How to correlate genes to transcription factors in RNA-Seq data?

Question

[R.Blues]

Hello everyone,

I have a set of differentially expressed genes in an "A" tissue when compared with other tissues, most of these genes being typical protein coding genes, and the rest of them being transcription factors. I have obtained this set from RNA-seq data from several samples of this "A" tissue and the rest of the tissues. This far, I can assume these typical protein coding genes are regulated by the transcription factors.

However, my problem is, is there any way or tool to know which genes are regulated by each transcription factor without using any database or experiment, but only my data?

Thank you very much for your help and kind regards.

[GouthamAtla]

Get the list of genes regulated by the Transcription factors and overlay them with the Differentially expressed genes from RNA-Seq data. But I do not think if there is any tool for your specific problem.

[komal.rathi]

You could do something like this:

1. Use Cytoscape to map your list of genes to available interaction networks (like Biogrid available from within Cytoscape). You can visualize if the TFs are directly or indirectly interacting with other protein-coding genes in your list.
2. You can calculate a simple correlation coefficient between all the genes based on the RNASeq data that you have. You can then overlay this information on the interaction network that you build in step 1.
3. Color the edges (relationship) of the interacting nodes (genes) based on high/low correlation. There is an option where you can set a color gradient for e.g. dark green -> dark red for high positive correlation -> high negative correlation.

Cytoscape is a little tricky. You will have to play with the options & read tutorials to get the hang of it.

Answer 1

[Devon Ryan]

If we rule out using any database (I don't know why we'd do that, using a database of TF binding sites would be convenient), then one possibility would be to look at covariation between the various DE genes (e.g., with WGCNA). The assumption would be that a TF driving a bunch of genes to be DE should be near the center of a module containing them and possibly be its eigengene.

[R.Blues]

Thank you very much! I ruled out using any database because I am trying to apply the method to other data too, and this data is not well annotated. WGCNA seems quite complex, but I'm sure it is worth a try.