Identification Of High Confidence RNA Regulatory Elements By Combinatorial Classification Of RNA–Protein Binding Sites

RNA-binding proteins(RBPs)are essential to sustain fundamental cellular functions,such as splicing,polyadenylation,transport,translation,and degradation of RNA transcripts.One study estimated that more than 1500 different RBPs exist in human.These RBPs cooperate or compete with each other in binding their RNA targets.Many RBPs are capable of binding different RNA targets,partially by associating with different co-factors.At the same time,some consensus RNA sequence motifs are recognized by homologous RBPs or homologous domains.Thus,proteins and RNAs appear to interact in a combinatorial manner.Crosslinking immunoprecipitation sequencing(CLIP-seq)technologies have enabled researchers to characterize transcriptome-wide binding sites of RNA-binding protein with high resolution.The CLIP-seq data of multiple RNA binding proteins have been curated and annotated in specific databases,such as CLIPdb,POSTAR,and STARbase.Several significant studies improved the prediction of individual RBPs' binding sites by training on CLIP-seq and RNAcompete datasets.Systematic assessment of combinatory regulation of multiple RBPs would be more beneficial to derive precious biological information from various high-throughput CLIP-seq data.We collected 327 CLIP-seq datasets in three cell lines generated from three technical approaches: PAR-CLIP,HITS-CLIP,and eCLIP.Deposited CLIP-seq datasets display significant variety.We apply a soft-clustering method,RBPgroup,to various CLIP-seq datasets to group together RBPs that specifically bind the same RNA sites.We provide a unified and high-confidence set of protein-binding RNA sites and clustered RBP groups,which were validated by the known physical interactions and co-IP experiments.The binding sites defined by our method were more enriched with known motifs and better correlated with RNA degradation data and alternative splicing data than the binding peaks of single RBPs.In summary,we show that integrating public CLIP-seq datasets can provide novel insights into the combinatorial classification of RBPs.We shared our method and code,as well as the derived RNA regulatory elements,with the RNA community via a web-based platform.