An Effective Graph Clustering Method to Identify Cancer Driver Modules.

Area of Science:

• Genomics
• Computational Biology
• Cancer Research

Background:

• Understanding cancer progression requires identifying molecular drivers.
• Existing methods often overlook crucial module properties, relying heavily on mutual exclusivity.
• Cancer Cell Fraction (CCF) and gene interaction networks offer additional layers of information.

Purpose of the Study:

• To propose MCLCluster, a novel algorithm for identifying cancer driver modules.
• To leverage somatic mutation data, CCF, and gene interaction networks for enhanced driver module detection.
• To improve upon existing methods by incorporating module properties like connectivity and functional similarity.

Main Methods:

• Utilized somatic mutation data and Cancer Cell Fraction (CCF) to select high-confidence mutations.
• Employed a weighted gene functional interaction network to quantify gene functional similarity within protein-protein interaction (PPI) networks.
• Applied a Markov-based clustering algorithm to extract candidate driver modules.

Main Results:

• MCLCluster successfully identified known oncogene driver modules in TCGA GBM and BRCA datasets.
• The algorithm detected novel modules comprising functionally associated driver genes.
• Comparative analysis on simulated datasets demonstrated MCLCluster's superior performance over Multi-Dendrix, FSME Cluster, and RME, especially with background noise and high passenger rates.

Conclusions:

• MCLCluster provides an effective approach to identify cancer driver modules by integrating multiple data types and module properties.
• The method enhances the understanding of cancer mechanisms and offers potential for developing targeted therapies.
• MCLCluster outperforms existing algorithms in identifying driver modules, particularly in complex genomic landscapes.