Differential binding with SICER/epic2

Question

[i.am.filippov]

I'm trying to analyse AcH3K9 and AcH3K12 ChIP-seq dataset for differential enrichment. I have used epic2-df tool which produces results similar to SICER. For comparison between WT and KO the results contains following columns:

• FC WT vs KO
• p-value WT vs KO
• FC KO vs WT
• p-value KO vs WT

Many people seem confused about the presence of two sets of FC and p-values, including me. What I want to do next is to find correlation coefficient between FC from ChIP and FC of RNA-seq gene expression. I'm not really sure which FC from ChIP-seq should be used in this case: WT vs KO or KO vs WT and how this would affect the results.

Hope the question makes sense.

Update: Volcano plot KO vs WT: https://prnt.sc/q661m3 Volcano plot WT vs KO: https://prnt.sc/q661si

[Nitin Narwade]

I think It should depend on your experimental condition.

I mean, It is just treatment over control comparison.

Now if you consider WT as your normal / control condition then you should use FC WT vs KO, which will be the differential binding of KO over WT.

If you check the results of FC KO vs. WT, it should be exactly opposite of FC WT vs. KO i.e. if a specific binding site in WT vs. KO comparison is having higher FC that will be lower in KO vs. WT comparison.

Thanks.

[i.am.filippov]

Thanks for the reply! That's what I'd expect as well, but plotting -log10(p-value) vs log2(FC) for both produces different plots. I've updated the post with screenshots.

[Nitin Narwade]

Both the scatter plots seem the same to me. I think there is a slight variation in p-value and it is obvious because of sample comparison (A_vs_B and B_vs_A).

[i.am.filippov]

I guess what I don't quite get is why the fold changes are not, how to say, "symmetrical"?

[Nitin Narwade]

Can you post the sample output of epic-df tool (FC values)?