RFMirTarget: A Random Forest Classifier for Human miRNA Target Gene Prediction

MicroRNAs (miRNAs) are key regulators of eukaryotic gene expression whose fundamental role has been already identified in many cell pathways. The correct identification of miRNAs targets is a major challenge in bioinformatics. So far, machine learning-based methods for miRNA-target prediction have shown the best results in terms of specificity and sensitivity. However, despite its well-known efficiency in other classifying tasks, the random forest algorithm has not been employed in this problem. Therefore, in this work we present RFMirTarget, an efficient random forest miRNA-target prediction system. Our tool analyzes the alignment between a candidate miRNA-target pair and extracts a set of structural, thermodynamics, alignment and position-based features. Experiments have shown that RFMirTarget achieves a Matthew’s correlation coefficient nearly 48% greater than the performance reported for the MultiMiTar, which was trained upon the same data set. In addition, tests performed with RFMirTarget reinforce the importance of the seed region for target prediction accuracy.