How new communities are propelling the future of tech + bio

Michael  Retchin:
Okay, one question is, I don't think I'm hearing anyone else in this.

Danny Crichton:
Well, that is, how's that?

Michael  Retchin:
Much better.

Danny Crichton:
Now we turned you on.

Michael  Retchin:
Oh Yeah.

Danny Crichton:
Now that's a great opening to a show.

Hello and welcome to Securities, a podcast and newsletter devoted to science, technology, finance, and the human condition. I'm your host, Danny Chrichton, and today we're talking about bio community and specifically how a new generation of engineers in bio and machine learning are fusing these disciplines together and redefining the future of therapies.

Joining me today is Nicholas Larus-Stone, the first software engineering hire at Octave Bio, a lux-backed synthetic biology startup that engineer sells to analyze all their activities and create therapies for complex diseases. He's also the founder of Bits and Bio, a global community of technologists devoted to. People who build tools that help scientists unlock new insights.

Second, Michael Retchin is a PhD student at Weill Cornell Medicine and built the foundational technologies at Biofirm Celerity, which uses high resolution cell data to create medicines from a cell-centric point of view. He's the founder of Nucleat, a free and collaborative student-run organization that facilitates the formation of pioneering life science companies. And these two have known each other for a while and for the first time in person, they're meeting at the Lux office. Joining us on Securities. Welcome.

Michael  Retchin:
Thanks for having us.

Nicholas Larus-Stone:
Super happy to be here.

Danny Crichton:
And in addition to me, I have Shaq Vayda, a member of our investment team here at Lux Capital whose focus is on the niche between bio meeting technology. All right, let's just dive in. We've normally talked in this industry about biotech, biotech, biotech. We've heard that for decades, and you're focused on something called Tech Bio, which to me sounds like someone did a little bit of an autocorrect problem at some point, added it to the dictionary and it's constantly getting fixed in iMessages. I'd love to just start out why this focus on tech bio, this rebranding of this field? Maybe Nicholas, you can get us started.

Nicholas Larus-Stone:
I think it's a broad term and a lot of what it's used for right now is actually founder-led biotech. Traditionally the model in biotech has been incubated out of venture capitalists. They kind of put in an executive team, et cetera. These days we're seeing more and more grad students and professors starting their own companies, kind of owning that, putting in their own contacts as opposed to really being incubated out of venture capital. That's coincided with the tech bio revolution and I think those two things feed on each other, right? Professors with good ideas, technologies and labs can then start companies and be backed by venture capitalists without having to be within the network already.

Danny Crichton:
And when I think about biotech, I think going back two decades ago, they were incubated at venture capital firms because you oftentimes had to buy IP. You would go to this lab, a bio professor would invent some new to tool or technique or make some discovery and you go through a commercialization process of the Bayh-Dole Act, finding it, spinning it out of university research. You have the scientists who become the CSO, the chief science officer, and then they sort of craft an executive team of bio practitioners around them to sell it and develop the research program. What makes this new field of machine-learning-meets-bio easier for professors and grad students to just enter right into it without going through that effect?

Nicholas Larus-Stone:
Honestly, the tech industry over the last 10 to 15 years, we've seen software companies do this and the biological sciences have seen that and said, "Okay, we can do this too." I think there's been a big influence from the software field on how people think about starting companies out of universities.

Shaq Vayda:
Something I think it's worth appreciating is in technology, what we've been able to establish is short feedback loops. We've been able to understand how to iterate quickly. We've gone from this waterfall style of development to much more of an agile level of development. Those same innovations haven't found their way into biotechnology. The idea of running an experiment, waiting a week, waiting two weeks, getting that data coming back, figuring out if it worked, making changes, that fundamentally slows down the process, which is why we see seven, 10 years before we can get a drug approved. This idea of taking machine learning and other technologies and applying them to drug discovery allows us to get those feedback loops in a much faster period of way, get that feedback incorporated and build upon that.

Danny Crichton:
Well I think that's a good time to talk just contextually that when it comes to the drug discovery process, costs have soared over the last two, three decades. The number of new entities going through the pipelines has decreased over time as those costs have risen. And so there is sort of this cataclysm coming up, which is that we obviously need more therapies than ever covering more diseases, but they're not a lot coming through the pipeline given the cost structure. To me, this has a huge opportunity to be extremely disruptive to the traditional model of biotech.

Nicholas Larus-Stone:
There's a name for this phenomena, which is Eroom's Law. People are familiar with Moore's Law, the concept that computing power will double every 18 months. Drug discovery is Eroom's Law. The inverse, which is drugs get exponentially more expensive over time, which is bad. And I see a large part of this tech bio movement is trying to fight against that with what kind of Michael and Shaq are saying with faster timelines, earlier decision points so you don't need to spend as much money and it doesn't take as long to actually discover these drugs.

