About miRNAs, miRNA seeds, target genes and target pathways

Abstract

miRNAs are typically repressing gene expression by binding to the 3' UTR, leading to degradation of the mRNA. This process is dominated by the eight-base seed region of the miRNA. Further, miRNAs are known not only to target genes but also to target significant parts of pathways. A logical line of thoughts is: miRNAs with similar (seed) sequence target similar sets of genes and thus similar sets of pathways. By calculating similarity scores for all 3.25 million pairs of 2,550 human miRNAs, we found that this pattern frequently holds, while we also observed exceptions. Respective results were obtained for both, predicted target genes as well as experimentally validated targets. We note that miRNAs target gene set similarity follows a bimodal distribution, pointing at a set of 282 miRNAs that seems to target genes with very high specificity. Further, we discuss miRNAs with different (seed) sequences that nonetheless regulate similar gene sets or pathways. Most intriguingly, we found miRNA pairs that regulate different gene sets but similar pathways such as miR-6886-5p and miR-3529-5p. These are jointly targeting different parts of the MAPK signaling cascade. The main goal of this study is to provide a general overview on the results, to highlight a selection of relevant results on miRNAs, miRNA seeds, target genes and target pathways and to raise awareness for artifacts in respective comparisons. The full set of information that allows to infer detailed results on each miRNA has been included in miRPathDB, the miRNA target pathway database (https://mpd.bioinf.uni-sb.de).

Keywords: microRNA; non-coding RNA; systems biology; target gene.

Figure 1

(A) For all 3. 25 Million pairs of miRNAs the similarity of the sequence, the seed sequence, target gene sets and target pathways is presented as histograms. Similarity scores (described in the Methods section) are in the range between 0 (no similarity) and 1 (identical). The bin width has been set to 0.01 (maximum of 100 bars per histogram). (B) Clustered correlation matrix. The correlation of the considered features are shown as heat map with dendrogram on the left side. Highest correlation was obtained for target pathway similarity compared to target gene set similarity followed by miRNA similarity to miRNA seed similarity.

Figure 2

(A) Scatter plot of miRNA sequence similarity score compared to miRNA seed similarity score. Since the plot contains 3.25 million data points, the density distribution is also shown in red. The green and blue sequences represent extreme examples, i.e. miRNAs with high sequence similarity compared to the seed similarity (blue) and miRNAs with high seed sequence similarity compared to the overall lower sequence similarity (green). (B) Scatterplots for the sequence similarity and the seed similarity, each compared to the genomic distance. The blue ellipses highlight miRNAs with similar sequences and similar seeds, respectively, that also have close genomic proximity.

Figure 3

(A) Scatter plots for the seed similarity compared to the target gene set similarity (left part) and the target pathway set similarity (right part). Again, extreme cases not following the expected pattern are highlighted. The area proportional Venn diagrams highlight the overlap in the gene set and the pathway set for the respective examples.

Figure 4

(A) Scatter plot for the target gene set and the target pathway similarity. The ellipse in the upper left part highlights miRNA pairs with overall low target gene similarity and high target pathway set similarity. The example on the right side presents two miRNAs (hsa-miR-3529-5p and hsa-miR-6886-5p) with low target gene similarity and significant target pathway similarity, as demonstrated by the area proportional Venn diagrams. (B) For the two miRNAs from Figure 4A the targeting of the KEGG MAPK signaling cascade is detailed. Targets of miR-3529-5p are highlighted in blue, targets of miR-6886-5p in red and joint targets of both miRNAs in purple. As the figure shows, both miRNAs target specific parts of the pathway such that overall a very large fraction of the path is targeted while the overlap between both miRNAs is small.