How Many Transcription Factors Are There With Known Transcription Factor Binding Sites?

Question

[D W]

According an the article (DOI:10/.1126/science.1066355) there are 2000 to 3000 transcription factors. How many of them have a known TFBS (transcription factor binding sites) in humans. This answer could be derived bioinformatically or by a citation.

[Neilfws]

Just to clarify, "derived bioinformatically" should mean "using a database of known TFBS". Predicted binding sites are not necessarily real binding sites.

[Jarretinha]

UHU! Tough question!!! No bounty?

[D W]

I have no ability to provide bounties sorry.

[Michael]

Agreed Neil, I think only experimentally verified sites can count as 'real' TFBS.

[D W]

So the value of 150-300 are values of predicted binding sites but the experimentally verified sites are lower? I guess the next question is how many of the 150-300 are experimentally verified?

[Casey Bergman]

I think this question needs to be posed more tightly to be answered properly. Which binding sites count: only those in the the genome or also those identified in vitro selection? Also, do data from footprinting, EMSA, mutational analysis, and/or Chip all count as valid TFBSs? Finally, are you asking how many sequence specific TFs have TFBSs or how many of all know TFs have a known TFBS (including TFs that do not have sequence specific action, e.g. histone modifying factors). Finally, and most difficult to answer, what is your definition of a TF?

[D W]

The context of this question is what would you say if you were asked this in a thesis defense? I was not, but told to be prepared for questions like that. All of the follow up questions you asked @Case Bergman are useful.

Answer 1

[Will]

I would check out http://jaspar.genereg.net/. It has collections of known transcription factor binding sites. I didn't actually count them but it looks like around 150-300.

Answer 2

[Farhat]

If I look at the TRANSFAC release dated Mar 2010, there are 1300 TFBS matrices given. This includes minor variations like V$P53_03 V$P53_04 V$P53_05 and so on. If I restrict my search to vertebrates the number is 908. If I ignore the part after the '_', the number works out to 598. While this is vertebrates and not just humans you could get some idea of the numbers involved.

Answer 3

[Larry_Parnell]

I have 245 human transcription factors with motifs described in TRANSFAC and/or JASPAR.

I also use data published by Xie, et al (Xie, Z., Hu, S.H., Blackshaw, S., Zhu, H. and Qian, J. (2009) hPDI: a database of experimental human protein-DNA interactions, Bioinformatics. (In press); http://bioinfo.wilmer.jhu.edu/PDI/) which gives sequence motifs for known and non-traditional DNA-protein interactions. From those data, I extracted 1015 genes.

[Casey Bergman]

+1 for PDI. Haven't seen this yet, very cool resource.

[Larry_Parnell]

Thanks! I heard one of the authors speak at RECOMB satellite conf in 2009. For me this is an interesting resource that has value for a given gene on a case by case basis. In other words, I am looking for other data on the gene before I put a lot of value on the PDI data. Nonetheless, their data make one think about the fine tuning of transcription based on a multitude of inputs from proteins with various functions.

Answer 4

[D W]

According to this site the answer is around 130: http://oreganno.org/tfview/cgi-bin/specieslist.pl

Does that seem reasonable? Are there different answers?

Answer 5

[Maximilian Haeussler]

The data is spread over at least three databases: Transfac, Jaspar and UniProbe. You could drop Transfac, it's a rather strange datasource. You would need to parse the (human) gene identifiers from Jaspar and Uniprobe and then get the union of them.

[Farhat]

I am curious about what makes TRANSFAC strange?

[Maximilian Haeussler]

Curated from the literature, so can pick up any mistake from the original authors. Unless the others don't have any data, I would avoid data from Transfac (apart from the fact that you have to pay for it to use it). So I guess you have a license. :-)