Bowtie: Faulty Combination of Paired Ends

[slink]

Hi All,

My problem is that Bowtie is aligning paired ends together that shouldn't be.

The alignment uses the options, -m1 -v2 --best --strata -X800

The beginning of the alignment output looks fine...

Ex Output at top of file:

Id0001 + chr1 10000 sequence quality
Id0001 - chr1 10200 sequence quality
...

The end of the file seems to align reads together that don't go together (the input is highly repetitive, so this will still align)

Ex Output at bottom of file:

...
Id0052 + chr1 10000 sequence quality
Id0124 - chr1 10200 sequence quality

So it doesn't look like it's just outputting singletons for these reads, but is actually trying to align them together.

Importantly I've double checked that the two paired end inputs are matched. EVERY READ in the input is in the same location in the first paired end file as it's counter part in second paired end file!

Lastly, I'm still using Bowtie 1 so this isn't something weird with Bowtie 2.

Thanks ahead of time !!!!

[slink]

I haven't solved this yet, but here's an update...

When I break the input fastq into smaller subfiles and run those with the same options, I don't come across the problem (as often).

Does anyone know if there's some sort of buffer or memory storage quirk with Bowtie that could be resulting in this? Again, I have very repetitive sequences, and I'm running them with tight restrictions, which is what I think might be causing the problem.

Again, thanks ahead of time. I hope someone out there can help

[apratap]

slink

Can you post the full command as well as the bowtie version ?

-A

[slink]

Sure thing!

The version is 0.12.7 (64-bit)

and the full command is:

bowtie --best --strata -m1 -v2 -X800 [hg19 reference] -1[input.fastq] -2[input.fastq] >[output]

[slink]

Okay maybe I'm going overboard with this... but after I split the reads into subfiles and realigned with the same options as previously, I graphed the number of alignments / subfile on one axis and the existence of mixed reads on an alternate y-axis.

As expected, the number of total alignments dropped when the file also contained these mixed reads.

What was odd, though, is that there seemed to be apprehension? The number of alignments in a subfile would drop but not have mixed reads. Then, the next subfile would have the drastic drop in alignment as well as an occurrence of mixed reads. Almost as if the program was struggling in the previous subfile's alignment, but not quite dead yet. Then...all would be good and everything would go back to normal until the next occurrence. Again, this kind of activity sounds to me like a buffer system, but I'm not sure and would love any other ideas anyone has. For now I'm going to run the subfiles separately, I'm getting enough data to work with, but would love a resolution to the problem.

Thanks again,

Sara

-- Attached is a picture of the graph