Question about BWA mapping quality

[oiiio]

I need to verify that BWA takes into account mate pair data, and the BWA paper that I found seems to be somewhat ambiguous in how it derives mapping quality for these paired reads. If someone could point me at some source code or describe how BWA deals with paired data that would be extremely helpful.

Also, I would feel much happier if I knew how the actual implementation of this statement from the paper worked:

BWA assigns a Phred-like quality score to each read (MapQ) based on match uniqueness, sequence identity, end-pairing, and insert size that is intended to indicate confidence of read placement accuracy in the genome

Again if someone could give a little bit of a more detailed explanation or point me at the source code file (or both) I would be very happy.

[nilshomer]

Check out the "pairing" function in "bwape.c".

[volks]

Check out the "pairing" function in "bwape.c".

which is quite hard to understand. can anyone translate?

[oiiio]

Thanks for the pointer Nils, I'll give it a look. And of course, I'd definitely like to hear any other input people may have as well.

[nilshomer]

which is quite hard to understand. can anyone translate?

Your local bioinformatician would be happy to help and receive the business.

[dp05yk]

Your local bioinformatician would be happy to help and receive the business.

To his credit, the code is kind of a mess, even if the theory behind it is solid.

[nilshomer]

To his credit, the code is kind of a mess, even if the theory behind it is solid.

That may be so, but then whos mess it to clean up?