Hi all,
I have ChIP-seq data sets for two different proteins (let’s call them A and B) in the same cell line that need to form a dimer in order to function as an active TF. Protein A can form either a homodimer (A/A) or a heterodimer (A/B) whereas protein B can only form heterodimers with protein A (A/B) or other proteins that are not of interest right now.
I called peaks with MACS2 and overlapped the peak sets for A and B afterwards using mergePeaks.pl from the HOMER tool to get peak files of regions that are only occupied by A or B or by A+B. Our assumption is now basically that regions only occupied by protein A (peak file A) represent in some way binding sites bound by the homodimer (A/A) and overlapping regions bound by protein A and B are binding sites possibly bound by the heterodimer (A/B). Is this assumption in some way acceptable or are we on the wrong pathway here?
Also, I thought about centering the peaks to the identified de novo motif and viewing the coverage from the two sets surrounding the centered motif. With this, I wanted to know if there is a slight shift in the coverage distribution depending on if a homodimer or heterodimer is binding (at least for the protein A which forms homo- and heterodimers). I don’t know if the question is understandable or makes sense but I would appreciate any recommendations/literature or comments on the idea of the strategy!
Thanks in advance!
Seems like a reasonable starting assumption. Are there any known, characterized targets of the A homodimer or heterodimer reported in the literature that you can use as a sanity check for this assumption?
Thanks for the reply! There a few target genes but as far as I know they always only took a look at A and not on any other partner. So it’s hard to say from their results wether this was the action of a homodimer or a heterodimer in my opinion (this is why we wanted to take a closer look). Also, according to my motif results and literature the binding motif for A/A and A/B seems to be more or less identical. So I would guess that a simple experiment like EMSA with A and B will give us no big insights into when a homodimer or heterodimer is binding to a specific region. I believe that this depends more on things like genomic context or potential co-factors rather than on the simple sequence. Therefore, I am hoping to get more information about this idea from the ChIP-seq too.
Adding that this also sounds like a decent starting point. You might want to see if any of the peaks are shifted by 1-10 bases between the A and B ChIP-seq datasets as they might bind downstream of each other, e.g. A-B
Thank you too for the suggestion! Have you ever done something comparable with "normal" ChIP-seq data? I tried to find some peaks that are shifted in the two datasets but I guess my resolution is simply too bad to make any statements regarding a shift in a range of 1-10 bp. Do you have any idea how one could filter the data or modify the analysis to get more information out of this or would you rather recommend performing another protocol to get better resolution such as ChIP-exo? Sorry for this type of question but I'm still relatively new to this field and I have the feeling ChIP-seq data are a little bit more tricky to handle especially as they are so noisy.
Yeah traditional ChIP-seq doesn't have the best resolution. I think your best bet is looking at positive control genes and coming up with some sort of "correction factor" to be able to compare the peaks from A and B.
I've heard through the rumor mill that ChIP-exo as a technique hasn't been the most reproducible method outside of the Pugh lab. Hopefully more resources have been generated since then.
Have you considered looking at some type of antibody blotting to see if you find bands approximately double in weight as the protein A (to confirm homodimerization)?