Chip-Seq replicates from different experiments

Question

[DavideBrex]

Hi,

I have two replicate Chip-seq samples of H3K27Ac in the same condition. However, they are from two experiments carried out at different times and on different mice. I would like to compare/merge them in order to create heatmaps etc.

The correlation is high between the two replicates. The problem is that I see a big difference in the signal intensity (see the scale of the signal on the left in the image below).

The bigwigs are created by applying coverage normalization (dividing by total number of mapped reads).

How can I handle these two replicates? Any ideas? Thank you!

Answer 1

[ATpoint]

I like to do this for proper normalization respecting different noise levels: ATAC-seq sample normalization

[DavideBrex]

Thank you for your answer. I am trying to understand whether I could apply your method to my data. To calculate the normalization factor, should I do that on the called peaks only (MACS2 output), or can I do it on the whole binned genome (With reads counts per bin)?

[ATpoint]

I typically find that a count matrix based on peaks works best. You could call peaks on all combined bam files, or call individually and take the overlapping peaks, or those that survive IDR.

[DavideBrex]

I got the peaks in common between the two replicates, then I applied featureCounts followed by TMM normalization. I got the scaling factors to generate the normalized bigwigs and now the signal is comparable, thanks so much. I would like to generate a heatmap by merging the replicates bigwigs. In this case, do you think that I could simply merge the bam files from the two replicates and get one single bigwigs per condition? Thanks for your help.