PyMut

PyMut: Prediction of Pathological Protein Mutations Using Machine Learning

PyMut predicts pathological mutations in proteins using machine learning models trained on curated mutation datasets. It updates the PMut predictor and supports Mendelian disease mutation classification through standalone training and prediction workflows.

Key Features:

• Standalone Training and Prediction Engine: Trains and validates custom predictive models using user-defined training sets and validation schemas, independent of the PMut web system.
• Pre-calculated Mutation Repository: Provides access to stored predictions for previously analyzed protein mutations.
• Performance Metrics: Default predictor achieves a Matthews Correlation Coefficient (MCC) of 0.61 in 10-fold cross-validation and 0.42 on a blind test using SwissVar 2016 mutations.

Scientific Applications:

• Mendelian Disease Mutation Analysis: Classifies protein variants as pathological or neutral to support genetic studies of inherited disorders.
• Custom Model Development: Enables generation of dataset-specific predictors for targeted mutation analysis.

Methodology:

Applies supervised machine learning algorithms to protein mutation datasets, training predictive models using cross-validation and independent blind testing against benchmark datasets such as SwissVar 2016, with performance quantified by Matthews Correlation Coefficient (MCC).