Bowtie1: Reads don't get randomly distributed on direct repeat

[Mammoth]

Hi all,

I am mapping small RNA sequencing reads with bowtie1.2.2 to a locus that encodes a direct repeat with the exactly(!) the same sequence repeated 19 times. I don't allow mismatches, and ask to get back only one alignment:

bowtie index clippedreads.fq --best --strata -M 1 -v 0 -S | \
samtools view -Sb -F 4 - |\
samtools sort - -o outfile.bam

(or alternatively --best -k 1 -v 0; however, it does not really make a difference in my hands)

A couple of thousand reads are mapping to that locus (all these reads have the same sequence of course), and based on the bowtie1 manual I would assume they are randomly (and roughly equally) distributed across the 19 repeat units. However, what I get is one or two of the repeat units are highly covered, and the rest is distributed as expected (see picture):


Weirdly, the peak shifts with different libraries, but is otherwise nicely reproducible.
I was wondering what is going on, why the repeat units are not equally covered, or how this can be handled?

Thanks a ton! Any suggestion is highly appreciated.

[pmiguel]

Looks more-or-less random to me. How many standard deviations from the mean number of hits/repeat are the high and low-hit repeat counts?

--
Phillip

[Mammoth]

Dear Phillip,

Thank you for your reply. The biggest peak is around 4 standard deviations larger than the mean, the lowest one 2 SD. That seems quite a bit for a random distribution, or not?