Chai-1

Overview

Exploratory protein complex modeling.
Protein-ligand, protein-nucleic-acid, glycoprotein, and covalent-adduct hypotheses.
Restraint-informed interface or pocket modeling.

Modes

Mode	Input shape	When to use it
`structure_prediction` Structure Prediction Default	Uses one selected file/source when file parameters are present.	Predict protein, nucleic-acid, ligand, glycan, or mixed-complex structures from one Chai FASTA file.

Canonical Job Configuration

These are the fields exposed by the default job configuration for chai-1. They are also returned by GET /api/v1/program/params?program=chai-1 and submitted as the params JSON object to POST /api/v1/job/submit.

Parameter	Type	Modes	What it does
`target_fasta` Target FASTA	Sequence file	All modes	Choose a Chai FASTA containing protein, DNA, RNA, glycan, and/or ligand SMILES records. Required; Files: .fasta, .fa, .faa, .fas
`use_restraints` Use Restraints	Yes/no	All modes	Guide folding with a Chai restraints table for contacts, pockets, or covalent bonds. Default: false
`restraints_file` Restraints File	Input file	All modes	CSV-format Chai restraints table. The official examples commonly use .restraints files. Required; Files: .csv, .restraints; Shown when use_restraints is true
`prediction_count` Predictions	Integer	All modes	How many candidate structures to sample. Chai-1 defaults to five. Default: 5; Range: 1-5

Advanced configuration fields

Parameter	Type	Modes	What it does
`use_msa` Use Local MSA	Yes/no	All modes	Use a prepared Chai MSA folder containing per-sequence `.aligned.pqt` files. Network MSA services are disabled. Default: false
`msa_directory` MSA Folder	Input folder	All modes	Folder of Chai `.aligned.pqt` MSA files named for the submitted sequences. Required when use_msa is true
`sampling_depth` Sampling Depth	Text	All modes	Fast uses fewer diffusion steps; Thorough spends more steps and recycles. Default: Standard; Options: Fast, Standard, Thorough
`random_seed` Random Seed	Integer	All modes	Set for reproducible sampling; leave blank for a random seed. Range: 1-999999999

Outputs And Metrics

Sampled CIF structures and matching score files.
Ranking scores, pTM, ipTM, per-chain pTM, pairwise interface scores, and clash information.
aggregate_score is the first ranking metric to compare; higher is better.
pTM is overall fold confidence; ipTM is interface confidence.
Local confidence is encoded in structure B-factors on a pLDDT-like 0-100 scale.

Common Examples

Protein-ligand complex: one protein record and one ligand SMILES record in the target FASTA.
Protein-protein complex: include all chains in one target FASTA and sample multiple predictions.
Restraint-informed model: add a CSV with contacts, pocket hints, or covalent links.

Example API params

{
  "mode": "structure_prediction",
  "target_fasta": "complex.fasta",
  "prediction_count": 5,
  "sampling_depth": "Standard"
}

Caveats

Restraints bias prediction but do not guarantee satisfaction.
Residue indices in restraint files are one-based and must match the submitted sequences.
Chemical inputs need correct SMILES, glycan syntax, and atom names for serious interpretation.
Offline template-hit inputs are not exposed because upstream still fetches referenced template structures.

Advanced Submit

Advanced submit is still available for direct program arguments through POST /api/v1/job/submit-advanced. Prefer canonical configuration unless you need exact low-level arguments or are reproducing a known command line.

Advanced submit passes supported Chai v0.6.1 arguments; remote MSA/template-server flags and URLs are removed.
Canonical configuration is preferred for normal FASTA-plus-restraints submissions.

curl -X POST https://subseq.bio/api/v1/job/submit \
  -H "Authorization: Bearer <api_key>" \
  -F program=chai-1 \
  -F 'params={"mode":"structure_prediction","target_fasta":"complex.fasta","prediction_count":5,"sampling_depth":"Standard"}'