A machine learning strategy to identify candidate binding sites in human protein-coding sequence

Background: The splicing of RNA transcripts is thought to be partly promoted and regulated by sequences embedded within exons. Known sequences include binding sites for SR proteins, which are thought to mediate interactions between splicing factors bound to the 5' and 3' splice sites. It would be useful to identify further candidate sequences, however identifying them computationally is hard since exon sequences are also constrained by their functional role in coding for proteins. Results: This strategy identified a collection of motifs including several previously reported splice enhancer elements. Although only trained on coding exons, the model discriminates both coding and non-coding exons from intragenic sequence. Conclusion: We have trained a computational model able to detect signals in coding exons which seem to be orthogonal to the sequences' primary function of coding for proteins. We believe that many of the motifs detected here represent binding sites for both previously unrecognized proteins which influence RNA splicing as well as other regulatory elements.