MACS and SAM flags for paired end

[crepaldi]

Hi,

I am analysing paired end sequences from a chipseq. I align with bowtie2 and feed the sam output to Macs.
What are the criteria by which macs excludes/includes alignments?
For instance, as far as I understand, flags for valid alignments are 83 and 99. How does macs consider alignments with "non valid" flags such as 81 and 97?
I could not find a clear answer in the manual.

Thank you for the help,

Luca

[crepaldi]

Following my question:
I have run some tests and it looks like peaks are generated only from reads with the "right" flags (99, 147, 83, 163) and the rest is ignored.
However, I would appreciate if someone could confirm this.
Cheers,

Luca

[harryzs]

Yes, Tao has already confirmed this:

https://groups.google.com/forum/#!searchin/macs-announcement/paired$20end|sort:date/macs-announcement/J9vbZz4s3BM/KKVWRXl24CYJ

----By Tao:
Both MACS14 and MACS2 favor SAM or BAM format for your Pairend data.

MACS1.4: Pair-end data will be read and simply filtered using flags in SAM file, only 5' ends will be kept.

MACS2: Option 1) Similar to MACS1.4; Option 2) MACS2 will bypass prediction for fragment size, and use real fragments during reads mapping to pile up signals ( done by Ben Schiller). The later option can only be found in MACS2 > version 2.0.10.09132012: http://pypi.python.org/pypi/MACS2/2.0.10.09132012. To turn it on, use --format BAMPE.

------