Fast and effective prediction of microRNA/target duplexes

Abstract

MicroRNAs (miRNAs) are short RNAs that post-transcriptionally regulate the expression of target genes by binding to the target mRNAs. Although a large number of animal miRNAs has been defined, only a few targets are known. In contrast to plant miRNAs, which usually bind nearly perfectly to their targets, animal miRNAs bind less tightly, with a few nucleotides being unbound, thus producing more complex secondary structures of miRNA/target duplexes. Here, we present a program, RNA-hybrid, that predicts multiple potential binding sites of miRNAs in large target RNAs. In general, the program finds the energetically most favorable hybridization sites of a small RNA in a large RNA. Intramolecular hybridizations, that is, base pairings between target nucleotides or between miRNA nucleotides are not allowed. For large targets, the time complexity of the algorithm is linear in the target length, allowing many long targets to be searched in a short time. Statistical significance of predicted targets is assessed with an extreme value statistics of length normalized minimum free energies, a Poisson approximation of multiple binding sites, and the calculation of effective numbers of orthologous targets in comparative studies of multiple organisms. We applied our method to the prediction of Drosophila miRNA targets in 3'UTRs and coding sequence. RNAhybrid, with its accompanying programs RNAcalibrate and RNAeffective, is available for download and as a Web tool on the Bielefeld Bioinformatics Server (http://bibiserv.techfak.uni-bielefeld.de/rnahybrid/).

FIGURE 1.

Artefacts of target/miRNA concatenation. The linker sequence is CGNNNNNNCG. Hybridized are parts of 3′UTRs from D. melanogaster and the miR-2b miRNA. (A) The structure (from RNAfold) exhibits hybridization between target and linker (arrow). (B) Corresponding prediction from RNAhybrid that shows no artefact. (C) The structure (from RNAfold) exhibits self-hybridization of the target (arrow). (D) Corresponding prediction from RNAhybrid that shows no artefact. The 3′UTRs are from CG1969-RB (A,B) and CG30120-RA (C,D). The structures are drawn counter-clockwise, with the target followed by the miRNA. For RNAhybrid, the target is shown with its 5′-end additionally marked.

FIGURE 2.

The four best minimum free energy (MFE) duplexes of the let-7 miRNA and C. elegans 3′UTRs (5′-end marked) from the UTR database (http://srs.ebi.ac.uk). The targets are 3CEL000274, lin-14, 3CEL000914 lin-41, 3CEL000790 daf-12, and 3CEL000772 hunchback-related protein hbl-1. The alignments show the complete miRNAs. The target UTRs are shown where they hybridize to the respective miRNA, plus dangling bases on either side. Each UTR was only searched for one optimal hit. Note also that in database search mode, RNAhybrid normally gives a textual representation of hybridizations to avoid the accumulation of a large number of plots.

FIGURE 3.

The top two hits of the let-7 miRNA in the 3′UTR of the C. elegans lin-41 (5′-end marked). The first hit was also among the top hits in the target database search (cf. Fig. 2 ▶).

FIGURE 4.

Extreme value distribution density functions. The location and shape parameters are mean values for Drosophila miRNAs. The left curve shows negative normalized MFEs of duplexes that are constrained to have a miRNA 5′-helix from nucleotides 2–7. The right curve shows such energies without a helix constraint, thus allowing unpaired nucleotides in all parts of the duplexes.

FIGURE 5.

Fitting extreme value distribution parameters. The crosses show the log(−log) transformed empirical cumulative density function of negative normalized MFEs from the bantam miRNA and 5000 random target sequences. The straight line is fitted to negative normalized MFEs larger than 2.0 without the top 1% of data points.

FIGURE 6.

Linear correlation between minimal duplex energies (MDEs) and extreme value distribution parameters. The top plot shows MDEs of Drosophila miRNAs (x-axis) and corresponding fitted location parameters (y-axis). The bottom plot shows the same MDEs and corresponding fitted scale parameters. The data points are highly linearly correlated with a correlation coefficient of −0.86 for both data sets.