Computational and Experimental Modeling of Alternative Polyadenylation

点击翻译按钮获取中文摘要

 DESCRIPTION (provided by applicant): Alternative polyadenylation (APA) is emerging as a pervasive mechanism in the regulation of most human genes under diverse physiological and pathological conditions. By changing the position of polyA site, APA can either shorten or extend 3' UTRs that contain many important cis-regulatory elements, such as miRNA binding sites. In this context, APA adds a new layer to how microRNA works, as mRNAs with shorter 3' UTRs will no longer be targeted, leading to higher expression. The role of APA in human diseases such as cancer is only beginning to be appreciated. Both proliferating and transformed cells have been shown to favor shortened 3′ UTRs, leading to activation of proto-oncogenes. In addition, our recent study (Nature 2014) identified CFIm25, a master APA regulator, as a glioblastoma (GBM) tumor suppressor, further underscoring the importance of APA in cancer development. However, in other disease models and cancer types beyond GBM, the critical target genes subject to APA, the functional consequences of APA and the mechanisms governing APA remain poorly understood. This is mainly because polyA profiling methods (PolyA-seq) have not been widely adopted. In contrast, RNA-seq has been widely used for gene expression analysis, yet most of these RNA-seq data have not been analyzed in an "APA aware" manner. Despite the above limitations, our preliminary data indicate that significant changes in APA usage between tumor and normal result in localized changes in RNA-seq read density within 3' UTR, which is readily detectable by our novel bioinformatics algorithm DaPars (Nature Commun. 2014). Therefore, we hypothesize that DaPars retrospective analysis of existing RNA-seq data can be used to study APA regulation in most cancer models and patient samples. The objective of this proposal is to reveal APA target genes, APA functional consequences and APA regulators by taking advantage of existing RNA-seq data of ~14,000 tumors across 33 cancer types, followed by experimental validation in cells and animal models. Furthermore, we will evaluate the efficacy of the bioinformatics method when measured in functional assays, which can then be used to further refine our analysis. Together, with novel bioinformatics methods, convincing preliminary results, big data analyses and functional validation, this proposal is uniquely positioned to make significant contributions towards our understanding of this new paradigm of gene regulation during tumorigenesis.