Transcription factor discovery from RNA-seq data; Is it practical ?

Question

[Farbod]

Dear Biostars, Hi ('m not native in English)

I have seen in some papers that via blasting the assembled transcriptome against some TF database, the authors have found the Transcription factor (TFs) of their (non-model) species.

here are my questions :

1) Is this approach (finding TFs from RNA-seq datas) practical ? wouldn't the TF binding sites be in promoter regions and therefore likely not be in the expressed portion that RNA-seq have sequenced?

2) if (1) is practical, which kind of blast (-query is de novo transcriptome assembly) is better ? blastX or TBLASTN?

3) I want to discover the sex-biased TFs (something same as this). I have the DEGs between two sexes. Can I use the Blast and TF database of a model organism and my DEGs as query to discover differentially expressed Transcription factors ?

4) Which TF database is most updated ? AnimalTFDB or REGULATOR OR something else ?

Thank you for sharing your knowledge

~ Best

Answer 1

[Devon Ryan]

1 No and I'd like to see a paper where they actually did this with demonstrable success.

3 You can't find differential transcription factor binding without direct evidence (i.e., ChIPseq). What you can do is to see if there are common transcription factors in one of the databases you listed and then perform ChIPseq with those.

[Farbod]

Dear Devon Hi,

Please have a look at this paper as an example, and search for:

"In order to identify the transcription factors represented in R. sativus transcriptome, all assembled unigenes were searched against the plant transcription factor database (PlnTFDB; http://plntfdb.bio.uni-potsdam.de/v3.0/downloads.php) by using blaxtx with a cut-off E-value of 1 e−5"

[Prasad]

1) Are you looking for TF encoding transcripts (What is there in the paper) or TF binding site? if you want to know which transcript codes for transcription factor follow the paper(results are homology based). if you are looking for later, as @Devon mentioned you need to have Chipseq data.

4) if you are working on plant- PlnTFDB, PlantTFDB and for animals apart from what you have mentioned, DBD

[Farbod]

Dear @Prasad and @Devon,

I am looking for TF encoding transcripts (as I DON'T have CHIP-seq data, but have RNA-seq), so

1- which kind of blast (-query is de novo transcriptome assembly) is better ? blastX or TBLASTN?

2- Does up-regulation of some TF encoding transcripts in one condition (e.g SOX9 in males) show that that transcription factor is upregulated in that condition ?

Thanks

[Devon Ryan]

Oh, you're looking for transcription factors that are up-regulated, not changes in transcription factor binding in genes that are up-regulated. That makes vastly more sense :)

1. blastx, since you have nucleic acid sequence and want to compare against a protein database.
2. It's not proof (if there's an antibody for another species that recognizes it then still do a Western), but it's certainly suggestive. If some of the DE genes are known to be regulated in an appropriate way by a DE TF in other species then all the better. Ideally you'd do some follow up experiments (e.g., ChIPseq (this would be difficult for obvious reasons)), but particularly in largely uncharacterised species I would think that most reviewers would have some understanding of the innate difficulty of expecting too much in this regard.

[Farbod]

Dear @Devon, Thank you for all your supports,

Is there any chance that by this approach (using blast and for e.g zebrafish TF database and my non-model fish transcriptome), any novel TFs will be discovered ?

(If Positive; How ?)

[Devon Ryan]

You're not going to find any TFs that aren't similar to a known TF, since that's how you're defining them. Now that doesn't mean that some of the DE genes won't be novel TFs, it's just that you'll not be able to tell.

[Devon Ryan]

Their analysis is laughably absurd. The reviewers should be ashamed (or the editor should if no biologists reviewed the paper).