Abstract 3897: Pharmacogenomic landscape of long non-coding RNAs in human cancers

Abstract

Abstract Extensive efforts in recent large-scale drug screens have enabled identification of protein-coding genes (PCGs) as biomarkers for anti-cancer agent sensitivity, but information on the role of long-non coding RNAs (lncRNAs) in drug response is sparse. Lower abundance, difficulty in expression profiling and limited functional knowledge have so far hindered global analysis of lncRNAs in drug screens. However, lncRNAs represent nearly 60% of the human genome (vs. 3% for PCGs) and contain a vast majority of pharmacogenomic SNPs. Thus, it is critically important to evaluate the role of lncRNAs in drug sensitivity. Here, we perform a comprehensive reanalysis of large-scale drug screens using novel computational methods to create a landscape of pharmacogenomic interactions of lncRNAs in human cancers. We developed an algorithm to accurately impute the lncRNA transcriptome using PCG expression. This approach allowed us to generate lncRNA transcriptome in high throughput drug screening data sets and enabled systematic evaluation of lncRNAs for hundreds of drugs. For example, we applied our method to the Genomics of Drug Sensitivity in Cancer (GDSC) study, where sensitivity to 265 anti-cancer agents were quantified in 963 cancer cell lines. Sparse regression analysis showed lncRNA transcriptome were similar in accuracy at predicting drug response compared to PCGs (P = 0.9), indicating lncRNAs may be equally important biomarkers. Subsequently, we identified drug-specific lncRNA biomarkers using ANOVA while adjusting for tissue type. Our analysis recapitulated literature-verified lncRNAs associated with drug response, e.g. HOTAIR expression with cisplatin resistance (FDR < 0.05). In addition, we identified several compelling novel lncRNA associations for specific drugs along with previously uncharacterized multi-drug response predictors. We next identified lncRNAs that may serve as biomarkers for drug response while controlling for known cancer functional events. As an example, we analyzed EGFR-inhibitors (erlotinib, gefitinib, pelitinib) while controlling for EGFR somatic point mutations and amplification status and identified an anti-sense lncRNA overlapping UGT1A locus that may be a potential regulator of erlotinib metabolism. These findings reveal important lncRNAs that predict response independent of known cancer functional events and are currently being evaluated in non-small cell lung cancer patients. In conclusion, our comprehensive analysis has generated unprecedented insights into the role of lncRNAs in anti-cancer drug response and will be an invaluable resource for researchers studying drug response in various cancer types. Citation Format: Aritro Nath, R. Stephanie Huang. Pharmacogenomic landscape of long non-coding RNAs in human cancers [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 3897.