Improving de novo protein binder design with deep learning
importance 7/10
partially solved
Protein Structure and Dynamics protein design stage: research open-source model
Problem
De novo design of high-affinity protein binders from target structure alone had a low overall success rate.
Why it matters
Protein binders are key to developing therapeutics and biotechnology tools.
What the AI did
AlphaFold2 or RoseTTAFold assessed whether designed sequences adopt the designed structure and bind the target; ProteinMPNN replaced Rosetta for sequence design.
Results
Design success rates increased nearly 10-fold; ProteinMPNN considerably increased computational efficiency over Rosetta.
Impact
Deep learning assessment markedly raised protein binder design success rates.
Record
| Venue | Nature Communications |
|---|---|
| Published | 2023 |
| Work type | article |
| Authors | Nathaniel R. Bennett, Brian Coventry, Inna Goreshnik, Buwei Huang, Aza Allen, Dionne K. Vafeados, Ying Po Peng, Justas Dauparas et al. |
| Citations | 396 (OpenAlex, at import time) |
| References | 44 |
| Model or method | AlphaFold2, RoseTTAFold, ProteinMPNN |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41467-023-38328-5 |
| Open access | gold |
| Keywords | protein binder design, deep learning, AlphaFold2, ProteinMPNN, de novo design |
Importance rating
7
significant — thousands to millions affected
Protein binder design is foundational to therapeutics and biotechnology; abstract shows 10-fold success rate improvement but states no scale of lives affected.