Study On Mirna Targets Prediction Based On Svm Algorithm

MicroRNA, which regulates the post-transcriptional expression of target genes and affects the growth and development of the organism, is an important class of small RNA. Identifying the target genes of certain microRNA is fundamental to understanding the mechanism of its work, however it become the bottle neck of microRNA research because of the laking of effective biological experiments and high accuracy prediction algotithms.This paper presents an algorithm basing on the features of microRNA terget binding site context region. The algorithm try to solve the microRNA target prediction accuracy issues by combining with the microRNA target sites relevant regional characteristics, inregration of biological information, as well as motif characteristics.1.Resource analysis:This paper overviews currently existing microRNA and microRNA targets database resources.An abroad and thorough research is carried out to survey a large number of existing computational miRNA target prediction algorithms. Basing on the investigation of the mechanisms of each algorithm the performance evalustion of several algorithms, the alaysis of non-3’UTR target prediction results, shortages of existing algorithms are sumrized. The possible research directions are point out as well.2.Data collection:The paper collected valided microRNA targets from authoritative database, downloaded each188,73,22miRNA genes of three species of human,mouse and Drosophila. I downloaded genome data from NCBI FTP and used perl script for analysis and processing. By Literature mining, I collected827miRNA validated target data of three species as positive set. By miRanda and TargetScan prediction, I used the results as negative set.3.Feature extract:By comparing the two kinds of targets, the paper finally selected17kinds of features as candidates. At last, I used8kinds of features for building the SVM classifier. These include GC content, folding free energy, motif number and target site position.4.Algorithm development:a SVM classifier based algorithm svMicroTar is proposed for target prediction which aiming at two types of non-3’UTR region targets in sequence. The paper used grid search method to optimize the SVM classifier parameters, then trained and evaluated SVM classifier. The simulation results show that this algorithm can distinguish between non-3’UTR targets a certain extent and improve the accuracy of the3’UTR target prediction. This algorithm can be applied to the target secondary screening by formatting any target prediction results. It has good generalization capacity.