15-Person Team's NovaAtom-Lite Hits 79% Antibody Prediction, Beating AlphaFold 3's 48%
QuanquanGu · x · 2026-11-11
NovaAtom-Lite is an all-atom structure prediction model built for drug discovery, capable of jointly predicting the structures and assemblies of proteins, nucleic acids, and bound molecules (including small molecules). Officially reported FoldBench antibody–antigen structure prediction results (DockQ ≥ 0.23):
- NovaAtom-Lite: 79.07% (n=172)
- IsoDDE: 75.58%
- AlphaFold 3: 47.90% (n=167)
- 77.06% on the protein–ligand task vs. AlphaFold 3's 64.90%
Evaluation used 5 seeds × 5 samples, with top-1 selected by each model's own ranking score. The model is now available via the Geodesic API and NovaDDE. The roughly 15-person team is regarded as a serious challenger to AlphaFold.
More from Research
- Stanford AI100 releases 2026 report covering the scaling era and generative AI rise — TobyWalsh · 2026-11-11
- Prism capability extraction: specialists carved from Qwen 27B at <20% memory match full model — tokenbender · 2026-11-11
- Data Colada Flags Fraud in Fifth Francesca Gino Paper, No Forensics Needed — danielrock · 2026-11-11
- Fable 5.2 tops GDPx chip-design benchmark, but task authors see zero gains — ycombinator · 2026-11-11
- Fortnow asks Claude about complexity in the Aronson-Ambainis proof: Promise-BQP-complete — fortnow · 2026-11-11
- Backdooring AI agents by reprogramming fewer than 10 dormant neurons — no training needed — DimitrisPapail · 2026-11-11