Prediction method for Intronic alternative polyadenylation sites.

Abstract

Alternative Polyadenylation (APA) of mRNAs has been proven as a considerable mechanism for post-transcriptional gene regulation. The interplay between Intronic APA and splicing may affect the isoforms of mRNAs. In this paper, we have found four prevalent motifs, i.e. AATAAA, TTTTTTTT, CCAGSCTGG and RGYRYRGTGG surrounding the polyadenylation sites; then we proposed a new computational method to identify the Intronic APA sites in the human genome, which is based on a Support Vector Machine (SVM) with weighted degree string kernel. Compared with other APA events, Intronic APA sites are likely to be with TTTTTTTT pattern. The proposed algorithm can correctly classify 89% of Intronic APA sites from the constitutive polyadenylation sites. In addition, the prediction accuracy of separating the Intronic APA from other types of APA could be up to 88.3%. The prediction results indicate that our computational method is promising for the identification of Intronic APA sites.