Generative Design of High-Affinity Peptides Using BindCraft

Abstract

Discovering high-affinity ligands directly from protein structures remains a key challenge in drug discovery. We applied BindCraft, a structure-guided generative modeling platform, to de novo design of peptide ligands for protein interfaces. Originally developed for miniprotein binders, we evaluated its use for shorter peptides (10-20mers) as peptides hold greater synthetic accessibility and therapeutic potential. For the oncoprotein MDM2, BindCraft generated 70 unique peptides; 15 were synthesized, and 7 showed specific binding with nanomolar affinities (KD = 65-165 nM). Competition assays confirmed site-specific binding for the intended target site. For another oncology target, WDR5, peptides were designed for the MLL (WIN) and MYC (WBM) sites. Of the peptides tested for each site, no validated hits were found for the WIN site, but six candidates bound the WBM site with sub-micromolar affinity (KD = 219-650 nM). Based on Bindcrafts structural prediction of the binding interface, we designed a stapled variant of the best WDR5 binder, improving the potency by 6-fold to a KD of 39 nM. Overall, our findings establish BindCraft as a powerful and accessible platform for structure-based peptide discovery, with remaining limitations, but with a promising success rate even for challenging targets.