PPSC

PPSC predicts changes in protein stability resulting from amino acid substitutions to assess stability effects for protein design, functional annotation, and interpretation of disease-associated variants.

Key Features:

• Structure-Based Coarse-Grained Model: PPSC employs a structure-based coarse-grained model to calculate amino acid contact energy (CE) changes due to point mutations, reducing model complexity by focusing on structural properties and decreasing the number of parameters while enhancing generalization potential.
• Support Vector Machine Classifiers: Two support vector machine (SVM) classifiers, M47 and M8, use contact energy (CE) and additional physicochemical properties of amino acids as input features.
• Performance Metrics: M47 achieves 87% prediction accuracy with a Matthews correlation coefficient (MCC) of 0.68 across a dataset of 1925 variants, and M8 exhibits superior performance on smaller datasets (e.g., 388 variants) when evaluated using 20-fold cross-validation.
• Comparative Advantage: The M47 classifier outperforms existing machine learning–based models and energy function–based protein stability classifiers across all six tested contingency table evaluation parameters.

Scientific Applications:

• Protein Design: Predicts the effects of amino acid substitutions on protein stability to inform design of proteins with desired structural properties.
• Functional Annotation: Assists in assigning biological functions by linking stability changes from substitutions to potential impacts on protein function.
• Disease Research: Provides insights into disease-associated genetic variations by predicting mutation-induced stability changes relevant to pathogenic mechanisms and therapeutic strategies.

Methodology:

Computations use a structure-based coarse-grained model to calculate amino acid contact energy (CE) changes caused by point mutations and two SVM classifiers (M47 and M8) that use CE and additional physicochemical amino acid properties as input, with M8 evaluated using 20-fold cross-validation.