SVM-based pre-microRNA classifier using sequence, structural, and thermodynamic parameters

Abstract

microRNAs are single-stranded noncoding RNA sequences of 18–24 nucleotide length. They play important role in post-transcriptional regulation of gene expression. Last decade witnessed immense research in microRNA identification, prediction, target identification, and disease associations. They are linked with up/down regulation of many diseases including cancer. The accurate identification of microRNAs is still complex and time-consuming process. Due to the unique structural and sequence similarities of microRNAs, many computational algorithms have been developed for prediction of microRNAs. According to the current status, 28645 microRNAs have computationally discovered from the genome sequences, and have reported 1961 human microRNAs (miRBase version 21, released on June 2014). There are several computational tools available for predicting the microRNA from the genome sequences.We have developed a support vector machine-based classifier for microRNA prediction. Top ranked 19 sequence, structural, and thermodynamic characteristics of validated microRNA sequence databases are employed for building the classifier. It shows an accuracy of 98.4% which is higher than that of existing SVM-based classifiers such as Triplet-SVM, MiRFinder, and MiRPara.