A SVM based approach for miRNA target prediction

MicroRNAs (miRNAs) are short RNAs that play important roles in post-transcriptionally regulation by binding to the target mRNAs. Although for a large number of animalspsila miRNAs have been defined, only a few targets have been known. Here we present a naive microRNA target prediction algorithm based on machine learning approach. SVM was used twice in our algorithm in order to make prediction for binding site and mRNA respectively. In order to avoid the loss of sensitivity, a set of seed match rules were defined base on observing experimentally validated targets to locate potential sites in 3psilaUTR sequences. TarBase and microarray data were used to build up database for training and evaluation of our algorithm. TargetScan and miRanda were implemented for comparison. The result shows that the performance of our algorithm is better than TargetScan and miRanda.