Metadata-Version: 2.1
Name: boltz
Version: 0.1.0
Summary: Boltz-1
Requires-Python: >=3.9
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: torch>=2.2
Requires-Dist: numpy==1.26.3
Requires-Dist: hydra-core==1.3.2
Requires-Dist: pytorch-lightning==2.4.0
Requires-Dist: rdkit==2024.3.6
Requires-Dist: dm-tree==0.1.8
Requires-Dist: requests==2.32.3
Requires-Dist: pandas==2.2.3
Requires-Dist: types-requests
Requires-Dist: einops==0.8.0
Requires-Dist: einx==0.3.0
Requires-Dist: fairscale==0.4.13
Requires-Dist: mashumaro==3.14
Requires-Dist: modelcif==1.2
Requires-Dist: wandb==0.18.7
Requires-Dist: click==8.1.7
Requires-Dist: pyyaml==6.0.2
Requires-Dist: biopython==1.84
Requires-Dist: scipy==1.13.1
Provides-Extra: lint
Requires-Dist: ruff; extra == "lint"

<h1 align="center">Boltz-1:

Democratizing Biomolecular Interaction Modeling
</h1>

![](docs/boltz1_pred_figure.png)

Boltz-1 is an open-source model which predicts the 3D structure of proteins, rna, dna and small molecules; it handles modified residues, covalent ligands and glycans, as well as condition the generation on pocket residues. 

For more information about the model, see our [technical report](https://gcorso.github.io/assets/boltz1.pdf).

## Installation
Install boltz with PyPI (recommended):

```
pip install boltz
```

or directly from GitHub for daily updates:

```
git clone https://github.com/jwohlwend/boltz.git
cd boltz; pip install -e .
```
> Note: we recommend installing boltz in a fresh python environment

## Inference

You can run inference using Boltz-1 with:

```
boltz predict input_path
```

Boltz currently accepts three input formats:

1. Fasta file, for most use cases

2. A comprehensive YAML schema, for more complex use cases

3. A directory containing files of the above formats, for batched processing

To see all available options: `boltz predict --help` and for more informaton on these input formats, see our [prediction instructions](docs/prediction.md).

## Training

If you're interested in retraining the model, see our [training instructions](docs/training.md).

## Contributing

We welcome external contributions and are eager to engage with the community. Connect with us on our [Slack channel](https://boltz-community.slack.com/archives/C0818M6DWH2) to discuss advancements, share insights, and foster collaboration around Boltz-1.

## Coming very soon

- [ ] Pocket conditioning support
- [ ] More examples
- [ ] Full data processing pipeline
- [ ] Colab notebook for inference
- [ ] Confidence model checkpoint
- [ ] Support for custom paired MSA
- [ ] Kernel integration

## License

Our model and code are released under MIT License, and can be freely used for both academic and commercial purposes.
