Chai-1

Chai-1 is a diffusion-based protein structure prediction model that generates high-quality 3D structures. It supports optional ESM embeddings, MSA server queries, and template-based modeling.

Quick example

from boileroom import Chai1

model = Chai1(backend="modal")
result = model.fold("MKTVRQERLKSIVRI")

print(result.atom_array)  # list of Biotite AtomArray objects
print(result.plddt)       # per-residue confidence scores

Methods

`.fold()`

Predict the 3D structure of one or more protein sequences.

result = model.fold(sequences, options=None)

sequences

str | Sequence[str]

required

A single amino acid sequence string or a list of sequences. Use ":" to separate chains in a multimer (e.g., "CHAIN_A:CHAIN_B"). Chai-1 currently supports a single batch entry.

options

dict | None

default:"None"

Per-call configuration overrides. Only dynamic config keys can be set here — static keys raise ValueError. See Configuration.

Returns: Chai1Output (see Output below)

Output

The Chai1Output dataclass returned by .fold().

Always included

metadata

PredictionMetadata

Prediction metadata with timing information. See PredictionMetadata.

atom_array

list[AtomArray] | None

List of Biotite AtomArray objects. Always generated.

Confidence metrics

pae

list[np.ndarray] | None

Predicted aligned error matrices (one per diffusion sample).

pde

list[np.ndarray] | None

Predicted distance error matrices (one per diffusion sample).

plddt

list[np.ndarray] | None

Per-residue pLDDT scores (one array per diffusion sample).

ptm

list[np.ndarray] | None

Complex pTM scores.

iptm

list[np.ndarray] | None

Interface pTM scores.

per_chain_iptm

list[np.ndarray] | None

Per-chain-pair interface pTM scores.

Structure representations

cif

list[str] | None

mmCIF-formatted structure strings. Only generated when include_fields contains "cif" or "*".

Configuration

These keys can be set via config={} at initialization or options={} per call (unless marked static).

Key	Type	Default	Static	Description
`device`	`str`	`"cuda:0"`	Yes	GPU device identifier
`num_trunk_recycles`	`int`	`3`	No	Number of trunk recycling iterations
`num_diffn_timesteps`	`int`	`200`	No	Number of diffusion timesteps
`num_diffn_samples`	`int`	`5`	No	Number of diffusion samples to generate
`num_trunk_samples`	`int`	`1`	No	Number of trunk samples
`use_esm_embeddings`	`bool`	`False`	No	Whether to use ESM embeddings as input features
`use_msa_server`	`bool`	`False`	No	Whether to query an MSA server for multiple sequence alignments
`use_templates_server`	`bool`	`False`	No	Whether to query a templates server for structural templates
`include_fields`	`list[str] \| None`	`None`	No	Which output fields to return. Use `["cif"]` to include CIF strings. Use `["*"]` for everything

Multimer prediction

Separate chains with ":" in the sequence string:

result = model.fold("MKTVRQERLKSIVRI:LERSKEPVSGAQLAEE")

print(result.atom_array)  # structure with both chains

Chai-1 currently supports only a single batch entry (one sequence or one multimer). If you pass a list of sequences, it must contain exactly one element.

Overview

Folding Models

Embedding Models

Reference

Quick example

Methods

`.fold()`

Output

Always included

Confidence metrics

Structure representations

Configuration

Multimer prediction

Overview

Folding Models

Embedding Models

Reference

​Quick example

​Methods

​.fold()

​Output

​Always included

​Confidence metrics

​Structure representations

​Configuration

​Multimer prediction

Quick example

Methods

`.fold()`

Output

Always included

Confidence metrics

Structure representations

Configuration

Multimer prediction