DrGaP

DrGaP is a computational framework for identifying cancer driver genes and driver signaling pathways from tumor genome-sequencing data by testing whether a gene’s nonsilent mutation rate exceeds an explicitly modeled background (passenger) mutation process. Its statistical models incorporate key biological determinants of somatic mutation patterns, including coding-sequence length, transcript isoforms, mutation-type spectra, heterogeneous background mutation rates, redundancy of the genetic code, and the possibility of multiple mutations per gene, enabling more realistic null expectations than simpler enrichment tests.

A central feature of DrGaP is a likelihood-based inference workflow that addresses the sparse, low-prevalence nature of somatic mutations within individual tumors by using a heuristic strategy to estimate the mixture proportion for the chi-square distribution of likelihood ratio test statistics, improving statistical power relative to standard asymptotic LRT assumptions. Beyond single-gene calls, DrGaP supports pathway-level driver discovery, leveraging aggregation across functionally related genes to increase sensitivity to rare drivers. The methods are packaged with auxiliary bioinformatics utilities and have been demonstrated via simulation and analysis of TCGA sequencing data to achieve high accuracy and sensitivity for driver gene and pathway detection.