Identification of Cancer Driver Modules Based on Graph Clustering from Multiomics Data

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Identifying cancer driver genes and modules is a significant challenge in cancer genomics.
• Current methods often focus solely on somatic mutation patterns (mutual exclusivity, high coverage) and neglect multi-omics data.
• Integrating diverse biological information is essential for robust driver module identification.

Purpose of the Study:

• To develop a novel computational approach for identifying cancer driver modules (iCDM).
• To integrate mutual exclusivity, coverage, and protein-protein interaction data for improved module detection.
• To evaluate the performance of the proposed method against existing advanced computational techniques.

Main Methods:

• Construction of an edge-weighted network by integrating mutual exclusivity, coverage, and protein-protein interaction data.
• Application of a graph clustering approach based on symmetric non-negative matrix factorization for iCDM.
• Testing the iCDM method on pan-cancer datasets and comparing results with other computational methods.

Main Results:

• The proposed iCDM approach demonstrated superior performance in recovering known cancer driver modules compared to other methods.
• Identified driver modules exhibited high accuracy in classifying normal and tumor samples.
• The integration of multi-omics data significantly enhanced the identification of functional cancer gene modules.

Conclusions:

• The developed graph clustering approach effectively identifies cancer driver modules by integrating diverse biological data.
• This method offers a more accurate and robust way to discover cancer drivers compared to traditional mutation-based approaches.
• The findings have implications for understanding cancer mechanisms and developing targeted therapies.