Protein-complex benchmark finds MSA depth matters more than correct pairing in AlphaFold3
ferruz_noelia · x · 2026-07-29
A new preprint benchmarks whether paired MSAs are really necessary for protein-protein complex prediction in AlphaFold3.
The authors test 439 non-redundant heterodimers and compare paired MSAs (pMSA) with unpaired concatenated monomer MSAs (mMSA), plus a shuffled-pair control (sMSA) that preserves depth and composition but destroys inter-chain coevolution.
Key finding: the small gain from pairing appears to come mostly from extra sequence depth, not from correct pairing itself. On the full benchmark, pMSA scores 0.613 DockQ vs 0.602 for mMSA, while sMSA matches pMSA almost exactly at 0.612.
They conclude that MSA depth matters more than enforcing pairing, which could simplify practical workflows for protein complex prediction.
More from Research
- DexRobotics open-sources a 50-episode SO101 robot fine-tuning workflow — AdinaYakup · 2026-07-29
- SmartPhotoCrafter lets a model diagnose a photo first, then auto-edit it — 量子位 · 2026-07-29
- A slide argues Bayes is only one lens, not the theory of modern AI inference — davidmanheim · 2026-07-29
- Brain-inspired AI can plan and solve problems with far less energy, study suggests — striketheviol · 2026-07-29
- From GPT-2 to KimiK3, a thread argues the story is bigger than scale — algo_diver · 2026-07-29
- Synthetic training data helps AI resolve messy dataset citations — RexDouglass · 2026-07-29