Large-scale cancer genomics projects like The Cancer Genome Atlas have generated a comprehensive catalog of somatic mutations and copy number aberrations across many tumor types, but the role of most frequently altered genes remains obscure. To better model the impact of these alterations, we developed a novel computational strategy for exploiting parallel phosphoproteomics and mRNA sequencing data for large tumor sets by linking dysregulation of upstream signaling pathways with altered transcriptional response through the transcriptional circuitry. Our modeling allows us to interpret the impact of somatic alterations in terms of functional outcomes such as altered signaling and transcription factor (TF) activity.

We used this novel machine learning strategy to train phosphoprotein-TF interaction models across 12 human cancers for which large reverse-phase protein array and RNA-seq data sets are available through TCGA. First, we used this approach to identify shared and cancer-specific roles of TF/signaling regulators across cancer types. Then we performed a statistical analysis to associate frequent somatic aberrations with alterations in inferred TF and signaling protein activities. From our analysis, we gained many novel insights into cancer biology. We identified both known (e.g FOXO1 for breast cancer) and novel TF regulators of cancer (e.g. ELK1 for head and neck cancer). Many of these identified TF regulators were significantly associated with survival outcome in bladder urothelial, renal cell clear and endometrial carcinoma.

Next we performed a comprehensive cross-cancer analysis and identified relationships between somatic alterations and downstream transcriptional effects and signaling pathway activation. We observed that specific molecular aberrations have different functional consequences in different cancer types. For example, PIK3CA mutations are associated with altered activities of a diverse set of TFs across cancers that are involved in cell cycle, apoptosis, metabolism and MAPK/ERK signaling. Notably, in cell line models, we validated some of the altered TFs predicted by our model. We showed that PIK3CA mutation leads to ELK1 activation in breast and head and neck cancer models. We further found that different set of TFs are associated with a specific mutation in different cancers due to the different background of genomic aberrations in each cancer. For example, KRAS mutations are associated with a distinct set of TFs depending of cancer-specific co-mutation profiles (e.g. co-mutations of KRAS and TP53 in lung adenocarcinoma and co-mutations of KRAS and APC in colorectal cancer).

Our analysis revealed both known and novel interactions of frequently altered genes with signaling pathways and transcriptional programs in a pan-cancer context. Patterns of co-alterations across cancers may provide new insights relevant to targeted therapy and may be crucial to optimizing combination therapies.

Citation Format: Hatice U. Osmanbeyoglu, Eneda Toska, José Baselga, Christina Leslie. Modeling the impact of somatic alterations across human cancers. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 781.