Overview
Chai Discovery provides Chai-1 code and model weights under Apache-2.0. Inputs describe complex components in FASTA, with ligand SMILES supported. The documented runtime requires Linux, Python 3.10+ and a CUDA GPU with bfloat16 support. Default inference uses embeddings without MSAs or templates; optional remote search sends query information to external services. Structural predictions require downstream assessment.
Key Features
- CLI folding from FASTA
- Python inference with optional MSAs, templates and restraints
Use Cases
- Propose protein–ligand complex structures
- Prepare predicted structures for pose checks
How to Use
Follow the repository installation guide; its pinned example uses pip install chai_lab==0.6.1. Run chai-lab fold input.fasta output_folder or use chai_lab.chai1.run_inference. Cache weights and provide local inputs when external sequence searches are inappropriate. Pin the package version.
Sources and verification
This entry describes capabilities documented by the linked project sources. Catalog edits are not evidence of a successful run, compatibility test or scientific performance evaluation.
Read the editorial and verification policy