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
. 2012 Oct;40(19):9404-16.
doi: 10.1093/nar/gks759. Epub 2012 Aug 16.

Toward a combinatorial nature of microRNA regulation in human cells

Ohad Balaga  1 , Yitzhak Friedman, Michal Linial
Affiliations

Toward a combinatorial nature of microRNA regulation in human cells

Ohad Balaga et al. Nucleic Acids Res. 2012 Oct.

Abstract

MicroRNAs (miRNAs) negatively regulate the levels of messenger RNA (mRNA) post-transcriptionally. Recent advances in CLIP (cross-linking immunoprecipitation) technology allowed capturing miRNAs with their cognate mRNAs. Consequently, thousands of validated mRNA-miRNA pairs have been revealed. Herein, we present a comprehensive outline for the combinatorial regulation by miRNAs. We implemented combinatorial and statistical constraints in the miRror2.0 algorithm. miRror estimates the likelihood of combinatorial miRNA activity in explaining the observed data. We tested the success of miRror in recovering the correct miRNA from 30 transcriptomic profiles of cells overexpressing a miRNA, and to identify hundreds of genes from miRNA sets, which are observed in CLIP experiments. We show that the success of miRror in recovering the miRNA regulation from overexpression experiments and CLIP data is superior in respect to a dozen leading miRNA-target prediction algorithms. We further described the balance between alternative modes of joint regulation that are executed by pairs of miRNAs. Finally, manipulated cells were tested for the possible involvement of miRNA in shaping their transcriptomes. We identified instances in which the observed transcriptome can be explained by a combinatorial regulation of miRNA pairs. We conclude that the joint operation of miRNAs is an attractive strategy to maintain cell homeostasis and overcoming the low specificity inherent in individual miRNA-mRNA interaction.

