Abstract A37: Cross-cancer evaluation of MYC synthetic lethal targets in The Cancer Genome Atlas

Abstract

Abstract Overview: Synthetic lethality in cancer has been investigated for the last decade, but only recently do we have the convergence of high throughput RNA interference screening capabilities with large genomic and molecular data sets derived from primary tumor samples that allows for more comprehensive characterization of putative therapeutic targets against multiple cancers. Accurately relating cancer cell line RNA interference synthetic lethal screening data to primary human tumors has proven to be challenging, yet critically important, for improving validation rates and therapeutic efficacy. To this end, we have combined siRNA synthetic lethal screening hits in MYC amplified cell lines [Toyoshima et al. PNAS 2012] with genomic and molecular data from twelve tumor types in The Cancer Genome Atlas (TCGA) to identify and prioritize cross-cancer therapeutic targets with direct evidence for synthetic lethal relationships in both cell line and primary tumor samples. Approach: In order to relate TCGA data to synthetic lethal siRNA screening in cell lines, we began by integrating all data types produced by TCGA, including: clinical data, sample characteristics, somatic mutations, gene expression, DNA methylation, miRNA expression, reverse-phase protein array data, and somatic copy number variation. From these data, feature matrices have been constructed which combine all available information regarding patients and samples for each given tumor type. For each such heterogeneous feature matrix, we have performed statistical analysis to infer associations in the data between MYC amplification and all other molecular features. Hits from MYC synthetic lethal cell line screening were evaluated for statistical significance in pairwise analysis of twelve TCGA tumor types, classes of pairwise data-derived associations as well as properties of network-level topologies were characterized, and relationships between genomic data and clinical features were investigated where possible. Results: Evidence for synthetic lethal relationships between MYC amplification and molecular features (e.g., gene expression and DNA methylation) for numerous experimentally validates hits was present across multiple TCGA tumor types. For example, statistically significant associations between MYC amplification and decreased CSNK1E methylation, indicative of increased expression, were present in head and neck squamous cell carcinoma, cutaneous melanoma, and stomach adenocarcinoma, corroborating the experimental findings of Toyoshima et al. Similarly, PES1 gene expression is increased in breast carcinoma, colon adenocarcinoma, and head and neck squamous cell carcinoma samples with MYC amplification. Synthetic lethal hits had differential expression profiles reflective of both increases and decreases in correlation with MYC amplification. Based on these preliminary findings, the intersection of siRNA cell line screening hits with accompanying statistical associations in primary tumor sample data may enable prioritization of preclinical models of MYC driven cancers. While the present approach has been applied to MYC amplification, synthetic lethal relationships are present for a multitude of significantly mutated, copy number aberrated, or epigenetically silenced genes in TCGA. Citation Format: Brady Bernard, Silvia Cermelli, Carla Grandori, Ilya Shmulevich. Cross-cancer evaluation of MYC synthetic lethal targets in The Cancer Genome Atlas. [abstract]. In: Proceedings of the AACR Precision Medicine Series: Synthetic Lethal Approaches to Cancer Vulnerabilities; May 17-20, 2013; Bellevue, WA. Philadelphia (PA): AACR; Mol Cancer Ther 2013;12(5 Suppl):Abstract nr A37.