Stanford team releases FoldDir, a Dirichlet flow-matching model for protein inverse folding

KevinKaichuang · x · 2026-09-16

A Stanford team (Tartici, Jewett, Altman et al.) released FoldDir, a flow-matching, structure-conditioned protein sequence design model for the inverse folding task — finding amino acid sequences compatible with a desired backbone.

The authors report strong computational benchmark results and several functional nanobody redesigns. The preprint, "Inverse FoldDir: Structure-conditioned Protein Sequence Design by Dirichlet Flow Matching," is on bioRxiv.

Related event: Stanford team releases FoldDir for structure-conditioned protein design(2 posts)→

Original post →

More from Research

Research channel →