Searching For Mouse Regulatory Motifs

Question

[Andrew Su]

Question: The last time I played with transcription factor motif searching (many years ago) I used the TRANSFAC database of motifs combined with the MATCH tool at gene-regulation.com. Has the state of the art progressed? I'm particularly interested in web servers, web services, or R packages to do this analysis.

Background: A collaborator approached us wanting to study the transcriptional regulation of his favorite gene. Specifically, he wanted to identify key regulators that bind in the ~100 KB upstream region. Because he wanted to cast such a wide net, we started by providing the ~2 KB of sequence (spread among four regions) with the highest phylogenetic conservation as calculated in UCSC's conservation track.

Based on these results, he generated transgenic mice with various combinations of these conservons knocked out, and many of these mice had dramatically altered expression patterns. The next step of course is to identify the specific regulators binding to these regions.

He is currently making additional mice that refine the 2 KB of sequence into smaller chunks. In parallel, we'd like to use bioinformatics to identify candidate binding sites and their corresponding regulatory proteins.

Edit: To clarify, there are many tools that take many coregulated genes and find enriched motifs, but these tools do not really address my particular need. I'm interested in the regulation of exactly one gene, and I'm interested in identifying candidate binding motifs corresponding to known TFs in that gene's upstream genomic sequence.

[Khader Shameer]

Thanks a lot for your excellent follow up with the question and answers. IMHO, this should be cited as one of the best interactive Q&A of BioStar.

Answer 1

[Neilfws]

Has the state of the art progressed? Not much. I think the only difference that you'll find from the days of TRANSFAC is that there a few more (open) database alternatives - particularly JASPAR - and a few more webserver tools. Few if any of the latter will have a web service, or even provide results to download as ASCII text. To my knowledge, there is no useful R package for this task.

We have used ConSite and the JASPAR website - to scan sequence at the latter, follow the link to the vertebrate core database. You could also look at the MEME suite; it is more geared towards de novo motif discovery, but has tools to scan for known motifs, provided that you convert them to the required format.

[Andrew Su]

In addition to providing the database of matrices, the scanning tool that JASPAR provides is certainly among the easiest to use. However, they note themselves the "abysmal selectivity" of their tool, and point the user to ConSite. Unfortunately, the ConSite site is down at the moment... I've emailed the administrators and will (hopefully) post another comment on ConSite once it's back up...

Answer 2

[Fred Fleche]

[?]

[?]

http://homes.esat.kuleuven.be/~saerts/software/toucan.php

[Andrew Su]

Gave TOUCAN a shot and I actually like it quite a bit! I was a bit wary of the webstart, but now I'm a believer. FASTA input was pretty straightforward, MotifScanner allows automated retrieval of several PWM databses (JASPAR, Transfac, etc.), the GUI is pretty nice for viewing the output, and the GFF export allows for easy downstream parsing. Two thumbs up here!

[Andrew Su]

After taking a look at all the great suggestions, I'm accepting this one as the answer -- nice GUI, easy access to TRANSFAC and JASPAR, many pre-built background sequences for motif scanning, nice visual display of results, and GFF export. Thanks Fred (and all)...

[dli]

link seems not work anymore?

[Neilfws]

So web search to track it down; here it is -http://med.kuleuven.be/lcb/toucan.php

[dli]

Thank you very much :)

Answer 3

[razor]

This is an overview of some of the tools available. "Assessing computational tools for the discovery of transcription factor binding sites" http://www.ncbi.nlm.nih.gov/pubmed/15637633. Most of them are command line tools, but there are web server versions for some of them too.

[Andrew Su]

Pretty lengthy list of tools that were compared here, but most fall into the class of tools for de novo motif discovery (as opposed to scanning known motifs). Interesting read though, thanks!

Answer 4

[Khader Shameer]

I think ORegAnno will be a good start. List of mouse TFs and TFBS are available here. If you are interested in regulatory regions in non-coding regions you may check Enhancer Browser or RSAT.

