mmCSM-PPI

mmCSM-PPI predicts the impact of single and multiple missense mutations on protein-protein interactions (PPIs) by estimating mutation-induced changes in binding affinity to support analysis of molecular mechanisms and disease-associated disruptions.

Key Features:

• Scalability and effectiveness: Employs a scalable machine learning model to assess changes in protein-protein binding affinity caused by single and multiple missense mutations.
• Expanded graph-based signatures: Uses expanded graph-based signatures to capture the physicochemical and geometrical properties of multiple wild-type residue environments.
• Integration with substitution scores and dynamics terms: Incorporates substitution scores and dynamics terms derived from normal mode analysis for comprehensive assessment of mutation effects.
• Performance metrics: Achieves Pearson's correlation up to 0.75 (RMSE 1.64 kcal/mol) under 10-fold cross-validation and Pearson's correlation 0.70 (RMSE 2.06 kcal/mol) on a non-redundant blind test set, outperforming existing methods.

Scientific Applications:

• Disease mechanism analysis: Predicts how missense mutations affect PPI binding affinities to aid study of molecular bases of diseases linked to disrupted PPIs.
• Therapeutic target identification: Supports identification of potential therapeutic targets by indicating mutation-induced changes in interaction strength.
• Complex genetic studies: Evaluates effects of multiple concurrent missense mutations to inform analyses in complex genetic contexts.

Methodology:

Expanded graph-based signatures, machine learning models, and dynamics terms from normal mode analysis are used to predict changes in protein-protein binding affinity.