tools
Chai-1
Chai-1 predicts molecular structures and interactions across proteins, small molecules, DNA, RNA, and covalent modifications.
Chai-1 is a multimodal foundation model for molecular structure prediction. It can predict proteins, small molecules, DNA, RNA, covalent modifications, and multimolecular interactions, including from single sequences without multiple sequence alignments.
Researchers and drug-discovery teams can use it through Chai Discovery's web interface or run the released software library and model weights. The web interface is free, while the official site does not list paid tiers or commercial pricing. Chai Discovery has since introduced newer model series, including Chai-2.
Features
- Predicts structures for proteins, small molecules, DNA, and RNA
- Models covalent modifications and other molecular components
- Predicts multimer structures from single sequences without MSAs
- Supports epitope conditioning and laboratory-derived restraints
- Provides a web interface for model access
- Releases model weights and inference code
- Software library released under the Apache 2.0 License
Use cases
- Predict protein and biomolecular structures for research
- Model antibody–antigen interactions and epitope contacts
- Explore molecular interactions relevant to drug discovery
- Run structure prediction without multiple sequence alignment searches
- Test predictions using laboratory-derived molecular restraints