Computational methods for the identification of mature microRNAs within their Pre-miRNA

Abstract

The urgent demand in miRNA research has call for the high performance computational methods for mature miRNA identification to supplement the biological experiment methods. In this study, we analyzed the secondary structure of pre-miRNA and extracted the important features. Then the current computational methods are investigated, and the flow chart of mature miRNAs location prediction methods is summarized. In addition, the current methods and algorithms are classified and assessed. Notably, we compare five machine learning algorithms of Naive Bayes, SVM, Random Forest, the Conditional Random Field and Adaboosting for mature miRNA-located prediction. Empirical findings indicated that SVM algorithm could achieve better performance than Naive Bayes method. And the Random Forest method is comparable to the performance of SVM, it shows good performance in this subject.