Five pharma companies fine-tune OpenFold3 federated, lifting ligand-pose accuracy from 29% to 47%
AllThingsApx · x · 2026-09-15
Five pharmaceutical companies fine-tuned OpenFold3 in a federated setting on 20,000+ structures, raising correct ligand-pose predictions from 29% to 47%. A comment notes that if protein-structure prediction is still in the early fun part of its scaling law, structure-free discovery holds many low-hanging gains — the bottleneck is getting the right data.
More from Research
- T02 system takes first place in VoiceMOS Challenge 2026 Track 2 — kastnerkyle · 2026-09-15
- Jon Barron decodes non-Gaussian 3D scene representation: spheres bound each texel — jon_barron · 2026-09-15
- ImageNet representational similarity is algorithmic, not mechanistic, argues neuroAI thread — ShahabBakht · 2026-09-15
- NeuroAI researcher: backprop conflates evolution, development and lifelong learning — ShahabBakht · 2026-09-15
- k-server Conjecture Proved, Closing a Classic Online Algorithms Open Problem — minilek · 2026-09-15
- Yale physicists re-grade AI benchmarks: most are broken, frontier models near saturation — inductionheads · 2026-09-15