Identification of 5´ Pre-miRNAs and 3´ Pre-miRNAs Employing Support Vector Machine and Local Structure Units

MicroKNAs (miRNA) are essential 21-22 nucleotides regulatory RNAs produced from larger hairpin precursors (pre-miRNA), and regulate gene expression through mRNA degradation or translational inhibition. We applied functional strand support vector machine (FS-SVM), a new method for prediction of functional strand on the miRNA precursors, to classifying 5´ and 3´ pre-miRNAs and achieved about 94% accuracy on human or mouse data. The FS-SVM classifier built on human and mouse data can correctly identify up to 90.9% of the pre-miRNAs from primates, and up to about 89.0% of the pre-miRNAs from other mammals.