Best tools for miRNA target prediction

Question

[simone.castellana]

Hi,

I have sequencing data for microRNA on common bean (Phaseoulus vulgaris), all stored in fastq format.

I have used sRNAtoolbox to identify the miRNAs; now I would like to understand which might be the target of this miRNAs in the common bean genome.

Can anyone suggest me which are the best tools for miRNA target prediction?

Thanks!

Answer 1

[Carlo Yague]

You can start by reading this paper: Tools for Sequence-Based miRNA Target Prediction: What to Choose?

The authors highlight three of the most mature miRNA target prediction tools: miRanda, targetScan, and DIANA tools. I have personaly used miRanda recently and it worked great for my purpose. It is very easy to set up even with non standard model organism (like common bean). Just input the reference transcriptome and the miRNA previously identified, and miRanda gives you all putative targets.

[simone.castellana]

Thank you! I ran miRanda, but the output can get really big. I used my known miRNA and looked for putative targets in the trascriptome of P. vlugaris and i got my output file, which mapped my miRNA to tons of different sequences.

Do you have any suggestion on how to process the output file in a nice way? Like to extract only matches with good scores, or something like that. How you process it usually?

Thanks again for the support!

[Carlo Yague]

So each output entry looks like this (at least it did for me) and there were many of them:

   Forward: Score: 170.000000  Q:2 to 19  R:3559 to 3579 Align Len (17) (82.35%) (88.24%)

   Query:    3' aguAGUCGUCAUCAUAGUUAu 5'
                   |:|||  |||||||||| 
   Ref:      5' tatTTAGCTATAGTATCAATg 3'

   Energy:  -18.719999 kCal/Mol

Scores for this hit:
>miRNA_x    gene_y  170.00  -18.72  2 19    3559 3579   17  82.35%  88.24%

Seq1,Seq2,Tot Score,Tot Energy,Max Score,Max Energy,Strand,Len1,Len2,Positions
>>miRNA_x   gene_y  170.00  -18.72  170.00  -18.72  186 21  3596     3559
Complete

To make a clean table out of it, I used grep to recover only the last line, starting with ">>". Then I sort by score (column 5). This gives me a nice table easy to parse in R, with the best hits on top.

grep ">>" miranda_output.txt | sort -k 5 -n -r > miranda_clean.txt

You can also use the -sc option in miRanda to limit the number of hits based on score. For instance with -sc 150 you only get hits with score higher than 150.

[simone.castellana]

Thanks, that was really helpful! I managed to get the output as you suggested. What if I want to have the alignment as well in the file I'm going to export? There's a way to do that? Sorry for all of the questions!

Thanks again

[Carlo Yague]

Well, the alignment takes multiple lines so it would be difficult to make it fit in a table with one hit per line.

[simone.castellana]

I have one last question. Could you help me out with the interpretation of the output from Miranda? I’m still the learning process so it’s still a little bit cryptic for me what to look at exactly to find the best predicted target.

Thanks again

[Carlo Yague]

In general, the lowest free energy ("Max Energy") reflects a more stable/specific interaction between the miRNA and the target. The higher score ("Max Score") is also desirable in general, but you will have to read the paper to know exactly how this score is calculated.

In my opinion, "Tot Score" and "Tot Energy" are not that interesting to look at. As best predicted target, I would rather choose a mRNA with only one region strongly targeted by a miRNA instead than a mRNA with multiple region weakly targeted by the same miRNA.

[simone.castellana]

Thank you for your support! You've been extremely helpful