bwa: Read unmapped in proper pair

[dariober]

Hello,

I came across a bam file with one read having flag 103, which means:
read paired
read mapped in proper pair
read unmapped
mate reverse strand
first in pair

I think there is something weird as a read cannot be "mapped in proper pair" and "read unmapped" at the same time. It wasn't me doing the alignment, but as far as I know it was done with bwa 0.6.1 against hg19. Everything looks fine in the bam file (sort & indexed ok) apart from this read.

This is the incriminated read pair:

1:488660:43	103	chr9_gl000201_random	36144	60	36M	chrM	267	0	CGATGGATCACAGGTCTATCACCCTATTAACCACTC	1:BDEFDHHHHABGH@>DHEEHJJJFHIFBFI>DD9	XT:A:U	NM:i:2	SM:i:25	AM:i:25	X0:i:1	X1:i:0	XM:i:2	XO:i:0	XG:i:0	MD:Z:0G3C31
1:488660:43	147	chrM	267	60	36M	chr9_gl000201_random	36144	0	TCCACACAGACATCATAACAAAAAATTTCCACCAAA	HCD:<BDFGGE:?AAGFFIGHF>BFCA<A=2?B?@?	XT:A:U	NM:i:0	SM:i:37	AM:i:25	X0:i:1	X1:i:0	XM:i:0	XO:i:0	XG:i:0	MD:Z:36

I think the problem comes from chr9_gl000201_random being 36148bp long and the read being aligned to the end of it (36144)?

Am I missing something or is this a bug in bwa? Any thought appreciated!

Dario

[TobyH]

I have seen these strange results from BWA mappings as well, so you are not alone!

It is my understanding that this behaviour is a result of something that happens during the reference indexing step. When indexing a reference file containing multiple sequences (e.g. >1 chromosome), bwa index concatenates the sequences together and indexes this conjoined sequence. Unless I am mistaken (definitely possible!), the inconsistent SAM flags that you get out are when a read pair happened to be mapped in a proper pair on the concatenated reference sequence, but actually - after the reference sequences are split up again - this pairing is split across two references.

It should be clear that this phenomenon is expected to occur only very rarely, so I assume that your single offending read pair is part of a much larger dataset?

[TobyH]

In fact, looking more closely at your SAM, I see that your reads are mapped very close to the end of one chromosome and very close to the start of another.

[dariober]

In fact, looking more closely at your SAM, I see that your reads are mapped very close to the end of one chromosome and very close to the start of another.

Thanks very much! Makes sense... Indeed this is the only read I found with flag 103 out of a few millions in this file.