compare ChIP-seq datasets in two cell lines

Question

[Bogdan]

.Dear all,

would you please advise, shall we do a comparison between two different ChIP-seq datasets,

that have been generated in two cell lines,

for example, H3K4me3 in HEK293 and H34me3 in IMR90,

in order to show that the data in HEK293 is better than in IMR60, which metrics would you choose ?

To keep in mind that the loci marked by H3K4me3 in HEK293 is very different than the sets of loci marked by H3K4me3 in IMR90.

Thanks a lot,

~ Bogdan

[seidel]

Do you have replicates in each cell line? You could look at replicate reproducibility, and examine IDR (Irreproducible Discovery Rate). What software did you use to call peaks? You could examine the distributions of Q-values and fold enrichment values for peaks between the two. Do you have any specific criteria you would like to use to define "better"? Or is it simply: lower p-values, better enrichment, better peaks. (?)

[Bogdan]

Thank you both for your answers and questions.

Let's consider this time the case of H3K27me3. :

<> in HEK293, we do have set of peaks SET1, that poorly overlaps with the set of peaks in IMR90 cells, i.e. SET2

<> the signal of H3K27me3 looks higher in HEK293 versus IMR90, by inspecting specific 3-4 regions.

Our goal is to show that the signal of H3K27me3 is higher in HEK293 versus IMR90 on a GENOME WIDE scale instead of picking specific regions. How shall I do that ? Thanks a lot..

[ATpoint]

Isn't that a classical example where a Western Blot would have down better quantitative job than NGS?

[Bogdan]

Yes, although we do not have access to the WB data. All we have is CUT&RUN data.