README.md

---
license: other
license_name: enzymecage-research-license
license_link: https://huggingface.co/GENTEL-Lab/EnzymeCAGE/blob/main/LICENSE
tags:
  - enzyme-retrieval
  - geometric-deep-learning
  - protein
---

# EnzymeCAGE pretrained checkpoints

Pretrained epoch 19 checkpoints for [EnzymeCAGE](https://github.com/GENTEL-lab/EnzymeCAGE), *A Geometric Foundation Model for Enzyme Retrieval with Evolutionary Insights* (Nature Catalysis, 2026). These are PyTorch state dictionaries for the repository's model code, with seeds 40–44 under `pretrain/`. The repository's demo and evaluation configs use `pretrain/seed_42/epoch_19.pth`.

## Download and run

From the EnzymeCAGE repository root:

```bash
pip install -U huggingface_hub
hf download GENTEL-Lab/EnzymeCAGE pretrain/seed_42/epoch_19.pth --local-dir checkpoints
python infer.py --config config/demo/infer.yaml
```

To download all five seeds instead:

```bash
hf download GENTEL-Lab/EnzymeCAGE --include "pretrain/*/epoch_19.pth" --local-dir checkpoints
```

The dataset and generated features needed for inference are described in the [GitHub README](https://github.com/GENTEL-lab/EnzymeCAGE#dataset). A file list with sizes and SHA-256 checksums is in [`checkpoint_manifest.json`](checkpoint_manifest.json). The original `checkpoints.zip` is also linked from the GitHub README.

## License

See [LICENSE](LICENSE) for the full terms, including restrictions on commercial use and drug development or discovery.