Mismatch options in bwa

[alexbmp]

Is it right, that the -n option of "bwa align" is restricted to only one of the paired-end reads?

I """think""" that I have found the fact that, when I give -n option, the option is set to only one of the reads.

Does anybody know how to set -n option to both of the reads?

I think the problem rises when I use "bwa sampe" to pair reads; bwa probably pairs without regarding the previous options given in "bwa aln", such as the -n option I mentioned.

[CTGF]

Yes, when you do a paired-end mapping, bwa sampe will force a non-optimal alignment regardless of the mismatch rate you set during the initial alignment.

[alexbmp]

Yes, when you do a paired-end mapping, bwa sampe will force a non-optimal alignment regardless of the mismatch rate you set during the initial alignment.

I got this quote from lh3:

http://seqanswers.com/forums/showthread.php?t=3017

lh3 mentions that though XM:i:# can seem nonoptimal, it actually isn't;
lh3 says that such "seemingly nonoptimal" reads are the result of pairing to semi-repetitive regions.

I BLATted some of the former suspicious reads, and they all turned out to be mapped in repetitive regions.

So I decided to ignore XM:i:# when pairing.

How do you think?

[CTGF]

In my case, if I use paired end mapping by bwa, more than 90% of the reads align to the reference genome; while if I map two ends individually, the alignment rate drops to 80%. So I think the paird end mapping is not that stringent.

[shoegame2001]

So does anyone know how to change the settings so that both reads in the read pair are required to have fewer than n mismatches? It was not obvious to me from the user manual...