Language, software code, DNA, etc.—all seem different, but LLMs equally help reveal the shared logic beneath.
This week I was privileged to meet with a few impressive biotech startups that are using advanced generative AI technologies to accelerate the discovery of complex cellular and genetic mechanisms. DeepLife's computational models of cellular systems and WhiteLab Genomics's algorithmic approach to gene therapy design both point to the same transformation: biology is increasingly being understood and shaped through the lens of data and algorithms. Thanks Jonathan Baptista and Timothé Cynober for showcasing your team's work, and congratulations for leading the way in this promising field.
One thing that struck me most is how transfer learning enables these AI models to process DNA—as if it were just another form of grammar—spotting hidden patterns in cell DNA or clustering cellular datasets to predict causal relationships with drugs.
All of this opens up great perspectives for life sciences, while also requiring us to proceed with a steady mind—and sometimes, a trembling hand—when it comes to the ability to read and even rewrite the "code of life." There's something deeply fascinating about the mathematical patterns that underpin both our ways of speaking and the workings of living systems, which echoes a thought-provoking idea I love, taken from Bruno Brunor's series of insightful graphic novels: "Chance doesn't write messages."
