Plant MicroRNA-target Prediction Research Based On The Integration Of Prediction Tools And Support Vector Machine

Confident identification of microRNA-target interactions is a vital way to study the function of microRNA (miRNA). Although several computational miRNA target prediction methods have been proposed for plants, the results of each predictor are inconsistent and usually lead to more false positive. To address these issues, an integrated model to identify plant miRNA-target interactions was developed.Traditional statistical approach and machine learning algorithm were integrated to identify and analyze Arabidopsis thaliana miRNA-target interactions. Three effective plant miRNA target prediction toolkits were used. The combination of psRNATarget, TAPIR and UEA provided a sufficiently large miRNA target candidate set. A support vector machine (SVM) classifier with features categorized by structural, thermodynamic, and position-based was implemented to reduce the false positive rate. In order to improve the performance of the proposed model, principle component analysis (PCA) feature extraction and self-training technology was introduced. Results showed that the proposed PCA-SVM strategy behaved better in detection and diagnosis with more accuracy comparing with a basic SVM approach. Moreover we evaluated the results using degradome sequencing supported Arabidopsis thaliana miRNA-target interactions. Further, the proposed model was run over Oryza sativa and Vitis vinifera to demonstrate gross usability of our model for other plant species. At last, we constructed gene regulatory network within miRNAs and target and co-expression network winth miRNAs.The integrated model of online predictors and local PCA-SVM classifier gained credible and high quality miRNA-target interactions. The supervised learning algorithm of PCA-SVM classifier was employed in plant miRNA target identification for the first time and its performance can be substantially improved if more experimentally proved training examples are provided.