Background | Frequently Asked Questions | Bugs


Score, energy, and P-values

Score The miRanda algorithm was used to scan all available miRNA sequences for a given genome against 3' UTR sequences of that genome. The algorithm uses dynamic programming to search for maximal local complementarity alignments, which correspond to a double-stranded anti-parallel duplex. A positive score is given for complementary base pairing, and a negative score is given to mis-matches, gap-opening and gap-extension. Importantly, scores derived from the 5' end of the miRNA where multiplied by a scaling factor of 4.0, to reflect the apparent importance of perfect Watson-crick base pairing which has been observed experimentally. The overall score for a hit is simply the summation of these derived scores across the total miRNA vs UTR alignment.

Energy is computed by the Vienna RNA folding routines and is a measure for the thermodynamic stability of a duplex. Energy is not used in the computation of P-values as described below.

Statistics We also compute P-values for all target sites. Only target sites for which P-org (see below) is below 0.05 are included. We closely follow the statistical model proposed by Rehmsmeier et al.

Marc Rehmsmeier, Peter Steffen, Matthias Hochsmann, and Robert Giegerich, Fast and effective prediction of microRNA/target duplexes RNA, 10:1507-1517, 2004. [ PUBMED ]

P-base is the base P-value; this is computed using distribution parameters derived from the genomic background of miRanda scores. The background is fitted according to an extreme value distribution. This functionality is now part of miRanda and will be available with its next release.

P-Poisson computes the probability that a single transcript has more than one (significant) hit, according to a Poisson distribution (applicable to rare events).

P-org or P-ortholous-group computes an estimated probability of the same microRNA family hitting multiple transcripts for different species in an orthologous group. This is basically done by taking into account the level of sequence conservation between all the 3' UTRs. The implementation of this step diverges from the approach taken by Rehmsmeier et. al. and will be detailed in a forthcoming publication.

Frequently Asked Questions

How do I use miRBase Targets to find out if my gene of interest has microRNA target sites?

If your gene has an EnsEMBL identifier or an official gene name (e.g. HUGO, MGI or ZFIN) then simply click the Search button on the miRBase Targets homepage. Once at the Search page select the desired genome you want to search, enter the EnsEMBL identifier or gene name and click search.

How do I query miRBase Targets for all the targets of a microRNA of interest?

If your miRNA has an official miRNA registry identifier then use the search page available from the miRBase Targets homepage, select the genome and enter the registry id in the box provided. Alternatively you can use the navigation bar on the left hand side of the screen.

How do I look at all miRNA targets for a particular genome?

By entering miRBase Targets a list of all the current genomes available will be displayed with a selection of summary statistics. Simply select the genome of choice via the GO button of the right hand side. This will display the main hit list view with all the potential miRNA targets displayed.

Can you search by Gene Ontology terms?

Yes, use the search page, select the desired genome and GO Class, and then enter the term in the box provided.

I'm interested in one target UTR but want to know more details, what should I do?

From the main hit list view, click the GO button on the far right hand side. This will take you to the detailed view for that UTR. Here you can view the alignment and look at general information about the target gene.

What are the coloured bars in the overview of the alignment viewer?

The colours on the sequence overview (large blue and red coloured object) represent sequence conservation where blue represents low conservation and red represents high conservation. The coloured bars above this sequence overview represent the location of the micoRNA hits. Each colour, in each stack, represent microRNA family members. If a stack has more than one family of miRNA hitting the query sequence then the subsequent family /families are represented with a different colour.

Can I load my own miRNA sequences into miRBase Targets to search for potential target genes?

At this initial stage the answer is unfortunately no. However we are actively developing the site with every release and hope to allow this functionality very soon.

Can I load my own UTR sequences into miRBase Targets to search them for potential miRNA targets?

As with miRNA sequences we do not currently offer this functionality, however we do plan to introduce it in future releases. You can use one of the existing algorithms for predicting miRNA targets by running it yourself. The version of the miRanda algorithm used in preparing the mirBase targets is due to be released shortly. The currently available version of miRanda is freely available and well-documented.

miRanda requires two files: a list of miRNA sequences and a list of potential target sequences, both in FASTA format. As explained in the documentation, different parameters allow you to adapt the algorithm to your specific needs.

Can I download the target information?

Yes, we currently support three download formats; these are gff, DAS and SQL. Simply click the Download button from the miRBase Targets homepage and select the genome and format.

Known Bugs

Vertical stripes.

MLAGAN sometimes fails to align one of the sequences in an orthologous group, presumably because a large part of the sequence is masked by 'N's. That sequence is then represented by (repeated occurrences of) the string 'not available'. The display program sometimes handles such multiple alignment incorrectly, necessitating a helper mending part to patch up the display. This results in red patches looking as vertically positioned stripes containing blanks.

Conserved hits on the far left or far right of the multiple alignment sometimes look mangled.

This can happen when a microRNA is aligned with one of the outermost bases of a UTR and there are further bases in the microRNA that virtually align with positions dropping off the UTR. This confuses the display program to such an extent that the display is locally mangled.