Nobel Laureate Uses AI to Create Gene Scissors from Scratch, More Precise than the One Evolved Over Billions of Years
Nobel laureate Jennifer Doudna and her team, along with the startup Profluent, have used AI to design novel CRISPR enzymes that do not exist in nature. These AI-generated gene-editing tools demonstrate superior performance compared to their natural counterparts. Doudna's team focused on engineering TnpB, a compact ancestor of Cas12, achieving functional enzymes with sequences approximately 30% different from natural versions—a drastic increase from the typical 1-2% modifications possible with prior methods. Meanwhile, Profluent generated entirely new Cas9 variants from scratch using protein language models. Their leading candidate, OpenCRISPR-1, exhibits higher on-target editing efficiency (55.7% vs. 48.3%) and a roughly 95% reduction in off-target editing compared to the commonly used SpCas9, despite having hundreds of mutations away from any known natural sequence. While these breakthroughs, published in *Science* and highlighted by *Nature*, mark a significant shift from discovering biological tools to actively designing them with AI, the results are currently confined to cellular experiments. The research underscores a future where AI accelerates protein design, but experts emphasize that deep biological insight remains essential to fully harness these models and advance towards therapeutic applications.
marsbit20 giờ trước
