GoKawiilNew AI method predicts protein conformational changes without lengthy simulations
Researchers describe a generative sampling approach that models transitions between molecular conformations using diffusion-based machine learning, bypassing the extremely long molecular dynamics runs traditionally needed to capture rare structural changes. The technique draws on diffusion probabilistic models and path-sampling theory to generate plausible transition pathways directly, rather than waiting for these events to occur naturally in simulation.