Convergent architecture of the transcriptome of human cancer

ABSTRACTSDespite large-scale efforts to systematically map the cancer genome, little is known about how the interplay of genetic and epigenetic alternations shapes the architecture of the tumors transcriptome. With the goal of constructing a system-level view of the deregulated pathways in cancer cells, we systematically investigated the functional organization of the transcriptomes of 10 tumor types using data sets generated by The Cancer Genome Atlas project (TCGA). Our analysis indicates that the human cancer transcriptome is organized into well-conserved modules of co-expressed genes. In particular, our analysis identified a set of conserved gene modules with distinct cancer hallmark themes involving cell cycle regulation, angiogenesis, innate and adaptive immune response, differentiation, metabolism and regulation of protein phosphorylation. We applied a network inference approach to nominate candidate drivers of these conserved gene modules. The predicted drivers have consistent cancer-relevant functions related to the specific hallmarks and are enriched with cancer consensus genes and significantly mutated genes. We showed genetic alternations of TP53 and other cell cycle drivers have major downstream transcriptional impact on cell cycle regulation. Collectively, our analysis provided global views of convergent transcriptome architecture of human cancer. The result of our analysis can serve as a foundation to link diverse genomic alternations to common transcriptomic features in human cancer.