Auto-Research Agent on MSA Organization to Improve Protein-Protein Complex Structure Prediction
The agent improved AlphaFold 3’s interface predictions while using an average of 60% fewer MSA sequences.
The authors tested it on 1,015 protein–protein interfaces across three benchmarks released after AF3’s training cutoff, reporting higher overall DockQ scores on each. The agent leaves AF3 frozen and adjusts sequence selection and cross-chain pairing for each target using label-free predictor signals. Gains were largest for inter-species heteromers: mean DockQ rose from 0.465 to 0.612 on FoldBench and from 0.482 to 0.562 on HD439. Effective choices included deliberately leaving sequences unpaired.
HF Daily Papers' note
The authors tested it on 1,015 protein–protein interfaces across three benchmarks released after AF3’s training cutoff, reporting higher overall DockQ scores on each. The agent leaves AF3 frozen and adjusts sequence selection and cross-chain pairing for each target using label-free predictor signals. Gains were largest for inter-species heteromers: mean DockQ rose from 0.465 to 0.612 on FoldBench and from 0.482 to 0.562 on HD439. Effective choices included deliberately leaving sequences unpaired.
HF Daily Papers' note
score 5