Classification of Real and Pseudo pre-miRNAs in Plant Species

Abstract

Micro-RNAs (miRNAs) are an important class of small non-coding RNAs which play a crucial role in gene regulation at the translational level. Both perfect and nearly perfect binding of miRNAs through base complementary could cause translation inhibition. Identification of miRNAs in organisms has mostly been focused on precursors of miRNAs (pre-miRNAs). These pre-miRNAs are found only in non-coding regions. However, pseudo pre-miRNAs, which are RNA sequences in the coding region, can be folded as hairpin structures. The classification of real and pseudo miRNAs is more complicated for plants, when compared to animals, due to their wider diversity. This study aims to extract the features of pre-miRNAs for classifying real and pseudo pre-miRNAs in plants and compare classification performance of five different machine learning approaches. Stochastic-based Random Forest and Naïve Bayes Classifier showed the best classification performance with over 90% accuracy using 10-fold cross-validation and over 85% accuracy using cross-dataset validation.