they would do a virtual screen for a pharma target and progress that.
Stacie-Calad ThomsonWhat I was doing there was running their AIMS program, which was working with academic collaborators. And we had like 700 of these partnerships where an academic would have a target of interest and an assay for that target. Atomwise would do the virtual screen and then we'd order a plate of compounds from, you know, a supplier, like enamine or such. And so I joined atomwise March of 2020, right when the world was changing.
Stacie-Calad ThomsonThat was the start of COVID So we. One of the kind of first projects that I worked on was we had all these people coming to us saying, I've got a target and an assay, and we're all hands on deck for doing something to help this Covid crisis.
Nick EdwardsAnd you guys are like, onboarding the assays and screening the.
Stacie-Calad ThomsonSo our partners did that. We had academic partners that were doing that for all the different types of targets, whether it was the main protease or some other targets that were being explored. Early on in the pandemic, there were new structures being thrown up on PDB basically daily with people doing research, sharing this research. And so what we did was we. We selected some of these partners. We selected about a dozen partners. We did the virtual screens and then we bought plates of compounds that they could screen in their assays.
Stacie-Calad ThomsonYou buy the plates of assays and it takes maybe eight weeks for them to format those plates of compounds. Maybe they had to synthesize some of the compounds. We'd, you know, maybe select ones that were in stock, but then they'd have to plate them Out. And I remember one academic partner that we worked with was like, look, I hired this like intern for the summer to work on this, like when am I going to get my compounds? And and were like in eight weeks. And he's like, my summer intern is going to be gone by then.
Stacie-Calad ThomsonLike do you not understand that a pandemic is going on and it's a.
Nick EdwardsState of the art thing that's like,.
Stacie-Calad ThomsonYeah, this is still the bottleneck. Right?
Nick EdwardsYeah.
Stacie-Calad ThomsonSo after Adam Wise I did do some consulting myself for a while. I worked with a few other biotechs that were working in AI drug discovery. I worked for a few VCs that were trying to evaluate companies that were doing AI drug discovery and really differentiate some of the hope from hype. But one of the companies that I was working with as a consultant that I ended up going to full time was Biosymetrics. And at Biosymetrics they were working earlier in the drug discovery process.
Stacie-Calad ThomsonWell I should say earlier and later because were doing phenomics work that could happen at the clinical stage. So understanding and identifying biomarkers to better inform kind of clinical studies or that could also be used for identifying and discovering new drug targets. And to me that was, this is getting at the heart of why those 90% of clinical trials fail is coming back to understanding the biology at that kind of phonemic level, but also at that patient level and being able to close that translation gap where something works in an assay or it works in animal model and then it doesn't work in a patient.
Stacie-Calad ThomsonSo having that through line of predictivity is what I was really interested in my role at Biosymmetrics. I was the chief strategy officer and head of drug discovery. And and so when we started running our drug discovery programs is when I was pulling in consultants from med chem or biological assays and were getting hits out of at a phenotypic screen. And we'd have to like either push those molecules forward into some developability or you'd also have to like deconvolute a target.
Stacie-Calad ThomsonAnd so you know, with limited budget you'd have to make decisions on what experiment to run next. And this is where I'm super excited about like the age of agentic AI that we're in right now and the AI scientists that are coming on board because I think this is where we're going to see the next breakthroughs happening. We would send off, we'd synthesize a compound and that would cost some money. And you'd have that done at a CRO, and then that CRO maybe would send it to the other CRO where were running our assay, and then we'd get that data back and be like, did it work?
Stacie-Calad ThomsonIs it black or is it white? And it's never black or white, it's gray. Right? And like, you'd get stuck and be like, oh, God. Okay, so we've told our investors or our pharma partners that we need to reach, like, this next milestone. And what's the fastest path to get to that milestone? And do we spend our limited budget on like, another chemical iteration, or do we spend it on, like, an orthogonal assay that's going to give us slightly different results and the cycle time for aggregating that data, for interpreting that data, for understanding it with our small, little limited team?
Stacie-Calad ThomsonThis is where I see a huge opportunity with AI scientists that are able to understand the context of the problem you're solving. You know, they're going to be able to have an understanding of all of the work packages that are typically done in an R and D program. They're going to have an understanding of what assays any CRO may have. They're going to have an understanding of what assays you can run in your own laboratory.
Stacie-Calad ThomsonThey're going to know what samples or starting materials you have within your disposal, and they're going to give you the most efficient path to either make or buy. They're going to be able to give you that critical path. Maybe they're going to be able to understand that it's actually going to be faster to make this in house because you've got these starting materials and a retrosynthesis model is able to point you to a more efficient path here than doing it over there. And then ultimately, in the future, be able to just have a conversation with your AI scientist bring together like, I've got this.
Stacie-Calad ThomsonI've got this target. Here's the target product profile that I want to achieve for this drug discovery program. Yeah, go and, like, have it. Be able to say, okay, these are the types of steps that I need to take. I think it's better to do developability early rather than later and this, and time all this out. And then ultimately, you know, if you've got a fully connected wet lab, be able to then talk to your orchestration agent that says, all right, let's go make that over here.
Stacie-Calad ThomsonLet's get the assay prepped over here. I'm also going to Know how much compound I'm going to need to make? Is it a hundred milligrams right now? Is it one gram that I need? And I think we are on the cusp of being able to outsource a lot of that orchestration to agents.
Nick EdwardsWhat do you think that means about, like, the future of work for scientists? Sometimes I think about. Well, I think a lot as we're building tools about, like, how would this have helped me? When I was doing research at the Bench, there was one time when I ran analysis on our agent and it reproduced the main finding from this Nature Neuroscience paper that I published in like five minutes. What took me, you know, at least a couple days. And, you know, my mind was.
