Obtaining precise targets of microRNAs
Print | Close | Text Size:A A A


Tongbin Li, Ph.D.


Chief Bioinformatics Officer, and Head of Bioinformatics and Computational Research Laboratory, LC Sciences, LLC, Houston, TX 77054


Hosted by Dr. Chung-I Wu




The human genome encodes >1,000 microRNAs (miRNAs), which in turn regulate thousands of protein-coding target genes. Abnormalities in miRNAs and their targeting are associated with multiple forms of cancers, genetic disorders, cardiovascular disorders and neurodegenerative diseases, thus it is critically important to understand how miRNAs recognize their target genes. Because experimentally verified miRNA-target interactions were scarce, in many targeting rule studies, researchers were forced to pool target genes of multiple miRNAs and study them together. Consequently, the targeting rules uncovered in these studies are broad-scale, “general targeting rules” without considering miRNA-specific effects. Unfortunately, these “general targeting rules”, as implemented in well-known miRNA target predicting programs including TargetScanS, PicTar, miRanda and DIANA-microT, are far from satisfactory, manifested by the fact that predictions made by different programs only have limited overlaps, and that even the most accurate predicting programs produce up to two-thirds false positive predictions.


Recently, as a collaboration between our group and the groups of Drs. James Ferrell and Pat Brown at Stanford University, and Dr. Yan Zeng’s group at the University of Minnesota, we have developed the Argonaute (Ago) immunoprecipitation (IP) array assay to screen confident miRNA-specific targets in a high-throughput manner, and a set of computational methods to uncover refined miRNA targeting rules based on the Ago IP array data. Through these novel developments, we have obtained more precise targeting rules at individual miRNA resolution, and have discovered that transcriptional regulation by the miRNAs takes place preferentially on target sites located in the 3’UTR of the target genes, while post-transcriptional regulation takes place in both the 3’UTR and the ORF. While G:U wobbles are detrimental to target matching most of the time, there are situations when they are beneficial to target identification. Moreover, we have found that 3’ compensatory target sites (i.e., the ones without perfect “seed matching”) account for >15% of total target sites, much higher than previously estimated. Moreover, our analyses have confirmed that strong miRNA-specific effects exist in target identification.


This project was supported by NIH/NCI, University of Minnesota and LC Sciences.