AI in Biotech
Wow, that was fun. I just spent almost 4 hours on a Thursday night with a diverse, exciting, and talkative bunch of scientists, software engineers, and other folks who are interested in the intersection of the life sciences and artificial intelligence. The event (AI in Biotech Engage & Connect) was hosted by Nucleate in partnership with Bits in Bio. Nucleate is a global student-led organization representing a global community of bio-innovators. Bits in Bio is a vibrant group of people who are interested in using software to push science forward. The event was sponsored by Union AI, University of Washington (UW) Department of Immunology, and J.P.Morgan.
The event started with free-form networking and after that we moved into the Orin Smith auditorium at UW. I would guess at least 100 people were there, showing the strong interest in AI, and biotech in the Seattle area.
The main event was a panel discussion. There were two co-moderators: Jag Singh (Software Engineer at Arzeda) and Ian Derrington (Principal Data Scientist at Roche Pharmaceuticals). Along with the moderators there were 4 panelists: Joe Horsman (Investor at Madrona Ventures), Jacob Lee (CEO & Co-Founder at Genemod), Kate Nelson (Executive Director, Life Science – Commercial Banking at JP Morgan Chase & Co), and Pryce Turner (Union AI).
Overall, I think everyone on the panel was quite excited to see where AI can help make progress happen--faster. There was discussion of how/where to apply AI, and that perhaps the best opportunities will come from organizations with carefully-curated “clean” data. That struck a chord with me—30 years ago I was part of the Human Genome Project, the first time the life sciences was generating big data sets that were being analyzed by computers. And one of the most pervasive issues we have to deal with was making sure the data sets were clean, even though the data sets were much smaller than today’s collections. “Garbage in, garbage out” still applies. The future can reflect the past.
Along with clean data sets there was discussion of standardization. Another pain point I remembered from the Human Genome project days, was when the messy, non-standard, organically evolved (inside joke intended) naming conventions of biology bumped up against the requirements of the compute world. At one point even the instruments did not talk to each other, requiring technicians to take images on one instrument, download the image to a floppy disk (think CD with very little capacity for those who have never dealt with "floppies") and sneaker net, ie walk the floppy drive over to the computers set up for analysis. Standardization was important then, and important now. But now there was a suggestion that perhaps Large Language Models could help with this pain point so I am curious to see how that could work.
Finally, there was also discussion of challenges when standing at the intersection of bio “tech” and “tech”. The investing cultures are not necessarily the same, and the working cultures aren’t either, so young companies need to think about what sector they want to approach for best-fit financing. And how to build the connections between the bio folk and the tech folk within the company for optimal results.
I know I did not capture all that was covered, I suppose if the event had been a zoom meeting and someone had deployed Read.ai or Fireflies.ai then they would have full coverage. But there is a lot happening in the AI x Biotech world, and I was glad to be there in person--I feel lucky to have a front row seat (figuratively, in reality I prefer to blend into the audience!).