A machine learning approach for miRNA target prediction

MicroRNAs (miRNAs) are 21 or 22 nucleotides noncoding RNAs known to possess important post-transcriptional regulatory functions. Identifying targeting genes that miRNAs regulate is important for understanding their specific biological functions. Usually, miRNAs down-regulate target genes through binding to the complementary sites in the 3' untranslated region (UTR) of the targets. Since the binding of the miRNAs of animals is not a perfect one-to-one match with the complementary sites of their targets, it is difficult to find targets of animal miRNAs by accessing their alignment to the 3' UTRs of potential targets. More sophisticated computational approaches are desirable and have been proposed as a result. The most popular algorithms include TargetScan, miRanda, and PicTar. However, they share similar methodology and are restricted by the human observation of conserved nature of miRNAs and their targets. In this article, we develop a statistical learning based approach that uses support vector machine (SVM) as a classifier to predict miRNA targets. SVM have been applied in many fields such as pattern recognition, computational biology, and medical image analysis. With SVM, information is gained automatically from relevant data and therefore human bias can be removed in the decision process.