PCA-MutPred

PCA-MutPred predicts changes in binding free energy (ΔΔG) caused by missense mutations in protein-carbohydrate complexes to assess mutation effects on binding affinity.

Key Features:

• Data-Driven Approach: Uses an experimental dataset comprising 318 unique mutants to correlate mutation-site properties with ΔΔG.
• Feature Analysis: Considers accessible surface area, secondary structure, mutation preference, conservation score, hydrophobicity, and contact energies as predictors of binding affinity changes.
• Predictive Modeling: Employs multiple regression equations to predict ΔΔG, achieving an average correlation coefficient of 0.74 and a mean absolute error (MAE) of 0.70 kcal/mol in 10-fold cross-validation.
• Independent Validation: Validated on an independent test set of 124 mutations across 62 unique sites, yielding a correlation coefficient of 0.79 and an MAE of 0.56 kcal/mol.

Scientific Applications:

• Protein Engineering: Guides design efforts by predicting how missense mutations alter protein–carbohydrate binding affinities.
• Disease Research: Identifies mutations that may disrupt protein–carbohydrate interactions and contribute to altered function or disease.
• Drug Design: Informs development of therapeutics targeting specific protein–carbohydrate interactions by predicting mutation-driven affinity changes.

Methodology:

Uses an experimental dataset of 318 mutants; computes features (accessible surface area, secondary structure, mutation preference, conservation score, hydrophobicity, contact energies); builds multiple regression models; evaluates performance with 10-fold cross-validation and an independent test set of 124 mutations across 62 sites.