EPIMUTESTR

EPIMUTESTR prioritizes rare cancer driver genes by integrating a nearest-neighbor machine learning approach with phylogenetic Evolutionary Action (EA) scores to distinguish driver coding variants from passenger mutations.

Key Features:

• Nearest-neighbor machine learning: Implements a nearest-neighbor ML approach to prioritize candidate cancer driver genes from mutation data.
• Evolutionary Action (EA) integration: Incorporates phylogenetic Evolutionary Action scores to model the fitness impact of coding variants.
• Modeling evolutionary dynamics: Captures evolutionary dynamics and selective pressures acting on coding variants to separate drivers from passengers.
• Large-cohort application: Applied to The Cancer Genome Atlas cohorts covering 33 tumor types.
• Driver gene discovery: Identified 214 previously inferred cancer driver genes and 137 novel candidate genes, seven of which are supported in the COSMIC Cancer Gene Census.
• Benchmarking performance: Demonstrated improved robustness and specificity relative to existing methods across multiple datasets.

Scientific Applications:

• Rare driver gene identification: Prioritizes low-frequency cancer driver genes that are difficult to detect with conventional methods.
• Cohort-wide cancer genomics analysis: Enables analysis of coding-variant impacts across TCGA's 33 tumor types.
• Novel candidate discovery and validation: Produces novel candidate driver genes for downstream experimental validation and comparison to the COSMIC Cancer Gene Census.
• Method benchmarking: Functions as a comparative method for assessing robustness and specificity in driver gene prediction studies.

Methodology:

EPIMUTESTR employs a nearest-neighbor machine learning algorithm that integrates phylogenetic Evolutionary Action (EA) scores to model fitness impacts of coding variants and distinguish driver from passenger mutations.