For all miRNAs, the top 100 target predictions are shown. They can be sorted and filtered using various parameters, described below.
Our screen was based on the few known miRNA-target pairs, comparison of Drosophila melanogaster and Drosophila pseudoobscura, and minimal assumptions regarding miRNA-target interaction (see Stark and Brennecke, et al.). We found a cutoff value of Z=3 most suitable to distinguish between confident sites (i.e. likely to be true) and random matches. Known miRNA-targets were always among the top predictions when sorted by the single best site (ZMax) or the sum of all confident sites (ZUTR). The user can choose to sort by ZMax or ZUTR using the radio-buttons. We emphasize that target-predictions based on a single site are not statistically significant.
Site conservation in Anopheles gambiae was used to increase the significance of the predictions. Although powerful, it is limited to (Drosophila) genes for which we could identify the orthologous last (3') exon in Anopheles. Conservation in Anopheles increases the likelyhood that a site is functional. However, we emphasize that the absence of an identified ortholog (labelled NF) should not be taken as evidence that the prediction is not valid.
We re-examined the 11 validated miRNA-target pairs (6 of which were validated in our study) for common features that were not required during the search. These were implemented as flags to identify sites that resemble valid targets. Flag H indicates an uninterrupted helix of a least 6 nts within the first 8 nts of the miRNA. C requires 100% sequence conservation of the target region paired to nts 2-7 of the miRNA 5'end; CS+ / CS- require no overlap of target sites with coding sequence (sense / antisense). As all validated miRNA-target pairs conform to these features, all flags are applied by default. They can however be freely chosen in any combination by selecting the corresponding buttons.
In addition to the flags, the user can require site conservation in Anopheles (at a given Z score). Use of the flags, especially requiring conservation in Anopheles, increases the specificity of the predictions and permits the use of a less stringent Z-score cutoff. We recommend Z>2 when the flags are implemented, but note that the assumptions underlying the flags have not been systematically tested yet and using them might omit functional sites.
The top 100 predictions for all miRNAs are provided as a table with a specific format. The columns are as follows:
The rank (#) is the rank for the single predicted target site if sorted by Z, but the rank for the target gene (all sites combined) when sorted by ZUTR or ZMax (see below).
Name and Accession of the predicted target. The GO-Terms indicate its functional classification according to the Gene-Onthololy (GO) consortium.
Z-scores are a measure of non-randomness that we chose to normalize the Mfold free energy dG. An average random prediction has Z=0 with inreasing Zs being more reliable. Specifically, Z-scores are number of standard-deviations above the mean of a random sample (10,000 randomly chosed sites). ZMax is the single best site of a predicted target gene/UTR. ZSum is the sum of all sites that lie above the selected Z-score cutoff. This value is recalculated according to the user defined flags and Z-score cutoff.
MFOLD provides a folding free energy dG (kcal/mol) with its secondary structure prediction. We use this measure to assess miRNA-target interaction.
The sequence numbering (start and end) for predicted target sites refer to the D.melanogaster 3'UTR
Additional features common to all valid miRNA-target pairs that were not required during the search. These are an uninterrupted helix of a least 6 nts within the first 8 nts of the miRNA (H), a 100% sequence conservation of the target region paired to nts 2-7 of the miRNA 5'end (C), no overlap of target sites with coding sequence (sense or antisense - CS+ or CS-). If these features turn out to be generally required for miRNA-function, the search will be adapted to exclude sites that don't fullfill these making screens more specific. Now, we only flag them and let the user choose to exclude them from the predictions. An additional flag (P) indicates cases where a favorable Mfold free energy dG arises from base-pairing within the target-site or the miRNA itself, i.e. not contributing to the duplex stability. Checking the boxes above each flag requires the feature and discards predictions that do not fullfill it.
+/- indicate the presence/absence of a conserved site that passes the user defiend Z-score cutoff. NF means that no orthologous 3'UTR could be identified. Requiring conservation by checking the corresponding box discards predictions that are not conserved with a Z-score above the selected Z-score cutoff (- category; retains NF).
We provide a symbolic representation of the miRNA-target RNA-duplex as predicted by Mfold. A star * represents a regular Watson-Crick basepair, a plus + a G:T pair and a dash - a non-paired nucleotide (mismatch or insertion). Note that a mismatch translates into two dashes and one dash is a single nucleotide bulge. The 5'end of the miRNA is on the right, the upper alignment corresponds to the D.mel, the 2nd to the D.pseudoobscura and the 3rd to the A.gambiae target sequence (when available), paired with the miRNA of interest. Each representation links to the sequence that was used to evaluate the target site (i.e. target site, hairpin-linker, and miRNA sequence).