Computational prediction of competitive endogenous RNA

MicroRNAs (miRNAs) play an important role in the post-transcriptional regulation of gene expression by pairing target messenger RNAs (mRNAs). As the abnormal expression of miRNAs has been implicated in various diseases, there has been many studies on regulating the expression level of miRNA, including “miRNA sponges.” miRNA sponges, which are artificial miRNA decoys, contain complementary binding sites to a target miRNA and regulate the expression level of target miRNAs. As competitive endogenous RNAs (ceRNAs) have been found in a recent study, there have been many efforts to find natural miRNA sponges. However, there are no related studies about the computational approach using the pairwise interactions of numerous mRNA-miRNA pairs. In this study, a computational approach to find candidates of natural miRNA sponges is proposed. Whole miRNA binding sites with query miRNA and the secondary structures of reference mRNA are predicted, followed by calulating the adjusted minimum free energy (AMFE) as the total score. We can quantitatively compare the interactions between miRNAs and target mRNAs by using this proposed approach. Thirty viral miRNAs and about 300 of thousands of human mRNAs are used in this study. As a results, the top 20 natural miRNA sponge candidates are recorded. The results are expected to provide appropriate knowledge before in vivo experiments to validate the identification of miRNA sponges.