Atom Level Enzyme Active Site Scaffolding Using RFdiffusion2

This webinar features Jason Yim and Woody Ahern from Valence Labs discussing their groundbreaking research on "Atom level enzyme active site scaffolding using RFdiffusion2." Explore how this new deep generative method overcomes two major limitations in de novo enzyme design: the inability to specify active site geometry at the atomic level and the requirement to pre-specify catalytic residue locations in the sequence. Learn about RFdiffusion2's impressive performance on a benchmark of 41 diverse active sites, successfully building proteins that scaffold all test cases compared to only 16/41 with previous methods. Discover how the team designed enzymes around three diverse catalytic sites and identified active catalysts while testing fewer than 96 sequences. The presentation demonstrates the potential of atomic resolution generative models for designing de novo enzymes directly from reaction mechanisms. Connect with the speakers and access more details through the Valence Labs Portal, the home of the AI for drug discovery community.