MicroRNA-target interactions databases

Question

[Joe Kherery]

Hello everyone,

I have a list of 30 microRNAs and I want to build a network of interaction with the targets.

Well, I'm thinking about use these databases:

1. TargetScan Release 7.2
2. mirTarBase v6.0
3. enter link description here
4. microT-CDS 5
5. miRWalk 3

And as a cut-off the target has to be present in 3 of these databases and have at least two microRNAs as targets.

First point: Does this approach look robust?

Second point: miRWalk 3 seems to use other bases (miRBase, TargetScan and miRDB) as Resources for miRNA-target data http://mirwalk.umm.uni-heidelberg.de/resources/ Am I right? If so, it seems redundant to use it here in my approach.

If anyone can give me some advice on how to do this analysis I will be very grateful, I have read several articles and each author does this analysis in a different way.

[Björn]

@kevin Blighe Nice thread..how to create the two graphs that you showed here. For example, I have two miRNAs: hsa-let-7a, and hsa-miR-4511

[Kevin Blighe]

Hello, the code is somewhat complex. In which plot are you more interested? I can post a tutorial over the weekend.

[Björn]

Hi, I am interested in the first one. Thanks

Answer 1

[Kevin Blighe]

Yes, that's the conclusion that I got, too, i.e., that each author does it a different way. You know what? - I also did it a different way and here's what I did:

Given that most of these databases provide in siico predictions, and in some cases have come under scrutiny and criticism, the more corroborative evidence that you can accumulate for each mir, the better. There is an R package that does this for you, called miRNAtap. Whilst it doesn't have all of the databases that you mention, it does look at:

• DIANA (Vlachos et al., 2012)
• PicTar (Krek et al., 2015)
• TargetScan (Agarwal et al., 2015)
• miRanda (Betel et al., 2008)
• miRDB (Wong & Wang, 2015)

The good thing here is that you can automate it and say that you want evidence for an interaction in at least 3 of these databases, or even 5 (all). miRNAtap also let's you to do gene enrichment of the genes targeted by each of your mirs of interest, whilst in addition the tutorial to which I've linked also shows how you can perform KEGG pathway enrichment of these, too (using KEGGprofile).

---------------------------------------------

miRWalk, I believe, was the database that came under some scrutiny in the past, but they appeared to then improve the data held in it. For one, they actually have validated mir-to-mir, mir-to-gene, etc interactions, i.e., from functional studies. In my study (yet to be published), I used miRWalk as a sort of secondary validation step for anything of interest found in the first part using miRNAtap.

---------------------------------------------------

After you have identified some key mirs and gene targets, you can develop some nice figures like a mirPrint (a term that I chose) and a graph (yet unnamed) that shows the key mirs and their gene targets, and also the layering of the graph indicates how many mirs target the same gene.

Just to give you some ideas.

Kevin

[Joe Kherery]

Dear Kevin,

Many thanks for the wonderful answers. As you rightly said there is no "right" or "wrong" way of doing this. I'm new in bioinformatics and R, So I did it all manually, there were few miRNAs (30) and made a merge using the venn diagram. So I import to cytoscape and construction my network.

So can I continue to use miRWalk within this kind of analysis I've described?

[Kevin Blighe]

Hey Joe, no problem. Yes, well, it looks like mirWalk has already been upgraded yet again since I used it (I used version 2; you mention version 3). I believe that the issues that I mentioned were related to the first version. It is very well cited in the literature and comes from a reputable organisation, so, feel free to use it.,

If you wanted help to generate these plots that I mention, then I can perhaps help. The mirPrint was built using functions in the ComplexHeatmap R package, whilst the other plot was generated using iGraph.

[Joe Kherery]

Hey Kevin, Yes it was last updated in Apr / 2018. I already built an interaction graph using cytoscape, it was similar to yours. As a beginner, I try to use the easiest tools. And about the ComplexHemapmap R package, But I do not know if I will have much time to use it. My time to finalize this data analysis is short.

Thank you very much.

Answer 2

[Sasha Fokin]

Hi, Joe Kherery!

We are collecting a similar database and creating an interactive tool for visualizing regulatory networks. But only for a Homo Sapiens.

ONCO.IO is an integrated knowledge database and network analysis tool for exploring complex gene regulatory networks. Database content is based on the manually curated annotation of experimentally verified molecular interactions described in the literature and includes microRNAs, lncRNAs, mRNAs, signaling proteins and transcription factors interactions.

[Kevin Blighe]

Looks useful - thanks Sasha!