How a Machine Learns to Pick the Next Batch of Genes in a CRISPR Screen
Biologists testing thousands of genes often face an expensive question: which untested genes should go into the next batch? A new machine-learning approach combines a general-purpose language model with a specialized policy that learns from past gene hits, screen descriptions, and round-by-round lab feedback. The goal is to turn noisy experimental results into better guesses about what to test next, while keeping choices grounded in known biology.
