MPA-MutPred

MPA-MutPred predicts changes in binding affinity (ΔΔG) of membrane protein-protein complexes upon mutation using sequence and structure-based features.

Key Features:

• Membrane Protein-Specific Modeling: MPA-MutPred is specifically developed for predicting ΔΔG in membrane protein-protein complexes, addressing a gap where existing methods target soluble protein complexes.
• Hybrid Linear-Nonlinear Approach: The method combines multiple linear regression for feature selection with gradient boosting regressor for final prediction to obtain a more robust model with improved accuracy.
• Sequence and Structure-Based Features: The model incorporates changes in stability of the complex, conservation score, electrostatic interaction, relatively accessible surface area, and interface contacts.
• Performance Metrics: The method achieved a correlation of 0.75 and mean absolute error (MAE) of 0.73 kcal/mol in jack-knife test on 770 mutants, and correlation of 0.85 with MAE of 0.77 kcal/mol on a blind test set of 86 mutations.

Scientific Applications:

• Disease Mechanism Studies: MPA-MutPred enables large-scale studies to understand the influence of binding affinity change on disease-causing mutations in membrane protein complexes.
• Therapeutic Target Identification: The method aids in identifying potential therapeutic targets by elucidating how interface mutations in membrane protein complexes contribute to human diseases through altered binding affinities.

Methodology:

MPA-MutPred uses multiple linear regression to extract informative features that influence binding affinity in membrane protein complexes, followed by gradient boosting regressor on the selected features to predict ΔΔG values.