Machine learning on biomolecules

Bioverse is a standardized framework for machine-learning experiments on biomolecules — proteins, RNA, small molecules, and related structures. It combines reusable benchmarks, transforms, and evaluation components with a config-driven CLI so you can train and evaluate models without reimplementing data loading, splitting, or metrics.

Quick start

pip install bioverse-ml
pip install torch lightning   # required for training
bioverse train experiment.yaml

The default data directory is ~/.bioverse. Override paths with environment variables such as BIOVERSE_ROOT and BIOVERSE_DATASET_ROOT.

Documentation

Full documentation is hosted on GitHub Pages:

https://borgwardtlab.github.io/bioverse/

• Quickstart — minimal B_AFCATH loader example
• User Guide — configuration and CLI workflows
• Developer Guide — architecture and extension points
• Implementations — auto-generated catalog of datasets, benchmarks, and components
• Citation

Build docs locally:

pip install -e ".[docs]"
cd docs && make html
python -m http.server --directory build/html 8080

Pipeline

Adapter  →  Dataset  →  Benchmark  →  Trainer + Model
                          (sampler,
                           task,
                           metric)

Citation

If you use Bioverse in your research, please cite:

@software{bioverse2026,
  author = {Kucera, Tim and Bioverse Contributors},
  title = {Bioverse: A standardized framework for machine learning on biomolecules},
  year = {2026},
  url = {https://github.com/BorgwardtLab/bioverse}
}

See the citation page for details.

License

BSD-3-Clause — see LICENSE.