TargetMM

TargetMM predicts the functional impact of missense mutations to distinguish pathogenic from neutral variants that alter protein function.

Key Features:

• Feature Extraction: Implements a feature extraction method that evaluates the impact degree of residues within the microenvironment range surrounding mutation sites.
• Classifier Ensemble: Employs an ensemble of three heterogeneous models—Support Vector Machine (SVM), Random Forest (RF), and Gaussian Process Classifier (GPC)—for prediction.
• Sequence Feature Integration: Integrates local sequence information including Position-Specific Scoring Matrix (PSSM) and predicted secondary structure with global features such as Amino Acid Pair Composition (APAAC) and Pseudo Amino Acid Composition (PAAC).
• Performance Evaluation: Evaluated using stringent cross-validation and independent testing on benchmark datasets, reporting improvements in Matthews Correlation Coefficient (MCC) relative to SIFT, PROVEAN, FATHMM, and PolyPhen-2.

Scientific Applications:

• Genomics: Prioritizes missense variants for downstream genetic analyses by distinguishing likely pathogenic from neutral substitutions.
• Proteomics: Assesses mutation impacts on protein function using local microenvironment and global sequence-derived features.
• Personalized medicine: Supports interpretation of patient-specific missense variants to inform potential therapeutic strategies.

Methodology:

Extracts residue microenvironment features; computes PSSM and predicted secondary structure and global descriptors (APAAC, PAAC); trains an ensemble of SVM, RF, and GPC on benchmark datasets using stringent cross-validation; evaluates performance with cross-validation and independent testing and reports MCC comparisons to SIFT, PROVEAN, FATHMM, and PolyPhen-2.