Inverse FoldDir: Stanford's Dirichlet flow-matching model designs protein sequences from backbone structure
KevinKaichuang · x · 2026-09-16
A Stanford team published the bioRxiv preprint "Inverse FoldDir: Structure-conditioned Protein Sequence Design by Dirichlet Flow Matching" (Alp Tartici, Mihajlo Stojkovic et al., with Michael Jewett, Russ Altman, and Microsoft's Bruce Wittmann).
- Task: protein inverse folding — designing amino acid sequences compatible with a target backbone, a core capability for protein redesign and de novo backbone generation with implications across materials, medicine, and energy.
- Method: Dirichlet flow matching conditioned on structure.
- Results: strong computational benchmarks plus several functional nanobody redesigns.
Related event: Stanford team releases FoldDir for structure-conditioned protein design(2 posts)→
More from Research
- François Fleuret: AI math is to math what numerical integration is to ODEs — francoisfleuret · 2026-09-17
- White paper reframes OpenAI-HuggingFace incident via semantic field mechanics — LopsidedLevel9009 · 2026-09-17
- UIUC Launches Center for Orchestrated Agentic Biology to Study Brain Aging with Multi-Agent AI — dotey · 2026-09-17
- ActionPiece rethinks action tokenization for VLA models, hits 94.8% on LIBERO — DeepCybo · 2026-09-17
- Berkeley's ComPO: zeroth-order preference alignment to mitigate likelihood displacement — Berkeley · 2026-09-17
- SpectralShift extends Gated DeltaNet context windows via spectral reparameterization — RUC-AIBOX · 2026-09-17