Abstract 1423: Integrative analysis predicts lncRNA regulating gene alternative splicing in breast cancer

Abstract

Abstract Background: Non-coding region occupies 98% of the whole human genome and plays a regulatory role for protein-coding genes. About 95% of the protein-coding genes undergo alternative splicing, however, limited understandings for the complexity of alternative splicing (AS) mechanisms. Recent studies reported that some aberrant alternative splicing are associated with cancer development. Cancer-specific AS could be regulated by long non- coding RNAs (lncRNAs) directly or indirectly through other intermediate molecules . LncRNAs regulate gene splicing either by binding to their splicing factor proteins or competing with miRNAs to influence their targeted genes. Therefore, it is critically important to develop an effective computational framework capturing the lncRNAs regulatory mechanism in the process of AS from heterogeneous molecular relationships. Method: To model the complicated regulation mechanism in splicing event, we developed a integrative analysis framework based on machine learning algorithms. We firstly identified breast cancer-specific lncRNAs and AS genes by edgeR differential gene analysis from TCGA RNAseq tumor (n=1101) /normal (n=139) samples. Then we built co-expression network based on Spearman correlation pairwise distances from these cancer-specific lncRNAs and AS. We ranked top lncRNAs regulating alternative splicing in tumorigenesis by an optimized random walk multi-graphic method from the integrative networks of co-expression network, publically curated epigenetic network (e.g. ENCODE) and protein-protein interaction network (e.g. STRING). Linear regression analysis further refined individual candidate lncRNA regulation relating to their most correlated target splicing genes. The secondary structure of binding lncRNAs to potential splicing genes was predicted by support vector machine (SVM) algorithm to confirm the most conserved targets. Results: We identified 496 lncRNAs and 418 coding genes with alternative splicing isoforms associated with breast cancer. Pathway analysis predicted the functions of 14 candidate lncRNAs potentially regulating AS, are associated with cell migration, cell cycle progression, and more. We detected a high confidence lncRNA MALAT1 that regulates PKM gene alternative splicing, exclusion of either exon 9 or 10, in breast cancer tumorigenesis, which confirms the predictive strength of our proposed method. Conclusion: Understanding cancer-specific splicing machinery is therapeutically crucial in targeting the molecules that can influence the splicing process. We developed a network-based integrative framework to predict lncRNA regulations on breast cancer-specific alternative splicing as the potential therapeutic targets. Citation Format: Yunyun Zhou. Integrative analysis predicts lncRNA regulating gene alternative splicing in breast cancer [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 1423.