Danny Crichton:
And let's talk about the community. You've both created slightly different versions, sort of target the same problem but from different models. I'm curious on why you constructed each show. Maybe Michael, we'll start with you. When you think about building Nucleat, you built it as a sort of student-run organization and facilitating sort of life sciences company. Talk a little bit about your model and how you built it.

Michael  Retchin:
Sure thing. I am a PhD candidate at Memorial Sloan Kettering Cancer Center. I'm one of hundreds of students in this organization. I wouldn't even say that I built Nucleat. I think we see it as really a collective achievement of what we've done together. But the original idea was just, hey, we have all these different schools and students and they're not really talking to each other, which is a little surprising in some ways.

I mean, even on the same campus, this originally started at Harvard Medical School and Harvard Business School, why aren't those people talking to each other more? And beyond Harvard in the Boston area, beyond Boston, all over the United States and now across the world actually. And I think that the big need that we see in the kind of public good that we're trying to create is this superhighway, this super connector so that students can connect with each other at the very beginning of their career.

Also postdocs and other folks in academia and really try to wrap their head around, how do I develop my career in life sciences? We see it as not just biotech founders actually, but biotech leaders, founders, early employees, investors, business development and helping people help each other.

Danny Crichton:
Nicholas, when we go to you, Bits and Bio, I think the first chapter was in San Francisco if I recall, or was sort of centralized on the West Coast, but it's now a global community. I'd love to hear the origin story there and then how it kind of metastasized or positively metastasized all around the world.

Nicholas Larus-Stone:
I was at Benevolent AI, which is a larger computational drug discovery startup. And then I moved to Octant for a software engineer at a very small biotech. And I felt like I was missing a community of people to talk about interesting software problems, data models, pipeline technologies, et cetera.

And I had friends who were in health tech and there are all sorts of slacks popping up and people going to meetups, and it felt like there's a gap here. And I said, "Well, why not have a Slack community around that?" It's kind of taken off from there. We are online first, we're organized around the Slack community. We have meetups in cities all across the world. SF is Boston, New York or big hubs, London.

We're hoping to expand to smaller cities as well. And it's really a place for people interested in this intersection of software and biology to come talk about their common problems, to introduce people more from the software side to what are the open problems in biology and over time also introduce people from the bio side on how software can help solve their problems.

Michael  Retchin:
You really know when you fit, I guess a community problem fit or something along those lines when you just slightly open the window and just like people flood in.

Danny Crichton:
You want to describe that, how you started the Slack channel?

Nicholas Larus-Stone:
Yeah, I mean I just posted on Twitter like, "Hey, I'm going to start a Slack community, anyone interested?" And then my DMs just kind of blew up. People really took off from there, which has been awesome to see. And it's very much a community. I kind of started the Slack and then people have started meetups, they've started presentation series, reading groups, all sorts of things that are completely community driven. I think that's awesome. And we're pretty big now and hoping to grow even further.

Danny Crichton:
That's amazing. When I think about meetups, I have not gone to a bio meetup. I do have a bio background, but I've never been to a bio meetup, but I have been to other technology meetups. Programming meetups and enterprise software meetups and the like. There's always sort of a unique set of needs. You're talking about CPM, this sort of community product fit or whatever, CPF, I guess we're sort of talking. I'm making this up, I'm going live.

When I look at an enterprise meetup, it's usually trying to meet customers, meet business development, figure out revenue metrics, and it's very metrics-driven, which is like, how do I think about AR? How do I get recurring revenue up? How do I get my net dollar retention rate up? And there's all these tactics. And when I think about bio, I think, wow, you don't even necessarily have metrics, you haven't even figured out the science of how to do this. You're not only not figuring out the science, you haven't even figured out how to connect computer science and biology necessarily for a lot of folks. I'm curious when you think about the needs that your community needs that you're serving between your two orgs, what comes to mind?

Nicholas Larus-Stone:
There are common lessons that people can learn and share and I think that's a big gap is because often your tech team is relatively small at a biotech compared to kind of a pure tech organization where you might be a third to half engineers, you also need chemists, you also need biologists. Your purely computational folks are probably a much smaller percentage of your organization. Where can you get new ideas? Where can you bounce your ideas off someone else? Bits and Bio, hopefully, is a place. And Nucleate, I'm sure on the bio side, is also a place that people can do that. That's where I see these communities fitting in is really providing growth opportunities for individuals as well as helping you be better at your job.

