Get your free personalized podcast brief

We scan new podcasts and send you the top 5 insights daily.

Shana Kelley says computation can take a more direct route than the random walks of experimental research, but results must go back to the lab for validation. She calls it a virtuous cycle of computation and lab work, not a replacement for the lab when bringing therapies forward.

Related Insights

The explosion of AI-generated molecular candidates creates a new problem: physically testing them at scale. A significant portion of R&D effort should be redirected from AI model development to inventing novel, high-throughput instrumentation and assays to experimentally validate the flood of predictions AI produces.

Most hard genetic cases stall at 'variants of uncertain significance' (VUS). Gamo Labs aims to solve this by creating a feedback loop: AI agents identify a VUS, robotic biology labs conduct experiments to determine its function, and the biological data is used to train and improve the AI models.

Contrary to the belief that AI will replace experimentation, A-Alpha Bio's CEO argues that as models improve, the industry becomes "hungrier" for high-quality wet lab validation data. Better computation creates a greater need for ground-truth data to train, validate, and refine the models.

Future progress in biology requires moving beyond static models. The new paradigm involves an AI that reasons over hypotheses, prioritizes experiments, learns from the empirical outcomes, and updates its internal world model. This creates a scalable, closed-loop system for scientific discovery.

While AI promises to design therapeutics computationally, it doesn't eliminate the need for physical lab work. Even if future models require no training data, their predicted outputs must be experimentally validated. This ensures a continuous, inescapable cycle where high-throughput data generation remains critical for progress.

AI models are trained on large lab-generated datasets. The models then simulate biology and make predictions, which are validated back in the lab. This feedback loop accelerates discovery by replacing random experimental "walks" with a more direct computational route, making research faster and more efficient.

Applying AI to biology isn't just a big data problem. The training data must be structured for reinforcement learning. This means it must be complete (including negative results) and allow for a feedback loop where AI predictions are tested in the lab, and the results are used to refine the model.

Building biologically relevant AI is not a one-off process. It demands a continuous "lab in the loop" system where wet lab experiments generate proprietary data to train models, whose outputs are then physically tested in the lab. This iterative feedback cycle constantly refines the model's predictive accuracy.

Alex Rives expects biology to follow a pretraining-plus-elicitation recipe, but says reinforcement learning must be verified in laboratories. Because biology is empirical, he envisions AI systems proposing real experiments, running them, learning from actual results, and recursively improving their biological models over repeated improvement cycles.

The primary obstacle in advancing protein design isn't creating better models, but the multi-week or multi-month delay in getting experimental validation from wet labs. This slow feedback loop fundamentally constrains the speed of research and model iteration, a problem the entire field is trying to solve.