Nick EdwardsBut I was like, I think if I could give this to myself in the past, it would be amazing. Right. But, like, I also wonder, like, what does that mean about the way that we do science in the future? What do you think?
Stacie-Calad ThomsonSo when I was an undergrad, there was a saying then that a day in the library would save you a week in the lab.
Nick EdwardsYeah.
Stacie-Calad ThomsonAnd that's because it was the early days of like, digitizing journals, all historical literature. You still had to go to the library to read if somebody had an idea for like a reaction that they wanted to do. You could just like, use your intuition and throw some things together and see if it worked. And it's not going to work the first time. So, like, maybe you'll get some insight though, and by the end of the week you'll have it worked out. Yeah. Or you could go look at what other people have already done and leapfrog to that.
Stacie-Calad ThomsonAnd by day two, you could be there because you spent the day in the library. Right now you don't need to spend a week in the lab, you don't need to spend a day in the library. You need to spend a few hours with an AI scientist that can reason through the historical digitized world that we live in now and be able to have a conversation with you about what is the best next experiment to do. So we talked before about all the heartache that we experienced in our PhDs.
Stacie-Calad ThomsonI'm excited by the pace of discovery. Instead of your AI scientist found that was able to recapitulate your research in a fraction of the time. That just means that you're going to be able to do more research. It doesn't mean that you're going to get your PhD in five hours instead of five years. It means that you're going to do so much more research. I really feel like we're at this inflection point where it's going to be a really exciting time to be doing science.
Nick EdwardsI like that vision.
Stacie-Calad ThomsonYeah.
Nick EdwardsIt's like endless collaborators.
Speaker 3Right?
Nick EdwardsThat's. That's how I think that. To me, that's. There's like, really a couple things that drive science forward, and I've probably repeated this too many times by this point, but it's. It's more and more collaboration and new tools. Right. New instruments and with agents, you kind of get both of those in one swoop. Right. When you think back, like, about those early days, being at the bench, I would imagine. Please correct me if I'm wrong, but I would imagine you did not plan on this path. You didn't know that this would be a path at the time.
Nick EdwardsAm I correct?
Stacie-Calad ThomsonCorrect. This was not. I did not envision. I did not have Nvidia on my. On my career bingo card, you know, when I was getting. When I was studying chemistry.
Nick EdwardsOkay. So if you think back about those early days of being at the bench, like, what. What's the advice you'd give to yourself, you know, especially if you were kind of going through graduate school these days?
Stacie-Calad ThomsonThere's so many times in people's careers where I think, especially in. In challenging science, that. That really, like every scientist I know has had some point where they've doubted themselves and whether they're doing the right thing. And I think that is a healthy question to ask because sometimes it may not be the right thing. And as much as I enjoyed the process of learning, chemistry isn't necessarily something that I am the most passionate about.
Stacie-Calad ThomsonI'm not still, you know, reading chemistry journals by my bedside night. Yeah. At night. Anymore. But it is something that. That got me to where I am now and has given me the tool set to understand the questions to ask and make connections that I wouldn't have otherwise been able to make. So I think my advice is to keep an open mind about what your passion really is and maybe differentiate the process of learning and exploring from the subject matter that you're diving into, because the process of learning and exploring is something that you'll be able to take anywhere in your career.
Stacie-Calad ThomsonAnd I think being open to when those pathways open is also something. It can also be hard. I think it was a big transition when I moved off of the bench. That was a challenging career rut that I almost felt like I was in. And it took a lot of mentoring, it took a lot of networking. To sort of make connections that enabled me to make that move.
Stacie-Calad ThomsonBut every move I've had after that has come along a little more naturally, I think, because I've been open to it. Because I've had a skill set then that hasn't been tied to some single subject matter expertise, but rather to a breadth of skills that are more around execution and understanding and delivery, rather than just one subject matter that I'm an expert at.
Nick EdwardsVery good advice. So, Stacy, really, thank you so much. I think it's going to resonate with a lot of people. It resonates with me for sure. This has been a lot of fun. I appreciate you coming on the podcast, sharing your thoughts. I think people will be very interested because got a very unique perspective. Thanks again. And we'll have to do it again sometime.
Stacie-Calad ThomsonYeah. Thank you, Nick. Always great talking to you. Thanks for the invite.
Speaker 3This is the Once A Scientist podcast. I'm Nick Edwards. We're back with new episodes, so keep an eye out and subscribe to the podcast if you haven't already.
Melissa Crisp, Ph.D. is a Senior Director in Biologics Screening and Automation Technologies at Eli Lilly and Company in San Diego, where she leads a team applying innovative automation and screening technologies to...
Renee Wegrzyn, PhD, is co-founder of Transfyr AI. Recognized as a TIME 100 Innovator, Rock Health Top 50 Digital Health Innovator, and recipient of the Superior Public Service Medal, Renee holds a BS and PhD in Applied...
Ron is Co-Founder and CEO of NOETIK, a next generation precision oncology company using multimodal foundation models pre-trained on large-scale human tumor data to discover cancer therapeutics. He holds an MD and PhD...
Chris Anderson currently serves as an advisor at Pillar VC and Renaissance Philanthropy. He was editor of WIRED and spent seven years at The Economist, where he worked as editor of both the technology and business...
Ira Hoffman is the CEO of HighRes. He has a M.S. in Computer Science and a B.A. in Physics/Mathematics from the University of Vermont. OUTLINE: [00:00] - Introduction to collaborative robots and safe human-robot...