Enhanced Viral Precursor MicroRNA Identification with Structural Robustness Features in Back-Propagation Neural Network

Abstract

This article presents a new viral precursor miRNAs identification tool using back-propagation neural network. The tool mainly discriminates the viral precursor miRNAs from coding sequences and other pseudo precursor miRNAs. It was trained with viral precursor miRNAs from miRBase, pseudo precursor miRNAs and coding sequences from ViralmiR and NCBI database, respectively. Top 20 features out of totally 115 features including sequence-based features, secondary structure features, base-pair features triplet sequencestructure features and structural robustness features were selected using GainRatio. In particular, the six structural robustness features are found to be the most informative features among 20 selected features. Using the same set of 20 selected features, five-fold cross validation was applied to choose a classifier algorithm with the highest performance among support vector machine, random forest, decision tree, naive Bayes, k-nearest neighbor, and neural network based on the receiver operating characteristics area and accuracy. The results demonstrated that the back-propagation neural network classifier has the best performance compared with other algorithms. It achieved up to 97% accuracy on the test set with only 1% false positive rate.