Revealing The Functional Cooperation Among Cancer-related Pathways Based On Mutation Profiles Of Cancer Genomes

Thounsands of cancer genes are being rapidly identified by high-throughputscreens of somatic mutations of genes in cancer genomes, providing us abundant datafor systematically deciphering the genetic changes underlying cancer mechanism. Ithas been suggested that mutations in multiple genes that participate in differentpathways are collaborative in conferring growth advantage to cancer cells. However,the functional collaboration of mutated genes is often neglected in current studies.Firstly, using four genome-wide somatic mutation datasets (glioblastoma, breast,colorectal and pancreatic cancer) and pathways defined in various databases, wedemonstrated that gene pairs significantly co-mutated in cancer samples tend todistribute between pathways rather than within pathways. Defining pairs of pathwaysfrequently linked by co-mutated gene pairs as the between-pathway models, weshowed they are also likely to be co-disrupted by mutations of the inter-pathway hubsof the coupled pathways, suggesting new hints for understanding the heterogeneousmechanisms of cancers. We also showed some between-pathway models consisting ofimportant pathways such as cell cycle checkpoint and cell proliferation wereco-disrupted in most cancer samples under this study.Secondly, because of the low statistical power of sparse mutation profile, weproposed a powerful pathway-based approach to study the functional cooperation ofgene mutations in carcinogenesis. We successfully identified many pairs ofsignificantly co-mutated pathways for a large-scale somatic mutation profile of lungadenocarcinoma. We clustered co-mutated pathways into co-mutated super-pathwaysand validated that the derived super-pathways also tend to be significantly co-alteredby DNA copy number alterations. Our results supported the hypothesis that comprehensive cooperation among a few basic functions is required for inducingcancer. The results also suggest biologically plausible models for understanding theheterogeneous mechanisms of cancers. We also suggested an approach to identifycandidate cancer genes from the derived co-mutated pathways.Finally, we proposed a novel analytic tool to search between-module models inthe human protein interaction network, where the modules in the models arefunctionally consistent and co-mutated in cancer. We assumed that the modulesidentified by our approach represent the segments of biological pathways. Onepathway pair, bridged by different pairs of co-mutated modules, is termed asbetween-pathway model. We identified co-function genes frequently interacting withboth pathways in a model (including overlapping genes between pathways of themodel). The co-functional genes also represent a pattern of functional coordinationand their dysfunctions deregulate both pathways, suggesting new hints for theheterogeneous mechanisms of cancers.All together, our proposed approach and derived results provide guidelines todistill the pathway collaboration in carcinogenesis from the complexity of cancersomatic mutation profiles and for identifying candidate cancer genes.