New TopHat version (2.0.7): Uses only one core?

[dfhdfh]

Hi there,
I just installed the new TopHat version 2.0.7 along with the new Bowtie 2.0.6. Everything works fine and without errors, except for the fact that it seems to run on only one core. Before this update, I used TopHat 2.0.6 with Bowtie 2.0.4 and -p 8 worked absolutely perfectly. Now with the new version, I use the exact same command, including -p 8, but the runs take forever and the CPU usage shows only 100% CPU usage instead of the expected 800%. Should I downgrade to Bowtie 2.0.5 (as this is the version stated in the news post of TopHat 2.0.7) or is this a bug? Does anyone else have this problem?

Jens

Edit: The complete command I'm using is the following:

tophat -r 50 --mate-std-dev 30 -p 8 --no-coverage-search --microexon-search --library-type fr-unstranded

[Jetse]

Not each step is multithreaded in tophat, did you check multiple times during different steps of the process?

[dfhdfh]

Yeah, I did. During the mapping it should be at 800%.
I downgraded to Bowtie 2.0.5 and now it shows a CPU usage of 800% at the mapping steps but it is unbelievably slow. The same run with the same options takes about 5-7x the time it did with TopHat 2.0.6/Bowtie 2.0.4. How can this be?

[benjamir]

dfhdfh,
Tophat 2.0.7 is the first version of the program I have ever used. I am also using Bowtie 2.0.6. My initial test run of Tophat has entered its fourth day.

The test is paired-end ( ~100bp qual trimmed x 2, 36.6E6 reads) with options that “may greatly increase mapping accuracy”, mapping every read to transcriptome, genome, and splice variants:
tophat -p 8 --library-type fr-unstranded --read-realign-edit-dist 0 --min-anchor-length 6 --microexon-search -G

One person in my lab suggested the long run could be due to the microexon search option. It seems in your case, this has been ruled out.

[dfhdfh]

It is really weird, indeed. I had some trouble with 2.0.6 (broken gzip pipe, CIGAR errors, quality and sequence inconsistent) that seem to be gone with the new version. Before, I always deleted the "faulty" reads (which were not faulty if you looked at them) and it then ran through or produced another error. With 2.0.7, these problems are gone, which is why I would like to use it as my daily solution from here on out. It seems that I have to live with it taking a lot more time than the previous version, though

[vpr8]

I also have bowtie 2.0.6 and TopHat 2.0.6 and it's also only using one core during mapping

[yzzhang]

I used tophat 2.0.7 with bowtie 2.0.6 it works well. for my 50 million *2 reads, it usually costs 9 hours when I use 12 cpus