A Systematic Identification of RBPs Driving Aberrant Splicing in Cancer

Alternative Splicing (AS) is a post-transcriptional process by which a single RNA can lead to different mRNA and, in some cases, several proteins. Various processes (probably many of them yet to be discovered) are involved in the regulation of alternative splicing. This work focuses on the regulation by RNA-binding proteins (RBPs). In addition to splicing regulation, these proteins are related to cancer prognosis and are emerging therapeutic targets for cancer treatment. CLIP-seq experiments target selected RBPs and result in uncovering the loci of the nascent transcriptome to where the RBP binds to. The presence of changes in the splicing status surrounding these loci is a good starting point to establishing a causal relationship. The selection of the specific RBP(s) to target in the CLIP-seq experiment is not straightforward; in many cases, this selection is driven by apriori hypotheses. In this work, we have developed an algorithm to detect RBPs that are likely related to the splicing changes between conditions. To do this we have integrated several databases of CLIP-seq experiments with an algorithm that detects differential splicing events to discover RBPs that are especially enriched in these events. This is a follow-up of a previous work that is refined by 1) improving the algorithm to predict the splicing events and 2) testing different enrichment statistics, and 3) performing additional validation experiments. As a result, the new method provides more accurate predictions, and it is also included in the Bioconductor package EventPointer. We tested the algorithm in four different experiments where seven different RBPs were knocked down. The algorithm accurately states the statistical significance of these RBPs using only the alterations in splicing. We also applied the algorithm to study sixteen cancer types from The Cancer Genome Atlas (TCGA). We found relationships between RBPs and several cancer types like CREBBP and MBNL2 alterations in adenocarcinomas of the lung, liver, prostate, rectum, stomach, and colon cancer. Some of these relationships have been validated in the literature but other ones are novel. AvailabilityThis method is integrated EventPointer, an available Bioconductor R package.