A multilevel analysis of hiv1-miR-H1 miRNA using KPCA, K-means, Random Forest and online target tools

Abstract

The goal of this study was to propose a workflow using machine learning to identify and predict the miRNA targets of Human Immunodeficiency virus 1. miRNAs which is ~21 nt long are attained from larger hairpin RNA precursors and is maintained in the secondary structure of their precursor relatively than in primary chain of successions. The proposition approach for identification and prediction of miRNA targets in hiv1-miR-H1is based on secondary structure and E-value through machine learning. Data Linearity of Length and e-value for sequence match with hiv1-mir-H1 is verified using Kernel PCA. miRNA targets were grouped into clusters thereby indicating similar targets using K-means algorithm. Classification model using Random Forest was implemented regards to each secondary features variable considering feature relevance. A learning methodology is put forward that assimilate and integrate the score returned by various machine learning algorithms to predict cellular hiv1-miR-H1 targets. Gene targets results using TargetScan, miRanda, PITA, DIANA microT and RNAhybrid are also explored for multiple parameters.