A Graph-Based Approach for Prioritizing Sets of Related Cancer Genes

Abstract Purpose: The identification of cancer-related genes with significant mutations is critical for deciphering the underlying mechanisms of tumor initiation and progression. Because of the infinite number of genes that are mutated at a low frequency, this is often a critical task in large-scale genomic analysis. To identify infrequently mutated genes, gene interaction networks have been combined with mutation data. Here, we introduce GBP-PR (Graph-Based Prioritization with PageRank), an efficient computational approach for prioritizing cancer-related genes. Methods: GBP-PR assigns a mutation score to each gene based on the type of mutation.Then the mutation neighbor influence of each gene received from their neighbors in the network is calculated via the asymmetric spreading strength computed from the consensus gene interaction network. To generate a set of the prioritized potential cancer genes, GBP-PR applies a PageRank algorithm with a gene-specific dynamic damping. Results: The experimental results with six types of cancer indicate the potential of GBP-PR to discover known and possible new significant cancer genes. Evaluation matrices with six types of cancer indicate that GBP-PR performs better when integrated with PageRank Algorithm compared with other rating algorithms (GBP-Keener, GBP-Colley, and GBP-Massey)