Is bowtie2 really this unstable?

[kgulukota]

I am having the devil's own time running bowtie2 on a relatively modest set (at most 17 Mill reads) of RNA-seq data.

I run it with the "-a" flag (report all alignments) or a "-k 2" (report 2) or the default to just report one alignment.

I have paired-end reads. I have run them as paired end or as unpaired reads.

I have run them against transcriptome or genome.

Only a couple of times have I seen it actually run to completion. When it fails, it does not even give me a segmentation fault. Just keeps spinning for days on end with no output (to stdout or stderror). I am trying to build a pipeline by looking at hits to both genome and transcriptome but have never had it complete the runs against both. (It has only completed the mapping against the transcriptome on a couple of occasions).

Is bowtie2 (supposedly still beta) really this unstable? Do others have experience with it running flawlessly?

[chadn737]

Which version are you running?

There were a number of bugs fixed in the beta2 and beta3 release and now they are on the beta5.

http://bowtie-bio.sourceforge.net/bowtie2/index.shtml

[kgulukota]

I am using beta 5.

bowtie2 --version
/research/software/Bowtie2/bowtie2-2.0.0-beta5/bowtie2-align version 2.0.0-beta5
64-bit
Built on compute-0-38.local
Thu Dec 15 01:02:29 EST 2011
Compiler: gcc version 4.1.2 20080704 (Red Hat 4.1.2-46)
Options: -O3 -m64 -msse2 -funroll-loops -g3
Sizeof {int, long, long long, void*, size_t, off_t}: {4, 8, 8, 8, 8, 8}

[amitm]

has worked well

I have run bowtie2 on two datasets of RNA-Seq (Illumina PE, 100x2) with each ~70million reads.
It did fine I think though I havn't assembled the transcripts yet. I aligned on hg19 and used 12 threads.

Specifically, I used the following options:
--local
-M 5

Both of the time bowtie2 took around 13 hours to complete. Reported ~99% ('proper' paired end ~88%) alignment in each. Insert size summarized from the BAM file was around 175 which is okay.
version 2.0.0-beta5, 64bit (compiled)

EDIT -

/home/amitm/bin/bowtie2-align version 2.0.0-beta5
64-bit
Built on atlas
Fri Feb 3 11:17:37 IST 2012
Compiler: gcc version 4.4.6 20110731 (Red Hat 4.4.6-3) (GCC)
Options: -O3 -m64 -msse2 -funroll-loops -g3
Sizeof {int, long, long long, void*, size_t, off_t}: {4, 8, 8, 8, 8, 8}

[mgogol]

I haven't really switched to it because I thought it was a) unstable and b) my reads are still pretty short.

[kgulukota]

A satisfactory resolution:

I have been in private correspondence with Ben Langmead. There might be something about my sequence file that is causing great bowtie2 inefficiency. The offending option appears to be "--reorder".

Since I can reorder my reads in many other ways, I removed that flag and all my processes ran to completion.

Thank you Ben!