Unable to find flag in SAM with bowtie2 - but can with BWA

[yekwah]

I'm currently using Bowtie2 to map my reads to a small region of a genome to find what region it's in by collecting the unmapped pairs of the reads which have mapped.

I'm currently using samtools flags to get the reads I want:

---a--- ......... -------------
|------------------------gene----------------|
-------..............----b-----
What we want is
a: retain unmapped reads whose mate is on reverse strand (-f 36)
b: retain unmapped reads (-f 4) excluding reads whose mate is on reverse strand or unmapped (-F 40)

samtools view -Sb -f 36 sam > NtermBam
samtools view -Sb -f 4 -F 40 sam > CtermBam

Previously when mapping with BWA both of these commands would work and I would get an N terminus bam and a C terminus bam. However, when I map with Bowtie2 (because the mapping itself is better), I get no reads in the N terminus bam and all the reads go to the C terminus bam.

Does anyone know why this might be? I'm pretty sure the samtools commands are right because it works with BWA. I'm thinking it has something to do with Bowtie2, maybe not recording flags correctly?

[dpryan]

A couple points. Firstly, what was the exact bowtie2 command that you specified? Depending on the flags you used, the unmapped reads may or may not even be included. Secondly, I hope you take the orientation of your genes into account when you start calling the files Nterm and Cterm. They'll be swapped if the gene is on the - strand.

BTW, the most straight forward solution is to just find a couple reads like this from the bwa alignment and use grep to see what bowtie2 did with them. Presuming they map the same, you'll then know if bowtie2 is doing something unexpected with the flags.

[yekwah]

The bowtie2 commands I specified were these:

bowtie2-build ref base_name
bowtie2 --sensitive-local -x base_name -1 fq1 -2 fq2 -S sam

I'm aware that Nterm and Cterm have particular meanings, I'm just not too worried about it at this point. Once I get past this section in my testing I'll make sure the orientation is taken into account.

[westerman]

BTW, the most straight forward solution is to just find a couple reads like this from the bwa alignment and use grep to see what bowtie2 did with them. Presuming they map the same, you'll then know if bowtie2 is doing something unexpected with the flags.

I agree with dpryan -- get a couple of reads that you like from BWA and see what Bowtie2 did with them. The sam/bam format is flexible enough to allow programs to have different opinions on what flags are appropriate. As an aside BWA and Bowtie2 will report different template lengths (column 9) for the same mappings which just shows how 'flexible' (or ill-defined) the sam/bam format is.