Cooperative driver pathways discovery by multiplex network embedding.

Area of Science:

• Bioinformatics
• Computational Biology
• Cancer Research

Background:

• Cancer pathogenesis research often relies on genomics data, neglecting pathway information and multi-omics data.
• Existing methods struggle to fully decipher the carcinogenic process due to limited data integration.

Purpose of the Study:

• To propose CDPMiner (Cooperative Driver Pathways Miner) for discovering cooperative driver pathways.
• To jointly model relational and attribute information from multi-type molecules using multiplex network embedding.

Main Methods:

• CDPMiner quantifies gene weights using pathway topology and optimizes gene-pathway relations.
• An attributed multiplex network of RNAs, genes, and pathways is constructed and embedded using deep joint matrix factorization.
• Pathway interaction networks are reconstructed to define pathway driver weights for cooperative pathway discovery.

Main Results:

• CDPMiner effectively fuses multi-omics data to identify more driver pathways in cancer.
• The discovered pathways cooperatively trigger cancers and are valuable for carcinogenesis analysis.
• Experimental results on Breast invasive carcinoma and Stomach adenocarcinoma datasets validate the method's efficacy.

Conclusions:

• CDPMiner provides a comprehensive approach to cancer analysis by effectively fusing multi-omics data.
• The method enhances the understanding of cancer pathogenesis through cooperative driver pathway discovery.
• The findings highlight the importance of integrating diverse molecular data for robust cancer research.