Combining Structure-Based Features and Conserved Data for MicroRNA Target Prediction by the Neural Network Method

Abstract

Most of MicroRNAs are thought to control post-transcriptional mechanisms by base pairing with MicroRNA recognition elements found in their messenger RNA (mRNA) targets. A new computational method we provide is to predict mRNA targets for human MicroRNAs. Combined structure-based features and conserved data across species, the overall results are 89.1% in sensitivity. This means, the about 30nt short sequences of nucleotides can be accepted by our system not only when they appear in the interior loops and bulges, but also in the G:U, A:U and G:C nucleotide pairs of RNA secondary structures. We also provide the computationally deriving that guides single MicroRNA for multiple target mRNAs recognition. Incorporation of computational procedures allows prediction of human MicroRNA containing multiple target mRNAs. The results suggest that by providing structure-based features and conserved data can improve the performance of predicting MicroRNA target mRNAs.