Identification of rare cancer driver mutations by network reconstruction

Abstract

Recent large-scale tumor resequencing studies have identified a number of mutations that might be involved in tumorigenesis. Analysis of the frequency of specific mutations across different tumors has been able to identify some, but not all of the mutated genes that contribute to tumor initiation and progression. One reason for this is that other functionally important genes are likely to be mutated more rarely and only in specific contexts. Thus, for example, mutation in one member of a collection of functionally related genes may result in the same net effect, and/or mutations in certain genes may be observed less frequently if they play functional roles in later stages of tumor development, such as metastasis. We modified and applied a network reconstruction and coexpression module identification-based approach to identify functionally related gene modules targeted by somatic mutations in cancer. This method was applied to available breast cancer, colorectal cancer, and glioblastoma sequence data, and identified Wnt/TGF-beta cross-talk, Wnt/VEGF signaling, and MAPK/focal adhesion kinase pathways as targets of rare driver mutations in breast, colorectal cancer, and glioblastoma, respectively. These mutations do not appear to alter genes that play a central role in these pathways, but rather contribute to a more refined shaping or “tuning” of the functioning of these pathways in such a way as to result in the inhibition of their tumor-suppressive signaling arms, and thereby conserve or enhance tumor-promoting processes.