ATAC mutisample coverage plot comparision

Question

[fayina.xiao]

Hi,

I used MACS2 to call the peaks of 8 different ATAC samples. I want to visualize coverage of peak regions over chromosomes, and make a comparision of diffirent samples. Because of the different library size of these samples, i think the reads count of peaks should be normalized before covplot, but i can't find any description about reads count normalization in chipseeker manual (http://www.bioconductor.org/packages/release/bioc/vignettes/ChIPseeker/inst/doc/ChIPseeker.html).

So i'm wondering is it nessesary to do normalization between mutiple samples before covplot? if it's necessary, I'd like to do it just as edgeR(calcNormFactors), is it right?

Any kind of help will be great. looking forward to hearing back from you. Thanks in advance, fayina

[ATpoint]

Yes, the norm. factors from edgeR are possible but they need a count matrix. Eventually, the procuedure is a per-million normalization followed by dividing these per-million counts by the norm. factor to correct for compositional changes. Normalization is always neccessary before making any comparison.

[fayina.xiao]

Nice, i had got norm factors for each sample by edgeR, and i'd like to apply it to the macs2 treat_pileup.bdg(no control) result by mutiple this factors to the readcount in this file, then visualize these normalized data in ucsc genome browser.

[ATpoint]

This is not how this factor works. Check the edgeR documentation. The factor is intended to divide the per-million normalized count data, not to multiply it with anything. Why don't you simply make a bedGraph file for each sample, normalized to the reads ber million (e.g. with deeptools bamCoverageusing a bin size of 1 and CPM normalization` and then divide the column 4 (=the coverage) by the edgeR size factor. This is how I get my normalized data for visualization and plotting.