Prompteina: all-atom protein design from natural language, first validated GM2A de novo binders
ChengleiSi · x · 2026-10-14
A year-long collaboration between UW's Institute for Protein Design and Ai2 introduces Prompteina, which designs proteins directly from natural language. Like the leap from class-label image models to expressive text-to-image models, it natively converts free-form instructions into protein structures, letting a single text prompt control diverse protein properties. It shows higher predicted binding confidence than baselines, reports the first experimentally validated de novo GM2A binders, and works especially well for agent-driven protein design.
Related event: Prompteina Enables Full-Atom Protein Design from Natural Language(2 posts)→
More from Research
- Mathlib Standards Too High for AI Wave, Argues Resolution's Geoffrey Irving — geoffreyirving · 2026-10-14
- Resolution Grants $1M to Mathlib's Formal Commons for the AI Math Wave — geoffreyirving · 2026-10-14
- Resolution Grants $1M to Mathlib Initiative to Prepare for AI-Formalized Math — geoffreyirving · 2026-10-14
- MSAR brings multi-head depth attention to residuals, beating AttnRes and mHC with 1.58x iso-loss speedup — SonglinYang4 · 2026-10-14
- Autoresearch agent eval: hypothesis-distinguishing tests up from 16.8% to 28.0% — anirudhg9119 · 2026-10-14
- MIRA-AC trains research agents on proxy feedback, lifting symbolic regression from 50 to 63 solved — anirudhg9119 · 2026-10-14