The Detection of Gene Modules with Overlapping Characteristic via Integrating Multi-omics Data in Six Cancers

Abstract

A large amount of applications on high throughput data are applied for cancer diagnosis, clinical treatment and prognosis prediction. Module network inference is an established and effective method to identify the biomarks for specific cancer and uncover the oncogenesis mechanism. Exploiting the overlapping characteristic between modules, rather than detecting disjoint modules, may broaden our understanding of the molecular dysfunction that govern tumor initiation, progression and maintenance. To this end, we propose a novel framework to identify gene modules with overlapping characteristic by integrating gene expression data and protein-protein interaction data. In our framework, social community and overlapping characteristic are introduced to construct disjoint and overlapping modules which can represent the relationship between diverse modules. Applying this framework on six cancer datasets from The Cancer Genome Atlas, we obtain functional gene modules in each cancer, which are more significantly enriched in the known pathways than identified by other state-of-the-art methods. Meanwhile, those identified modules can significantly distinguish the survival prognostic of patients by Kaplan-Meier analysis, which is critical for cancer therapy. Furthermore, identified driver genes in the network can be considered as biomarkers which can distinguish the tumor and normal samples. Uncovering the overlapping feature in gene modules will help elucidate the relationship between modules and could have important therapeutic implication and a more comprehensive interpretation on carcinogenesis.