The unique complement of genes expressed in a given cell type modifies the response to a specific somatic mutation. Thus, each tumor type would present with a different set of mutations that modulate the activity of common factors regulating genes which influence the cancer phenotypes (e.g. the hallmarks of cancer). We assessed the validity of this hypothesis by applying the SYstems Genetics Network AnaLysis (SYGNAL) pipeline to the ~10,000 multi-omic profiles across 33 tumor types from The Cancer Genome Atlas (TCGA). The SYGNAL pipeline synthesizes correlative approaches with mechanistic knowledge to construct a gene regulatory network. A basis of the SYGNAL approach is the discovery of biclusters: gene modules that are both co-expressed and co-regulated over subsets of tumors, by specific transcription factors (TFs) and miRNAs. On top of the biclusters we discover potential mechanistic influences of somatic mutations modulating the activity of these regulators. We discovered 284 somatically mutated genes modulating the activity of 187 transcriptional regulators (175 TFs and 12 miRNAs) which regulated 7,786 significantly co-expressed biclusters that were significantly associated with patient survival. The discovery of 1.5 fold more somatic mutations than transcriptional regulators is consistent with our hypothesis that mutations will be unique to tumor types while the regulation of the genes involved in the hallmarks of cancer will be common across tumor types. We also find that only 15% of causal flows (mutation->regulator->gene) are observed across multiple tumor types. We observed only four causal flows across three tumor types (NFE2L2->BCL6B->HSPG2 for BLCA, ESCA, and HNSC tumor types; TP53->ERG->MTX1 for BRCA, LUAD, and STAD tumor types; TP53->ZNF511->NAA38 for LUAD, PRAD, and READ tumor types; and TP53->AEBP2->THOC6 for COAD, KICH, and READ tumor types). There were no causal flows observed across more than three tumor types. Using our top-down systems approach we are able to catalog the impact of somatic mutations on the transcriptional regulatory network across 33 tumor types. We are working to replicate the co-expression and survival relationships we discovered in independent cohorts for each tumor type. In addition, we are working to identify potential therapeutic targets at both the regulator and gene level. Future studies will use predictive models based on our detailed transcriptional regulatory networks to identify transcriptional regulators driving cancer phenotypes that won't have general toxicity in normal cells.
Citation Format: Christopher L. Plaisier. Genomic drivers differ across 33 tumor types, but factors regulating hallmarks of cancer are similar [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 1471.