macs2 normalization on broadPeak

Question

[firatuyulur]

Hi,

I have a number of atacseq bams and their total read numbers differ. I did peak calling with and without --SPMR for normalization and there is no difference between the two ......broad_peaks.broadPeak files. yet there is a difference in the resulting two ....broad_treat_pileup.bdg files. While using callpeak, does --SPMR have no effect on the signalValues? If that is the case, should I do another peak calling from the bdg files?

My end goal is to see what happened to some specific regions, whether they got condensed or decondensed and I cannot make sure whether this read depth difference effect the signalValues.

Answer 1

[James Ashmore]

Whether you use the SPMR flag or not, MACS2 will normalise the data internally to call peaks. The SPMR flag only effects the signal track produced. With the flag present the signal will be normalised to reads per million, this is for comparison with other samples which have been sequenced to different read depths.

[firatuyulur]

So what you are saying is the signalValues that I see in broadPeaks file are independent of the read depth as those are the values obtained from a normalised peak calling and I can compare two samples according to their signalValues although they have different read depths?

[James Ashmore]

The signalValue in the broadPeak file produced by MACS2 is not a measure of how many reads are aligned within the peak. It is a measure of fold-change above background.

[firatuyulur]

Isnt fold change related to the number of reads aligned to the peak?

[firatuyulur]

James thanks for the help. After this quick conversation I started looking at the regions of peaks on igv. I put the spmr called bedgraphs to igv and checked broadpeaks beds by eye. Still cannot strongly argue about it, I suppose you are right about peak calling normalizes while doing its job.