bowtie map only first 50 nt of reads

[jgibbons1]

Hi all,

My Data:
Illumina whole genome data from ~160 individuals. Reads were initially 100 nt but after quality trimming I have a distribution of read lengths between 50-100 nt.

My Problem:
I am interested in regions of the genome that are not annotated. The loci are very small (~80 bp). Because some of my reads are larger than my reference, I can not map them against my reference loci. My solution was to only map the first 50 nt of my reads. However, I can't figure out an efficient way to do this. Bowtie1 has an option to trim bases from the 5' or 3' end of the read (-5/--trim5 <int> and -3/--trim3 <int>), but instead of trimming off bases from a side, I'd rather just use the first 50 of each read.

Is there a way to do this using bowtie? I'm trying to avoid going back to the raw data...

Thanks,

John

[Brian Bushnell]

John,

If you map them with BBMap, they can go off the ends of the scaffolds without a problem.

bbmap.sh in=reads.fq ref=loci.fa out=mapped.sam minratio=0.5

The "minratio" flag specifies what fraction of a read must be inside the scaffold. You could also, of course, just grab the loci with more padding on the sides...

Well, there's also a tool in that package that can make the reads 50bp:

reformat.sh in=reads.fq out=trimmed.fq ftr=50

That will trim the right side of reads down to length 50. But that seems like a waste of data, and it will make mapping less precise, so I don't recommend it. If I were you, I WOULD go back to the original data and map it untrimmed using BBMap.

[jgibbons1]

Brian, thanks for the heads up about BBMap! The "minratio" flag sounds like a very nice alternative to arbitrary trimming. I think I'll try both methods and see how they compare.

I was asking if this could be accomplished in bowtie specifically, because I've used it quite extensively to measure other elements in the genome and I'm afraid if I switch to a different mapping program I will have to redo all of the analysis.

[Brian Bushnell]

If you're locked in to bowtie, then just use reformat.sh to first trim reads to the fixed length, then map with bowtie. Of course even trimmed, some of them will still hang off the sides, which may bias the analysis depending on what you're trying to measure. For example, a shorter scaffold will end up with lower coverage.

[dpryan]

If you're willing to use bowtie2 rather than bowtie1 then just use --local alignment. You'll never get global alignment to work if your reads are longer than the regions you map against otherwise (unless you pad things, of course).