Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2013 Oct;41(18):8421-33.
doi: 10.1093/nar/gkt629. Epub 2013 Jul 17.

MREdictor: a two-step dynamic interaction model that accounts for mRNA accessibility and Pumilio binding accurately predicts microRNA targets

Danny Incarnato  1 , Francesco Neri, Daniela Diamanti, Salvatore Oliviero
Affiliations

MREdictor: a two-step dynamic interaction model that accounts for mRNA accessibility and Pumilio binding accurately predicts microRNA targets

Danny Incarnato et al. Nucleic Acids Res. 2013 Oct.

Abstract

The prediction of pairing between microRNAs (miRNAs) and the miRNA recognition elements (MREs) on mRNAs is expected to be an important tool for understanding gene regulation. Here, we show that mRNAs that contain Pumilio recognition elements (PRE) in the proximity of predicted miRNA-binding sites are more likely to form stable secondary structures within their 3'-UTR, and we demonstrated using a PUM1 and PUM2 double knockdown that Pumilio proteins are general regulators of miRNA accessibility. On the basis of these findings, we developed a computational method for predicting miRNA targets that accounts for the presence of PRE in the proximity of seed-match sequences within poorly accessible structures. Moreover, we implement the miRNA-MRE duplex pairing as a two-step model, which better fits the available structural data. This algorithm, called MREdictor, allows for the identification of miRNA targets in poorly accessible regions and is not restricted to a perfect seed-match; these features are not present in other computational prediction methods.

