/helix

Primary LanguagePythonApache License 2.0Apache-2.0

logo

✨ Run large protein models in less than 30 seconds with Modal. Open an issue if it takes longer! ✨

PyPI version

Running large models and code that scales for big datasets in this repository is enabled by Modal (no affiliation). It allows us to run code in the cloud on thousands of containers and GPUs without having to think for a second about infrastructure.

Currently implemented

Up Next

  • RFDiffusion
  • EvoProtGrad
  • Evodiff

⚙️ Getting started

  1. Create an account at modal.com.
  2. Open a terminal and run (requires Python 3.10+): pip install helixbio
  3. Run modal token new

🧬 Run your first model

Let's predict a protein structure using ESMFold. This also works in parallel for multiple sequences.

modal run helix.esm::predict_structures_from_fasta --fasta-file "my_lovely_proteins.fasta" --output-dir "my_lovely_structures"

Examples

Generate protein variants using EvoProtGrad.

modal run helix.evoprotgrad::get_evoprotgrad_variants --sequence "MSGKIDKILIVGGGTAGWMAASYLGKALQGTADITLLQAPDIP"  --max-mutations 4 --num-chains 20 --n-steps 200 --output-csv-file "my-variants.csv" --output-fasta-file "my-variants.fasta"

Contributing

We welcome contributions of any size! Below are some good ways to get started.

  • GitHub Discussions: A great way to talk about features you want added or things that are confusing/need clarification.
  • GitHub Issues: These are an excellent way to report bugs. Additionally, you can try and solve an existing issue and submit a PR.

We are actively looking for contributors, no matter your skill level or experience.

License

Helix is open-source and licensed under the Apache License 2.0.