Auto-Research Agent improves AF3 protein interfaces
Original title: Auto-Research Agent on MSA Organization to Improve Protein-Protein Complex Structure Prediction
Agent refines MSA for AlphaFold 3 using biological hypotheses and label-free signals.
Protein-complex structure prediction with AlphaFold 3 depends on MSA content and cross-chain pairing. Uniform recipes fail to optimize target-specific needs. An autonomous agent treats frozen AF3 as black-box, formulating hypotheses, reorganizing MSA, and refining sequence selection using predictor signals. On 1,015 protein-protein interfaces, it improves DockQ scores while using 60% fewer sequences, with largest gains on inter-species heteromers.
- improves DockQ on every benchmark
- uses 60% fewer MSA sequences
- gains largest on inter-species heteromers
- from 0.465 to 0.612 on FoldBench
- from 0.482 to 0.562 on HD439
- Organizations
- Hugging Face
- AI models
- AlphaFold 3
From the original article
1 min read at huggingface.co
Protein-complex structure prediction with AlphaFold 3 depends on MSA content, depth, and cross-chain pairing. Current pipelines apply uniform recipes, typically maximizing depth and pairing by species, yet even whether cross-chain pairing helps AF3 remains contested. We show that the value of both depth and pairing is strongly target-dependent, so no single recipe is optimal across interfaces.
By Zilu Zhang, Jing Bao, Lifeng Qiao · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.