DriverRWH

DriverRWH identifies cancer driver genes by applying a random walk on a weighted gene mutation hypergraph that integrates somatic mutation data and molecular interaction network data to prioritize candidate driver genes.

Key Features:

• Random Walk Algorithm: DriverRWH employs a random walk algorithm on a weighted gene mutation hypergraph to prioritize candidate driver genes by integrating somatic mutation and molecular interaction network data.
• Co-mutation Utilization: It leverages co-occurrence mutation information from individual samples to capture sample-specific mutation patterns and reduce false positives.
• Data Integration: The method integrates somatic mutation data with molecular interaction and gene functional network data for comprehensive gene prioritization.
• Performance on TCGA: When applied to tumor samples from The Cancer Genome Atlas (TCGA) across various cancer types, DriverRWH achieves higher area under the curve (AUC) scores and recovers a greater cumulative number of known driver genes among top-ranked candidates compared to state-of-the-art methods.
• Robustness and Versatility: The approach exhibits robustness to perturbations in mutation data and gene functional network data and maintains effectiveness across different cancer types.
• Discovery of Potential Drivers: DriverRWH has identified potential driver genes that are enriched in cancer-related pathways.

Scientific Applications:

• Driver Discovery: Identification of potential cancer driver genes enriched in cancer-related pathways for biological interpretation.
• Pan-cancer Analysis: Application to TCGA tumor samples across multiple cancer types to recover known driver genes and benchmark performance against other methods.
• Candidate Prioritization: Prioritization of candidate driver genes for downstream experimental validation and therapeutic exploration.

Methodology:

Random walk on a weighted gene mutation hypergraph; integration of somatic mutation data with molecular interaction and gene functional network data; use of co-occurrence mutation information from individual samples; evaluation using area under the curve (AUC) and cumulative recovery of known driver genes on TCGA tumor samples; robustness testing via perturbations to mutation and network data.