mutedpy is a Python library for analyzing mutational datasets of protein sequences and structures. It provides a suite of machine learning models and tools for datasets ranging from $10^2$ to $10^6$ samples, with a focus on both shallow and deep learning approaches.

If you use mutedpy in your research, please cite our paper:

• Vornholt & Mutny (2024)

• Neural Networks: Feed-forward, convolutional, and graph-based architectures
• Linear Models
• Gaussian Processes:
  • Multiple similarity metrics
  • Amino acid embeddings
  • ESM and data-driven embeddings
  • Geometric features
• Directed Evolution Simulation: Tools for simulating directed evolution campaigns
• Sequence Dataset Manipulation:
  • Format conversion
  • Scanning and enumeration search
  • Train/test data splitting

Clone the repository:

git clone https://github.com/Mojusko/mutedpy

Install in editable mode:

pip install -e .

The -e flag installs the package in "editable" mode, so updates to your local copy are immediately reflected. Requires Python 3.6+.

// ... Add a brief description of the main folders and their purpose if desired ...

• 21/05/2024 – initial commit
• 24/08/2025 - project structure improved

Classical:

• pytorch, cvxpy, numpy, scipy, sklearn, pymanopt, mosek, pandas

Special:

1. pytorch-minimize
2. stpy

We use pytest for testing. To run the test suite and check coverage:

pytest --cov=mutedpy tests/

Test coverage reports are generated using the pytest-cov plugin. Please ensure new contributions are covered by tests.

Contributions are welcome! Please open an issue or pull request.

Author: Mojmir Mutny