Identifying mutated core modules in glioblastoma by integrative network analysis

Glioblastoma multiforme (GBM) is the most common and aggressive type of brain tumor in humans. Distinguishing “driver” mutations from passively selected “passengers” is a central challenge in computational cancer biology. Because of mutational heterogeneity, analyses that extend beyond single genes are often restricted to examine known pathways and functional modules for enrichment of somatic mutations. In this paper we present a network-based method to identify mutated core modules for tumors without any prior information other than the data of somatic mutations and gene expressions from tumor patients. Firstly, two networks with weighted vertices and weighted edges are constructed by using the mutations and expressions, respectively. Then these two networks are combined to get an integrative network, for which an optimization model is used to identify the most coherent subnetworks. With the significance and exclusivity tests we get the core modules for tumors. By applying our method to The Cancer Genome Atlas (TCGA) GBM data, we obtained three core modules, which contain not only oncogenes and tumor suppressors that have been previously implicated in GBM pathogenesis (e.g., EGFR, TP53, PTEN, NF1 and RB1), but also some genes which have not or rarely been reported earlier in the context of glioblastoma multiforme (e.g., DST, PRAME and SYNE1). Thus, in addition to present generally applicable methodology, our findings provide several GBM candidate genes for further studies.