MHCflurry documentation

MHCflurry predicts MHC class I binding affinity, antigen processing, and peptide presentation.

Start here

• Introduction and installation installs MHCflurry and makes a first prediction.

• Command-line tutorial scores peptides and scans proteins from the shell.

• Python library tutorial does the same from Python.

Choosing and trusting models

• Downloading and selecting model weights lists the available weights and shows how to select an older release.

• Evaluation of the 2.3.0 weights compares the released weights with other predictors on identical data.

Training your own models

• Training models fits custom models from your own measurements.

• Evaluating trained models compares a trained model with a released one.

Reference

• Command-line reference documents the command-line options.

• API Documentation documents every Python class and method.

• Configuration and performance covers hardware autosizing, environment overrides, and reproducibility.

Advanced topics

• Percentile calibration for all predictors explains custom percentile calibration.

• Auditing training-sample overlap checks whether evaluation samples overlap any compared model’s training data.

Contributors and release maintainers can start with Maintainer workflows.