PubMed Disclaimer

Figures

Figure 1.
Figure 1.
PRE is enriched in validated and predicted thermodinamically inaccessible regions. (A) Correlation of different features to miRNA binding effectiveness measured on the training dataset. Local target accessibility (ΔGaccess) and AU content exhibit the greatest correlation. (B) Cumulative frequency plots show that functional targets (positive data set) are more likely to reside within 3′-UTR regions with a higher AU content, and that there is higher target accessibility with respect to non-functional targets (negative data set). P-values are given by Welch’s t-test. (C) Motif discovery analysis performed using MEME shows an enrichment for the PRE motif belonging to the Pumilio family proteins (e-value: 1.2e-22). Top enriched mRNA bearing PRE motifs are shown. (D) Presence of the PRE motif well stratifies the functional interactions between highly accessible and poorly accessible targets. The P-value is given by Welch’s t-test. (E) Cumulative frequency plot of top 20 PRE-associated human miRNA families. Seed-matches with nearby PRE motifs are enriched within poorly accessible regions with respect to those lacking a close PRE. The P-value is given by the Kolmogorov–Smirnov test. (F) The base probability of the seed-match positions for the top 20 PRE-associated miRNA families in human. Sites with nearby PRE motifs exhibit a higher probability of being already engaged in a bond within the local 3′-UTR secondary structure. Positions m8 to m1 are paired, respectively, to miRNA’s seed positions 1–8. P-values per base are given by a Chi-Squared test.
Figure 2.
Figure 2.
Schematic representation of MREdictor’s pipeline. Given a target to test and a miRNA, MREdictor obtains sequences for the 3′-UTR from UTRdb and for the miRNA from miRBase. For every possible seed-match, the local accessibility is evaluated, and regions exceeding the ΔGaccess energy cost of −10 kcal/mol are subjected to a Positional Weight Matrix scan for possible PRE motifs. If no PRE is discovered, the site is discarded. Sites that pass this first filtering step are then subjected to a simulation of duplex formation and are filtered according to their free energy (ΔGduplex).
Figure 3.
Figure 3.
Two step calculation of miRNA-MRE pairing. (A) Two-step model of the miRNA–mRNA interaction. (B) Comparison of a Smith–Waterman-like approach and the MREdictor method for miRNA-MRE duplex modeling, which consist of the nucleation of the miRNA-MRE duplex at the seed level followed by an extended 3′-end pairing (see ‘Materials and Methods’ section), applied to a well-known worm MRE. Previous studies have shown that, in Caenorhabditis elegans, the lin-4 miRNA can target lin-14, forming a bulged structure within lin-14 3′-UTR (25,33,34). A Smith–Waterman-like approach tends to maximize only the number of interactions between the miRNA and the target, minimizing the number of gaps and the extension of the gaps; thus, it fails to correctly predict the duplex secondary structure and underestimates the duplex free energy.
Figure 4.
Figure 4.
Validation of MREdictor method. (A) Comparison of standard performance measures of state-of-the-art algorithms and MREdictor. (B) Schematic representation of two non-canonical MREs predicted by MREdictor but not by other tools, in the absence of a perfect seed-match. Hsa-miR-122* was predicted to target REST (ΔGduplex = −20.9 kcal/mol), whereas mmu-miR-667* was predicted to target Sirt1 (ΔGduplex = −34.7 kcal/mol). (C) Dual luciferase assay validation of the two predicted MREs. Data are averaged over four replicates, and error bars are given for S.Ds.
Figure 5.
Figure 5.
Validation of PUM-dependent sites predicted by MREdictor. (A) Accessibility plot of a DDX58 3′-UTR window containing two MREs for miR-297 and miR-410, which are predicted to be dependent on Pumilio proteins binding to their cognate PRE motif. (B) Schematic representation of the two MREs, as predicted by MREdictor. (C) Dual luciferase assay validation of the two predicted MREs within DDX58 3′-UTR. Data are averaged over four replicates, and error bars are given for S.Ds. (D) RT-qPCR analysis and western blot of HEK293 FT cells transfected either with a negative control or two siRNAs targeting PUM1 and PUM2.
Figure 6.
Figure 6.
Microarray validation of MREdictor predictions. (A) Heatmap showing the two clusters of genes that emerge after microarray analysis of HEK293 cells transfected with miR-297 in the presence or absence of PUM1/2. In the PUM-dependent cluster, the miR-297-mediated repression is abolished (or significantly reduced) following PUM1/2 double knockdown. (B) Cumulative frequency plot of the PUM-dependent, PUM-independent and non-regulated gene clusters. Genes within the PUM-dependent cluster are more likely to reside within poorly accessible regions compared with the PUM-independent cluster, which is similar to the non-regulated genes. The P-value is given by the Welch t-test. (C) Schematic representation of the 3′-UTR of five genes from each cluster, and RT-PCR validation of microarray analysis. Data are averaged over triplicate experiments, and error bars are given for the S.Ds. (D) Venn diagram showing the overlap between the predictions for miR-297 of different tools and MREdictor compared with microarray downregulated genes. MREdictor successfully predicts 99 unique targets, which are not predicted by other programs. (E) Comparison of standard performance measures of state-of-the-art algorithms and MREdictor predictions for miR-297, calculated over microarray data.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Lambert NJ, Gu SG, Zahler AM. The conformation of microRNA seed regions in native microRNPs is prearranged for presentation to mRNA targets. Nucleic Acids Res. 2011;39:4827–4835. - PMC - PubMed
    1. Lewis BP, Shih IH, Jones-Rhoades MW, Bartel DP, Burge CB. Prediction of mammalian microRNA targets. Cell. 2003;115:787–798. - PubMed
    1. Frank F, Sonenberg N, Nagar B. Structural basis for 5′-nucleotide base-specific recognition of guide RNA by human AGO2. Nature. 2010;465:818–822. - PubMed
    1. Brodersen P, Voinnet O. Revisiting the principles of microRNA target recognition and mode of action. Nat. Rev. Mol. Cell Biol. 2009;10:141–148. - PubMed
    1. Wang Y, Sheng G, Juranek S, Tuschl T, Patel DJ. Structure of the guide-strand-containing argonaute silencing complex. Nature. 2008;456:209–213. - PMC - PubMed

Publication types