Comparison of replicates from NGS : merge before or after peak calling ?

Question

[anais1396]

Hello everyone !!

I have a question for you : I have biological replicates from Atac-seq experience and I want to perform a peak calling and then a comparison of the overlaps between my files. I have 5 samples for healthy people and 9 samples for patients and I have 2 strategies :

• first, I have bam files from alignment and I merge the 5 bam for healthy together and make a second merge for the 9 samples of patients (with samtools merge) and thus I obtain 2 files (bam) and I perform peak calling on each and then search for overlap between the two sets of peaks with intersectBed

• second, I have bam files from alignment (so I have a total of 14 bam files), I perform peak calling on each of the 14 samples and then I make a merge (with bedtools merge) of the 5 peak files of healthy and a second merge of the 9 peaks files of patients and then I search for overlap between the 2 sets with intersectBed.

Which one is the best strategy and why ?? Should I merge before or after the peak calling ?

Thank you in advance !!

Anaïs

Answer 1

[h.mon]

It depends on the sequencing coverage of each sample, but ideally each sample have been sequenced to an appropriate coverage, map and call peaks separately, and then merge peaks. If you have low sequencing coverage per sample (I hope you don't), merge before calling.

For more extensive answers, this issue has been covered before, more than once. Searching for merge peak calling would have led you to several posts, such as:

ChIP-Seq: Calling peaks with replicates

how to pool together biological replicates?

ChipSeq: merge bam file before peak calling

Chip-Seq merging peak files

[anais1396]

How can I simply check the coverage of my samples ?