Hello All,
When running bwa aln the output sam file contains a series of bwa specific fields, one of which, XA, describes alternative
alignment locations for the given read.
I have a couple of questions relating to how bwa reports the multiple alignment results.
Firstly:
Is the alternative location in anyway relative to the preferred (by preferred I mean the alignment position in the
compulsory sam fields) alignment location?
To expand, imagine the preferred alignment position, alignment A, has an edit distance of 1. Is it true that any alignment, alignment B, in XA will have an edit distance of >=1. i.e bwa will choose the best (lowest edit distance) alignment and report in the preferred position, and all alignments in XA will be scored worse or equally to that in the preferred position.
Finally:
Given a set of equivalent (identical edit distance) alignments. How does bwa determine which alignment enters
the preferred position, and which alignments enter XA?
To expand, all important things pertaining to alignment quality being equal (I say this to capture other factors apart from edit distance) how does BWA pick between such equivalent alignments? Is choice based on alphanumeric ordering of the RNAME field, or is it effectively random? I'm not suggesting that bwa has a random generator for such events, but rather, the implementation does not care about such a case and ordering could be based on some non-determinstic factor - like file ordering etc.
Thanks,
Izaak
Do you have had any answer for this yet? I am also wondering.... (my impression is that XA represent eqaully optimal alignment and the assignment is random)