tophat / bowtie mapping discrepancies

[vineeth_s]

I am mapping 50 bp SOLiD RNA-Seq reads to the mouse genome.

When I use bowtie, I get ~60% mapping (I am running bowtie with -e 950 as has been suggested elsewhere for mapping SOLiD runs with bowtie)

As there was no direct way to give tophat the -e option, I changed line 872 in the tophat python script from

bowtie_header_cmd = [bowtie_path, "--sam"]

to

bowtie_header_cmd = [bowtie_path, "--sam", "-e 950"]

When I run tophat with this option, I still get very poor mapping, of about 25%

Why is there this discrepancy, as one would assume that all of the unspliced mapping that bowtie does should be reproducible by tophat ?

Vineeth

[DZhang]

Hi vineeth,

Please verify with the author(s) if your way of passing "-e 950" from Tophat to Bowtie. From the Tophat manual, it has very limited options to pass parameters to Bowtie. The manual says Tophat is currently optimized for 75bp or longer SOLID reads. I am not sure what it actually means but it might have contributed to your observation.

Douglas

https://www.contigexpress.com

[vineeth_s]

That really should not be an issue as this is the pythonic way (using the subprocess module) of building a command for executing something on the command-line.
Further when I look at the run logs in the logs folder that TopHat creates, I see the bowtie command formed without issue.

What I am wondering is since TopHat works serially by first aligning to the genome, then taking the unmapped reads and trying to align it to the splice junction which is where the read length matters (so even if the 75 bp recommendation makes a huge difference) I should only see an increase in mapping as opposed to just using bowtie; what confounds me is why I see the mapping decrease when compared to just using bowtie

[DZhang]

Hi Vineeth,

Conceptually you are right but I am not sure how exactly tophat implements the concept. Again, please send your request to the tophat user forum or the author for a direct answer. At the same time, you may compare the mapping rates b/t Bowtie and Tophat without the option "-e 950" to see the trend? Occasionally these logical machines do illogical things.

Douglas

https://www.contigexpress.com