Michael  Retchin:
I think it even goes back even kind of a priority, what is the reason to connect people in the first place? Of course I have a perspective I guess as a graduate student that it can be a little lonely and it's not so dissimilar from being a founder. I think being a founder can be lonely as well. By connecting people almost serve as these role models, whether or not they're like mentors or even just peers that show you this is possible, this is a human being, they achieve this and it shows you lots of different examples of what could be possible for yourself as well.

And I think even just the act, this sort sort of audacious act of putting people together, it presents new possibilities for people in their own lives. And then once they see what's possible, they get inspired, they start to come up with their own ideas and they say, I could pursue that as well. Then you say, okay, well then who should I talk to and how can I solve these real problems? Even then you return back to the community and get stronger.

Danny Crichton:
That brings up a huge question. Obviously I think Silicon Valley and Pure Tech has drawn a huge amount of the talent of the major schools over the last decade. There seems to be this crux moment or this pivot moment where crypto's drawing some folks, you've got some folks going into the metaverse and the content world, you're going into climate and climate change has really affected I think a lot of folks. And then in my three Cs I always add a B for bio as sort of this wild card of factors that's going on in the next decade.

I'm curious, when you think about the communities people are joining and how you draw them in, are the folks here true believers who are totally into the biotech world and are looking to move forward or are you starting to get to that next rung of folks who are saying, God, I'm kind of Bio-curious so to speak, where they might have heard something or they saw a news story or a new science journal article got posted and they're like, that's super interesting. I want to do something other than get my tweets to be more optimized or something like that.

Nicholas Larus-Stone:
When recruiting software engineers, I talk to lots and lots of software engineers who say I'm interested in doing something to help make the world a better place. I'm interested in bio, but there's no real resource on how to start. What are the open problems here?

What technologies would I even need to know? How could I be useful? We're hoping that Bits and Bio provides a community where we can answer questions and you can be introduced to people who can help answer those questions as well as starting to produce resources so you can refer to that. And I think Nucleates actually done a really good job of this on the bio side and we take inspiration from them.

Michael  Retchin:
Yeah, I appreciate that. Yeah, it's been interesting. I mean the community kind of takes itself, it becomes this almost living organism or an ecosystem that kind of transitions in certain directions.

Shaq Vayda:
And it's no secret that the big tech companies are working on these types of problems themselves. They've found that in order to attract the best and get them occupied and interested, they need to work on tough problems. And things like protein folding are coming out of places like Google. Facebook has a huge dedicated division of their AI team working on chemistry.

Michael  Retchin:
And just to add to that, maybe one other way to think about it as well is software goes, in some cases, where the data is. Certainly machine learning goes where the data is.

Danny Crichton:
And why is that? Why is the data just exploding?

Michael  Retchin:
A lot of reasons. I think sequencing costs have come down. You're seeing even more competition coming into the space, new tools for generating data and interpreting data, the sheer number of molecules that exist in our bodies and how you could just try to intervene and measure them. It gets combinatorial and very large, very quickly. And really you're only limited by your imagination and the questions that you can ask.

Danny Crichton:
Oh yeah, we went from the cellular genomics versus the human genome project in the 90s where I think the cost per sequence was what, three and a half billion or something on that order. And I believe we can do a full genome test today for what, a hundred dollars or so to sequence the entire thing? Presumably it will even go below that in the next couple of years.

Our ability to scan almost every genetic code we go through the RNA and the proteomics and different cells. We're just adding all these different instrumentations. I guess it's not just that we're getting more data from each of these, but we're also expanding across so many different categories. And I know all the omics. We've got genomics, we've got proteonomics and metabolonomics, metabolomics, sorry, but so many omics and we have so much from there.

We have tough problems and we have a tool set of solutions. You're starting to build best practices, knowledge sharing, et cetera, in these communities. But obviously there's got to be still challenges to going and building these companies. I'm wondering what the friction points are today, even with the community in place, they're stopping people from starting companies or joining them in the biotech world.

Nicholas Larus-Stone:
The biggest challenge is that drug discovery is really, really hard. I had a colleague who used to say, "we work on an impossible problem." And it really feels like that sometimes. And as Shaq was saying, people want to work on impossible problems. Those are the fun problems, but there's so many different parts to drug discovery that we don't have time to go in now, but there's computational problems. There's problems at a fundamental understanding of biology, chemistry, chemistry in vitro in a test tube versus in vivo in an animal and then translating from animals to humans. All of these parts of the process, none of them have been solved. If they had been solved, we would be able to print drugs, which we can't do yet.

Shaq Vayda:
And not to even mention reproducibility, I think we were having a conversation about this, but how few experiments can even be reproduced in the exact same manner in which the data is generated in the first place.

