Large scale analysis reveals novel insights in the characteristics of miRNA targeting.

MicroRNAs (miRNAs) have emerged as central post-transcriptional regulators of gene expression in higher eukaryotes. The about 23 nucleotides long RNA molecules regulate gene expression by binding preferably to the 3'untranslated region (3'UTR) of protein-coding messenger RNAs (mR-Nas). Generally, miRNA-mediated regulation represses protein synthesis either by inhibition of translation or by degradation of mRNAs [FBS08]. To understand the functional role of miRNAs it isessential to have knowledge of their regulatory targets. Although the number of experimentally validated miRNA targets is increasing, the majority of interactions are still unknown. Computational miRNA target prediction has emergedd to uncover unknown miRNA-mRNA interactions. Meanwhile several rules that reflect miRNA targeting have been suggested and subsequently adopted for computational target prediction [Bar09]. Of particular importance for target recognition are sites of length 6 to 8 nucleotides that pair to the 5' region of the miRNA, which is referred to as the miRNA seed [CZMD09]. Specificity of prediction is improved by considering further targeting features such as conservation of the target site, position of the site within the 3'UTR [GFJ+07] as well as the structural accessibility of the target region [KIU*07]. Although the accuracy of prediction methods was steadily growing the number of false positive predictions methods was steadily growing the number of false positive predictions is still very high. Consequently, the existing models of target recognition by miRNAs are still insufficient to describe the reality of miRNA-target interaction. Recently, a miRNA-mRNA interaction map has been published containing a set of verified target sites for 20 miRNAs in the transcriptome of the murine brain [CZMD09]. The large size of the map and the precise localization of miRNA target sites allowed us to study miRNA target recognition as well as the performance of target prediction methods in an unprecedent way