Accurate transcriptome-wide prediction of microRNA targets and small interfering RNA off-targets with MIRZA-G

Abstract

Small interfering RNA (siRNA)-mediated knock-down is a widely used experimental approach to characterizing gene function. Although siRNAs are designed to guide the cleavage of perfectly complementary mRNA targets, acting similarly to microRNAs (miRNAs), siRNAs down-regulate the expression of hundreds of genes to which they have only partial complementarity. Prediction of these siRNA 'off-targets' remains difficult, due to the incomplete understanding of siRNA/miRNA-target interactions. Combining a biophysical model of miRNA-target interaction with structure and sequence features of putative target sites we developed a suite of algorithms, MIRZA-G, for the prediction of miRNA targets and siRNA off-targets on a genome-wide scale. The MIRZA-G variant that uses evolutionary conservation performs better than currently available methods in predicting canonical miRNA target sites and in addition, it predicts non-canonical miRNA target sites with similarly high accuracy. Furthermore, MIRZA-G variants predict siRNA off-target sites with an accuracy unmatched by currently available programs. Thus, MIRZA-G may prove instrumental in the analysis of data resulting from large-scale siRNA screens.

Figure 1.

Value of t-statistic in comparing the mean values of features used in the model (rows) among functional and non-functional miRNA seed-complementary sites across 26 experiments (columns). The data from the experiments labeled in blue were used to train the model and those from experiments labeled in red were used in testing the model.

Figure 2.

Comparative evaluation of various models. (A) Models’ performance in predicting mRNA down-regulation following miRNA transfection. The expectation is that a model performs well when its top predicted targets undergo the strongest downregulation after miRNA transfection. (B) Estimated number of functional targets for different methods as the function of the number of top predictions. Variants of the MIRZA-G model are described in Table 2. The other tested models are TargetScan Context+, TargetScan PCT, DIANA-microT-CDS and miRanda-mirSVR (the most conservative predictions). See text for additional details on these methods.

Figure 3.

Relationship between the prediction scores obtained with different target prediction methods and the extent of down-regulation of target mRNAs upon siRNA transfections. (A) Average over the siRNAs in the data set of Birmingham et al. (12). (B) Average over the siRNAs from Jackson et al. (13). (C) Data from an individual siRNA identified by van Dongen et al. (14) to have prominent off-target effects. (D) Data from an individual siRNA identified by van Dongen et al. (14) to have modest off-target effects. See also Table 2 and the text for details on the methods.

Figure 4.

SiRNA off-targets in the TGF-β pathway. (A) Schema of the TGF-β pathway drawn based on the figure provided by the DAVID server (42,43). Genes predicted to be off-targets of the top 100 siRNAs with the strongest effect in the screen are marked with red boxes. (B) Correlation between the z-score of an siRNA in the screen (y-axis) and the score that our model assigns to the interaction of the siRNA with TGFBR2 (x-axis). (C) Scatter plot of the predicted activities of the top 100 most active siRNAs on TGFBR1 and TGFBR2.