OncodriveCLUST

OncodriveCLUST identifies genes whose somatic mutations exhibit significant spatial clustering within protein sequences to prioritize potential oncogenes and other cancer driver genes.

Key Features:

• Spatial clustering detection: Detects significant clustering of somatic mutations along protein coordinates to pinpoint putative functional hotspots.
• Background model using coding-silent mutations: Constructs a neutral expectation by analyzing coding-silent mutations presumed not to be under positive selection.
• Statistical comparison to baseline: Compares observed mutation distributions against the coding-silent background to identify genes with a statistically significant bias toward clustered mutations.
• Focus on gain-of-function signals: Targets clusters consistent with gain-of-function mutations that may reflect positive selection during tumor clonal evolution.
• Applicability to large cancer datasets: Applied to and compatible with mutation collections such as COSMIC and TCGA for cohort-level analysis.
• Prioritization capability: Prioritizes Cancer Gene Census candidates and can recover cancer-related genes missed by frequency- or functional-impact-only methods.

Scientific Applications:

• Driver gene discovery: Identification of oncogenes and other cancer drivers by detecting mutation clustering indicative of selection.
• Cohort-level analysis of cancer mutation datasets: Analysis and reanalysis of COSMIC and TCGA mutation data to detect clustered mutation signals.
• Prioritization of Cancer Gene Census entries: Ranking and highlighting genes with clustered mutations, including dominant and some recessive genes with localized clusters.
• Complementary driver detection: Supplementing frequency- and functional-impact-based methods to detect driver genes overlooked by those approaches.

Methodology:

OncodriveCLUST builds a background model from coding-silent mutations and compares observed mutation patterns across protein sequences to that baseline to identify genes with statistically significant clustering.