removing reads that map to more than one location from gsnap-aligned bam files

[efoss]

I would like to remove reads that map to more than one location from bam files I created using gsnap. Does anyone know how to do this? I would prefer not to have to realign. Also, does anyone know what MAPQ 0 means in files that have been aligned using gsnap? As I understand it, the meaning of MAPQ 0 can change depending on which aligner was used to generate the bam file.

Thanks.

Eric

[Richard Finney]

see here : https://www.biostars.org/p/56246/

samtools view -bq 1 file.bam > unique.bam

[efoss]

Fantastic! Thanks so much, Richard Finney.

As I understand this command, it's saying to filter out anything with a mapping quality (MAPQ) score that is less than one and output that as a bam file. Is it true that, regardless of which aligner you use to create your bam file, a read that maps to more than one location will have a MAPQ score of 0?

[Richard Finney]

Not necessarily.
The tags (I think) are optional and not all alignment programs go the extra mile to make sure the tags are thorough.
I'm not sure how orthodox GSNAP is on this matter; you may wish to view the sam output tags to make sure they're what they should be.

[Brian Bushnell]

Typically, a read that maps to multiple locations with a similar (internal) score will get a mapq of 3 or less, as 3 indicates at most a 50% chance that a given alignment is correct. But it varies greatly by aligner; some will always give mapq 255 for any mapped read, for example.

More importantly, even if an aligner does assign a read to multiple locations, they are not necessarily equivalent; the primary might be much better than the secondaries (and as such perhaps get a mapq well above 3). There is not a simple, universal way to ensure that you remove all reads from a sam file that map to multiple locations when processing it as a stream without tracking names to see how many times they occur, though you could make this process efficient if the file is sorted by name.

It's trivial to filter out all secondary alignments, though, with samtools.