PubMed Disclaimer

Figures

Figure 1.
Figure 1.
A workflow for assessing the performance of miRNA predictions. Analysis of miRror2.0 is illustrated for genes as an input (Gene2miR mode, default parameters). A filter for the collection of the top predictions according to internal scores was activated. Stringency refers to parameters including the choice of P-value threshold. Additional parameters that provide fine-tuning for the platform are ignored (tissue of interest, highly expressed subset, MDB choices). By changing these parameters, a relaxed or a strict search protocol can be activated. We assigned several categories for success according to the rank of the prediction (prediction ranked 1, 2, 3–5, 6–9, ≥10).
Figure 2.
Figure 2.
Success rates of MDB’s prediction. (A) 10% of the down-regulated genes from 30 experiments were subjected to miRNA prediction by the 12 MDBs and miRror. The color intensity is according to the indicated predictions’ quality. miRror results are shown in purple (Supplementary Table S5). (B) Reanalysis of the same input as in (A) with top predictions (25%) by the internal scoring system for each MDB. Only 10/12 MDBs provide an internal score (Supplementary Table S5). For example, PicTar-4 performs poorly when the top 25% of its internal score is applied. However, without such filtration, a good performance of PicTar-4 is recorded (compare Figure 4A with Figure 4B). (C) Average success rates based on the 12 MDBs for each of the 30 transcriptomics datasets. A miRNA overexpression experiment that is marked as 99% (green line) means that the rank across the 12 MDB is in the top 1% predictions. The colors mark the prediction success as very high (green), high (blue), moderate (orange) and poor (red).
Figure 3.
Figure 3.
miRror internal scoring. (A) The overexpression of hsa-miR-124 transcriptomic data (10% of the down-regulated genes) is used to predict the top 100 miRNAs. (A) The fraction of MDBs (from 12) that agree on each of the miRNA prediction (orange color), and the fraction of genes from the input genes that are explained by the miRNA prediction (blue color). A minimal agreement of two MDBs and of two genes from the input is requested. The unified miRIS is the averaged of these components. The 100 predicted miRNAs are sorted according to miRIS. (B) The prediction list, ranked by miRIS for hsa-miR-7 (GSE14507). The highest miRIS identified hsa-miR-7 whereas the other MDBs failed to predict it. The average rank for hsa-miR-7 by all the MDBs is 38.4 ± 30.6. (C) Predictions of hsa-miR-124 (GSE6207). The miRIS indicated hsa-miR-124 as the second prediction and hsa-miR-506 as the first one. These two miRNAs belong to the same family according to TargetScan. Insets display a zoomed view on the top 10 predictions. Note the abrupt drop in score among the top 10 predictions.
Figure 4.
Figure 4.
CLIP data analysis. (A) A cumulative view of the number of miRNAs regulating each gene, from the CLIP experiments in StarBase (52). The cumulative view demonstrates the prevalent regulation of genes by multiple miRNAs. The Total number of genes in the analysis is 6287. (B) A scheme for miRror interpretation of the CLIP data. Using miR2Gene mode to assess the recovery of a gene from the collection of miRNAs that are based on CLIP experimental data. (C) Sets of miRNAs that are regulated by 30–40 miRNAs per gene are the input for miRror application. 362 such genes were analyzed and success is categorized as two levels (top 1–9 and top 10–20 genes). Y-axis is defined as the fraction of success prediction from all selected targeted genes (total 362). The prediction success of miRror (purple) and the 12 MDBs are shown (Supplementary Table S7). Inset, the fraction of identified genes in the entire prediction list. miRror outperformed the other 12 MDBs in both success measurements.
Figure 5.
Figure 5.
Assessment of the combinatorial nature of miRNAs. (A) miRNA regulation of miRNA-Duos according to the JI. The JI combines the intersection and the union of the targeted genes by the miRNA-Duos. Data on the miRNA-pairs are extracted from the CLIP-based experiments. The average JI value for several of the most successful MDBs is shown in view of the observations from the CLIP experiments (from StarBase). Average JI calculated for all 12 MDBs are shown in Supplementary Table S8. (B) The distribution of miRNA-Duos target overlap for several of the most successful MDBs is shown. Note that each MDB is characterized by a unique distribution. The CLIP data shows 17.5% of miRNA-Duos as having no joint targets, in contrast with the MDBs who show no such pairs. Still, a generally low overlap dominates both the MDBs and the experimental data. (C) An illustration of a scheme for the intersection of any six pairs of miRNA (for miRNAs that are marked miR-a, miR-b, miR-c and miR-d). In this scheme, the interesting candidate pair that reaches a maximal coverage (measured by JM) is the combination of miR-a and miR-b. Note that miR-c and miR-d are actually included in the predicted list of genes of miR-a and miR-b. (D) Global analysis of 20 experiments in manipulated cell-lines (Supplementary Table S9). The top 5 miRNA predictions are listed as 10 pairs (total of 200 data-points). The joint miRIS is plotted in view of the JI. miRNA pairs that are within the statistical variance of the data are highlighted in light yellow. Several extreme cases are emphasized. Note the 4 pairs that involve miR-218. The combination of some miRNAs with miR-218 led to a substantial increase in the number of predictions relative to the input list.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Bartel DP. MicroRNAs: genomics, biogenesis, mechanism, and function. Cell. 2004;116:281–297. - PubMed
    1. Landgraf P, Rusu M, Sheridan R, Sewer A, Iovino N, Aravin A, Pfeffer S, Rice A, Kamphorst AO, Landthaler M, et al. A mammalian microRNA expression atlas based on small RNA library sequencing. Cell. 2007;129:1401–1414. - PMC - PubMed
    1. Calin GA, Ferracin M, Cimmino A, Di Leva G, Shimizu M, Wojcik SE, Iorio MV, Visone R, Sever NI, Fabbri M, et al. A MicroRNA signature associated with prognosis and progression in chronic lymphocytic leukemia. N. Engl. J. Med. 2005;353:1793–1801. - PubMed
    1. Grad Y, Aach J, Hayes GD, Reinhart BJ, Church GM, Ruvkun G, Kim J. Computational and experimental identification of C. elegans microRNAs. Mol. Cell. 2003;11:1253–1263. - PubMed
    1. Friedlander MR, Chen W, Adamidi C, Maaskola J, Einspanier R, Knespel S, Rajewsky N. Discovering microRNAs from deep sequencing data using miRDeep. Nat. Biotechnol. 2008;26:407–415. - PubMed

Publication types