Nicholas Larus-Stone:
The set of problems is so large and it spans so many different fields that getting one person's head around all of that is impossible. You need to work in cross disciplinary teams. That's always a challenge. Each of these fields has technical challenges. Working together has its own set of challenges, and that's what makes drug discovery so hard and so fun.

Shaq Vayda:
My sort of future world that I want to live in is similar to how kind of cloud computing was able to really change the game for any developer who can grab a laptop, push a few lines of code, and all of a sudden can have a web application, how do we build something similar for drug discovery? Now, there are many, many problems that both Michael and Nicholas have sort of highlighted from us getting from where we are today to getting there. But imagine a world in which there's a completely automated lab based somewhere else and you're able to grab a laptop, push a few lines of code, and you can ultimately get a compound back. Now that's powerful, that is a fundamental cost savings change that would allow us to just do more experiments, get more data, and ultimately generate sort of a whole new paradigm for what technology and biology could look like.

Michael  Retchin:
Four years ago I had a homework assignment and I proposed in the homework assignment like a Jupyter Notebook. You click enter and then it runs an experiment and there was some other nuances to it, but then I got the homework back. It got pretty bad marks because the professor said, "This is really unrealistic." I think we're getting closer to it.

Danny Crichton:
Now let me ask you on this. I've always been very curious. You have the wet lab sort of version of bio, which is doing stuff in the physical, visceral, real world, and then you have sort of the computational world where everything happens "In the cloud." You have sort of the idea of a computational mouse where you could run a drug encode on a coded mouse as opposed to an actual lab animal and be able to run experiments.

I'm curious, when you think about the direction that the field is going in the future, do you see this going to an all virtual world or sort of a hybrid where you're sort of coding in a Jupyter Notebook, it's running basically a remote lab, maybe complete automated with robotics? Or are we going to continue to use labs as we do today where actual humans are going to be monitoring the lab, doing work pipetting as they do today maybe with a more fancy machine? I'm curious where on that spectrum you think the field's going to go?

Nicholas Larus-Stone:
My unsatisfying answer is all three.

Danny Crichton:
All three. Your fake binary choice doesn't work for me.

Nicholas Larus-Stone:
I think as a computer scientist, the ideal is, okay, let's simulate everything. Let's just run it in code. I want to be able to run biological experiments purely in code and we're not close to that. Maybe someday in the future, sure. But there's so much we don't understand about biology and as Shaq was pointing out with the reproducibility crisis, there's a lot that is hard to trust when it's outside of your kind of strict domain.

That's why I don't think purely virtual, nor even kind of fully remote, is going to take off entirely. I think there will be certain experiments that you want to run locally, especially things that you're developing, and then maybe at some point you can kind of package that up and have someone else run it for you. But I think you're going to see all three in the future and the purely virtual stuff, we still have a ways to go on that.

Michael  Retchin:
That's absolutely true. I would say, okay, what is maybe an underappreciated aspect of being able to have a fully virtual kind of experiment? I guess they say there's no such thing as a perfect model. All models are wrong, some are useful. That said, if you have the right model, I think something you can do that's really interesting compared to that, that actually actual experiments we'll never be able to capture is on the infinite horizon of counterfactuals, simulating path dependent decision making infinitely.

And this is something that I'm thinking about a lot in the context of my PhD, actually, in my academic work is can we try to understand how, for example, medicinal chemists think about decision making in a drug discovery context? You couldn't run the same drug discovery campaign, let's say, that costs you a hundred million dollars. You're not going to say, "Well, we just did that. That went really well. Let's see what would happen if we went back to step number five and tweaked and chose this instead of that." That's to me, one of the interesting superpowers of computation, is you can ask those counterfactual questions and then try to transfer them over into the experimental decision making.

Nicholas Larus-Stone:
I think that's a key point is we will use whatever technology is necessary to improve scientific decision making. And I think that's the core thesis of tech bios of our community is whatever you can do to empower scientists to make better decisions, that's what we should do.

Shaq Vayda:
It's fascinating to me. The more I learn about the space, the more I realize what we don't know, and the fact that we have the world saving sort of lifesaving therapies that we have are just in some ways a miracle.

Danny Crichton:
Well, it is miraculous. I mean, I think the miracle of life, all these different layers, building together to create what we see in living organisms is truly miraculous. And on top of the miracle of life, we have this miracle of these two different communities coming together self-organizing and leading the field and the charge going forward. But that's all the time that we have. Nicholas and Michael, thank you so much for joining us.

Nicholas Larus-Stone:
Thank you for having us.

Danny Crichton:
Appreciate it.