News

Deep-learning pipeline RFpeptides designs high-affinity macrocyclic peptide binders from scratch

Technical 92/100
June 20, 2025↗ source

A denoising diffusion model from David Baker's Institute for Protein Design generates macrocyclic peptide binders against arbitrary protein targets. Tested on four targets, it produced nanomolar-affinity binders from fewer than 20 designs each, with crystal structures matching the computational models to within ~1.5 Å. Published in Nature Chemical Biology, the work signals that AI can now yield drug-like cyclic peptides on demand.

Every claim on this page is sourced. Reliability scores are tier-capped and server-normalized (0–100; green ≥70, amber ≥40, red <40). Drill through to the source page to verify.