BWA with no mismatches - n=0?

[dsher]

Hello,
We have a sample containing several bacterial species and we want to uniquely map RNA-seq reads to the genomes of each of our organisms to get the expression patterns of each organism separately. We tried to use BWA (implemented in "Galaxy", poor GUI-users that we are) with the “edit distance” (aln -n in the command line version) set to 0 but none of the reads were mapped (all had the SAM tag set to “4’). This is an artifact since running BLAST with some of the sequences showed that they have 100% identity to one of our genomes and not any others, so they should map uniquely.

Does the "aln -n" option really determine the number of mismatches? Any ideas why BWA will not run well using –n=0?
Thanks
Daniel

[xied75]

-n 0 should work at least I tried that before. How long is your reads? When you don't do -n 0, did you get any good alignment, what percentage?

Best,

Dong

[dsher]

Hi Dong,
Thanks for your reply. Our reads are 40bp long after QC. When we allow n=1 we get many millions of mapped reads(~90% of the reads map well), and increasing n from 1-5 increases the number of mapped reads but not by very much, so everything is looking OK - apart from when we use n=0.
Are there any other parameters we should be changing together with n?
Daniel

[xied75]

How about you try to run BWA directly? See what happens.

[dsher]

We might try running it directly but part of the idea is to implement BWA as part of a workflow in Galaxy (we are poor "wet" biologists trying to survive in a world of millions of reads...).

[xied75]

No no I wasn't suggesting you always do this, this is only to troubleshooting where the problem might be, by running directly with BWA you can figure out if this is something to do with BWA or it's GALAXY.

[dsher]

Hi Dong - we don't have BWA installed anywhere as a stand-alone yet. Do you have a standalone BWA installed? If so, if I send you a small file with some sequences and the appropriate genome, is there any chance you can try mapping them with n=0 and send me the results?
Daniel

[xied75]

Ok. You can leave your file over dropbox.

[dsher]

Thanks! What email should I send the dropbox invitation to?

[xied75]

You can send private msg or email via seqanswers in fact.

[dsher]

Thanks Dong - I just sent a private email with the data (we will follow up with the results on this thread). Daniel

[jenhjackson]

Feedback from Galaxy team

Hi Daniel -

As we posted back on the galaxy-user list, the exact problem you were having was a combination of an issue with the BWA binary itself and the design of the Galaxy BWA wrappper.

The original reply and current work-around is available here:

http://lists.bx.psu.edu/pipermail/galaxy-user/2013-March/005981.html

We will be addressing this as a priority. Anyone interested in tracking the progress can follow our Trello development ticket here:

https://trello.com/c/Q1FKwmOs

You uncovered what we believe is a novel issue. Perhaps not commonly run into during command-line usage, but still unexpected. Thanks again for reporting the problem so clearly,

Jen
Galaxy team