Prediction of functional microRNA targets by integrative modeling of microRNA binding and target expression data

Abstract

We perform a large-scale RNA sequencing study to experimentally identify genes that are downregulated by 25 miRNAs. This RNA-seq dataset is combined with public miRNA target binding data to systematically identify miRNA targeting features that are characteristic of both miRNA binding and target downregulation. By integrating these common features in a machine learning framework, we develop and validate an improved computational model for genome-wide miRNA target prediction. All prediction data can be accessed at miRDB ( http://mirdb.org ).

Keywords: CLIP-seq; MicroRNA; RNA-seq; Target prediction.

Fig. 1

The impact of miRNA seed types on target downregulation. Six seed types were evaluated (see Table 2 for seed definitions). a Percentage of downregulated genes containing individual seed types in relation to gene expression changes. All 25 miRNAs were included in the analysis. b Analysis of a subset of 8 miRNAs that do not contain 5′-U

Fig. 2

Comparison of miRNA target prediction algorithms using the HITS-CLIP dataset. MirTarget and four other target prediction algorithms were included in the analysis. a Receiver operating characteristic (ROC) curve analysis to evaluate the rate of false positive prediction in relation to the rate of true positive prediction. b Precision-recall (PR) curve analysis to evaluate prediction precision in relation to the recall rate

Fig. 3

Comparison of target prediction algorithms using microarray data. Microarray profiling data were analyzed to identify target upregulation resulting from concurrent inhibition of 25 miRNAs. a Correlation of target upregulation and target prediction scores computed by 5 individual algorithms, as measured by Pearson correlation coefficient. b Average level of expression upregulation for predicted targets. For each algorithm, 100 top-scoring predicted targets per miRNA on average were included in the analysis