[Andrew Su]

ORegAnno appears to be a database of regulatory elements, but it doesn't appear to have a tool to search those elements in sequence. Also, it appears that JASPAR is a more commonly used database of binding motifs, no?

[Andrew Su]

VISTA's Enhancer Browser appears to contain only human elements -- potentially good but not directly useful for my mouse example...

[Andrew Su]

Trying RSAT's matrix-scan which appears to be the most promising so far. The "scanning options" don't really make sense to me at first glance, so I hope the defaults are reasonably chosen...

[Khader Shameer]

Thanks for the feedback. I suggested ORegAnno, because I have used the TFBS information from ORegAnno to search in upstream of several genes of interest using hmm. True there is no search program available at ORegAnno, but you may perform a pattern search using hmm or other approaches using TFBS reported in ORegAnno.

[Khader Shameer]

I suggested Enhancer because the website explicitly mentions that "The VISTA Enhancer Browser is a central resource for experimentally validated human noncoding fragments with gene enhancer activity as assessed in transgenic mice" . I thought that means these elements must be conserved across the species.

[Khader Shameer]

I didn't used JASPAR before, but on a quick look I can see that ORegAnno reports more TFBS than JASPAR. As suggested Neil, you could use MEME to search for the TFBS from OregAnno in the upstream of your gene of interest.

[Andrew Su]

Great! thanks for the follow up... Good points all...

[Khader Shameer]

I would also like recommend you to take a look at one of my paper where we looked at the upstream of few arabidopsis genes involved in abiotic stress response using HMM models derived from known TFs. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2561162/

Answer 5

[Darked89]

Even restricting yourself to highly conserved regions, 100kb is a lot to search for motifs.

Check:

Prediction of over Represented Transcription Factor Binding Sites in Co-regulated Genes Using Whole Genome Matching Statistics. by Pavesi. Link

After his seminar I updated this page (see the top):

EDIT

I did not use it yet, but Pavesi has a PSCAN server

[Andrew Su]

Looks interesting, but ideally I could do this one-off analysis through a web interface, or alternatively a web service or an R package.

[Andrew Su]

Looks interesting, but ideally I could do this one-off analysis through a web interface, or alternatively a web service or an R package. I'm not quite so motivated to implement a published algorithm. (Added this context to the question above...)

[Darked89]

I hope there is no need for reimplementing the algorithm. See PSCAN link in the edited answer above. One-offism is also what I often do :).

[Andrew Su]

The PSCAN site looks good (thanks!), but it is geared toward a slightly different use case than what I'm looking for. See my comment to Greg Tyrelle's answer and my edit above...

Answer 6

[Greg Tyrelle]

There are many tools that will take sequence and look for TFBS based on various searching approaches e.g. PWM. In fact too many. I will second the recommendation for TOUCAN if you want a GUI tool. I had the need to integrate this kind of search in to a processing pipeline. For that I used TAMO, which is a python toolkit. I can recommend this if you like python. There is now a web based tool, webmotifs, that is based on TAMO, so you're covered if you want a one-off analysis. On the commercial side Biobase have good integrated web based tools for TF searching and analysis. The interfaces are a bit of a mess (organic growth of the product), but all the tools are there.

[Andrew Su]

Thanks for the links. TAMO and webmotifs look like they are targeted toward finding enriched motifs based on a putatively coregulated set of genes, but I'm interested in finding instances of known motifs in the upstream region of only one gene. Slight different use case in the general space of promoter bashing. I've added some clarification to my question above.

[Greg Tyrelle]

TAMO will do motif search and discovery.

[Andrew Su]

Greg, thanks for the follow up. I tried entering one refseq on this page (http://fraenkel.mit.edu/webmotifs/form.html) and get this error: "You requested motif discovery on fewer than 10 sequences. WebMOTIF only supports motif discovery on sets of 10 sequences or more."

Answer 7

[Ejm]

Maybe worth checking out Biostrings for doing this in R? It can do some PWM manipulations, including matching, and downloading PWMs you need from somewhere (i.e. JASAPAR) and getting them in there should be pretty straightforward.