wKinMut

wKinMut predicts the pathogenicity of sequence variants in the human kinome using a random-forest classifier to prioritize disease-associated kinase mutations.

Key Features:

• Random-Forest Methodology: Employs a random-forest classifier composed of 26 decision trees to predict pathogenicity of kinase variants.
• Multi-Level Feature Analysis: Evaluates gene-level features (KinBase groups, Gene Ontology terms), PFAM domain-level features, and residue-level features including amino acid properties and annotations from UniProt, Phospho.ELM, and FireDB.
• Performance Metrics: Reports accuracy 0.88, precision 0.82, recall 0.75, F-score 0.78 and Matthews correlation coefficient 0.68 on cross-validation using 3689 human kinase variants annotated as neutral or pathogenic.
• Benchmarking: Benchmarked against SIFT, PolyPhen-2, MutationAssessor, MutationTaster, LRT, CADD, FATHMM, and VEST, with kinase-specific features identified as most informative.
• Independent Validation: Validated on independent kinase-specific mutation sets Kin-Driver (643 variants) and Pon-BTK (1495 variants).

Scientific Applications:

• Variant Classification: Classifies kinase variants as pathogenic or neutral to prioritize disease-associated mutations.
• Mechanistic Insight: Supports investigation of molecular mechanisms linking kinase variants to complex traits and diseases, including cancer.
• Therapeutic Targeting: Aids identification and prioritization of potential therapeutic targets in kinase-driven pathologies.
• Family-Specific Analysis: Enables family-specific analyses to support personalized medicine and targeted drug development.

Methodology:

Uses a random-forest classifier of 26 decision trees with features computed at gene, PFAM domain, and residue levels (including UniProt, Phospho.ELM, FireDB annotations), trained and cross-validated on 3689 human kinase variants, benchmarked against SIFT, PolyPhen-2, MutationAssessor, MutationTaster, LRT, CADD, FATHMM and VEST, and independently validated on Kin-Driver and Pon-BTK.