bowtie: mapped read not in paired alignment

[bowtietester]

Hi,
I have a problem with bowtie and paired end reads:
A read is being reported as unmapped because its mate can not be mapped.
What am I doing wrong?

I have a test genome:

>chr1
AAGCTGGGATCGATGCCACGTAGACGTAGTTTAGTGAAACTGATGATTCCCTGATAGTAGCTGACTGATGGGGCAGTAAAAGTCCCCAGTCAGTTGACCTGAC

and two test paired read files with reads taken from genome (m1a,m2a,m1b) and mutated/not matching m2b:
read1:

>m1a
TGGGATCGATGCCACGTAGACG
>m2a
TGGGATCGATGCCACGTAGACG

read2:

>m1b
GGGCAGTAAAAGTCCCCAGTCAGTT
>m2b
GGGCAGTTTTTGTGGGGAGTCAGTT

When I align each file separately I get expected results (3 matching, one not):

@HD     VN:1.0  SO:unsorted
@SQ     SN:chr1 LN:103
@PG     ID:Bowtie       VN:0.12.5       CL:"bowtie/bowtie-0.12.5/bowtie -f -S genome r1.fasta"
m1a     0       chr1    5       255     22M     *       0       0       TGGGATCGATGCCACGTAGACG  IIIIIIIIIIIIIIIIIIIIII  XA:i:0  MD:Z:22 NM:i:0
m2a     0       chr1    5       255     22M     *       0       0       TGGGATCGATGCCACGTAGACG  IIIIIIIIIIIIIIIIIIIIII  XA:i:0  MD:Z:22 NM:i:0

@HD     VN:1.0  SO:unsorted
@SQ     SN:chr1 LN:103
@PG     ID:Bowtie       VN:0.12.5       CL:"bowtie/bowtie-0.12.5/bowtie -f -S genome r2.fasta"
m1b     0       chr1    71      255     25M     *       0       0       GGGCAGTAAAAGTCCCCAGTCAGTT       IIIIIIIIIIIIIIIIIIIIIIIII       XA:i:0  MD:Z:25 NM:i:0
m2b     4       *       0       0       *       *       0       0       GGGCAGTTTTTGTGGGGAGTCAGTT       IIIIIIIIIIIIIIIIIIIIIIIII       XM:i:0

When I align them as paired reads, suddenly m2a does not match anymore. I would have excpected it
to be reported as m2a mapped but mate (m2b) unmapped:

@HD     VN:1.0  SO:unsorted
@SQ     SN:chr1 LN:103
@PG     ID:Bowtie       VN:0.12.5       CL:"bowtie/bowtie-0.12.5/bowtie -f -S -m1 --best --strata --ff genome -1 r1.fasta -2 r2.fasta"
m1a     67      chr1    5       255     22M     =       71      91      TGGGATCGATGCCACGTAGACG  IIIIIIIIIIIIIIIIIIIIII  XA:i:0  MD:Z:22 NM:i:0
m1b     131     chr1    71      255     25M     =       5       -91     GGGCAGTAAAAGTCCCCAGTCAGTT       IIIIIIIIIIIIIIIIIIIIIIIII       XA:i:0  MD:Z:25 NM:i:0
m2a     77      *       0       0       *       *       0       0       TGGGATCGATGCCACGTAGACG  IIIIIIIIIIIIIIIIIIIIII  XM:i:0
m2b     141     *       0       0       *       *       0       0       GGGCAGTTTTTGTGGGGAGTCAGTT       IIIIIIIIIIIIIIIIIIIIIIIII       XM:i:0

[Ben Langmead]

Hi,

Bowtie's paired-end mode is only for finding paired-end alignments. If you're after both paired and unpaired alignments for the same set of reads, you can run Bowtie in paired-end mode with --un <file>, then run bowtie in unpaired mode with <file> as the input.

Thanks,
Ben

[bowtietester]

But the SAM format is complex enough to capture this information. Would you consider implementing this if one files a bug report?
To align the same input three times (paired, read1, read2) and then combine the outputs seems like a detour to me.

thank you very much for the quick answer,
ido

[Ben Langmead]

I bet it's already in there as a feature request, but feel free to file one if it's not. Please don't file as a bug.

Thanks,
Ben