miRHunter: A Tool for Predicting microRNA Precursors Based on Combined Computational Method

摘要

MicroRNAs (miRNAs) are small endogenous non-coding RNAs known to post-transcriptionally regulate gene expression in a broad range of organism. Since the discovery of the very first miRNAs, lin-4 and let-7, computational methods have been indispensable tools that complement experimental approaches to understand the biology of miRNAs. In this article, we introduce a web-based computational tool, miRHunter, that identifies potential miRNA precursors (pre-miRNAs) in the genomic sequences by using a combined computational method. The method coupled ab initio method with homology-based and hairpin structure-based methods. The miRHunter consists of five modules: 1) a preprocessing module, 2) an evolutionary conservation filter module, 3) a hairpin structure filter module, 4) a support vector machine module that evaluates preliminary pre-miRNA candidates derived from the previous two filtering modules, and 5) a post-processing module. The miRHunter system yielded the following average test results: 96.16%/93.23%, 96.00%/94.68%, and 95.87%/93.57% which are sensitivity (Sn) and specificity (Sp) for animal, plant, and overall categories respectively. The miRHunter system can complement experimental methods and allow wet lab researchers to screen long sequences for putative miRNAs as well as pre-testing miRNAs of interest. The microarray profiling experiments have supported that the clusters of proximal pairs of miRNAs are generally coexpressed. Therefore, the clustering or spatial localization information will be used to improve the accuracy of our system in further work. The miRHunter is available at http://www.bioinfoworld.com/.