How does MACS2 treat paired-end reads?

[lamon]

Hi all,
I cannot find good documentation on MACS2 and I do not understand how it treats paired-end data. I can see that I can use -f BAMPE to indicate that the reads are paired but I am not convinced that it is really using that information. I have removed all duplicates by pair but MACS2 is reporting that I still have 38% duplication which is what you would find if you only looked at one end of the pair, not both. Is anyone else having this same experience? I have tried both sorted and unsorted bam files in which the mates appear together.
Thanks!
Lynn

[lamon]

follow-up

Okay, so I can see that it is using the pair information because the peaks seem to span areas that lack reads but which are spanned by the insert length of the pair. Still, this duplication estimate is not correct.
Lynn

[idonaldson]

Hi,

I think the following answer by the author himself might help:
"In BAM mode, only the 5’ end of fragment will be recorded. In BAMPE mode, the 5’ end plus the observed template length will both be recorded so in later analysis, MACS2 piles up the actual entire observed fragment/template instead of estimating a fixed DNA fragment length. Technically, MACS2 will pick only the alignment with flag indicating it’s the first segment in template and both ends have been properly aligned, then read the TLEN value then take the absolute value."

from https://groups.google.com/forum/#!se...E/TUjLeQtsCfkJ

I hope this helps.