bowtie2: running out of memory?

[impeachme2]

Hello,

I'm attempting to align 30 million paired-end reads to hg19 from a 50-cycle illumina run with bowtie2.

I'm doing this one a computer with 12 cores and 48 GB ram.

[user@host folder]$ bowtie2 --dovetail -x /hg19/bt2_index/hg19 -1 219_0h_R1_val_1.fq -2 219_0h_R2_val_2.fq -p 12 -S 219_0h.sam
terminate called after throwing an instance of 'std::bad_alloc'
  what():  std::bad_alloc

I also get this error when using -p 4

The sam files are getting up to ~ 7 GB in size before it throws that error and stops. Based on comparing the last alignment names in the sam and the fastqs, it's getting about halfway through the fastq.

Am I running out of memory? When I check my free memory during alignment, it doesn't appear that I'm coming close to running out.

[user@host folder]$ free -g
             total       used       free     shared    buffers     cached
Mem:            47         12         34          0          0          6
-/+ buffers/cache:          6         41
Swap:            0          0          0

What can I do to get this to complete?

[GenoMax]

You may be running out of temporary disk space (/tmp) rather than memory. Bowtie and TopHat are know to need a lot of tmp space.

[impeachme2]

Hmm, I don't see anything in /tmp/ that bowtie2 is creating.

[GenoMax]

Hmm, I don't see anything in /tmp/ that bowtie2 is creating.

Is the job running now? The files may be cleaned up after the job quits.

[impeachme2]

Yes, it is running now, and there is nothing in /tmp

Does it always write to /tmp regardless if where bowtie2 is running from?

I've aligned to several genomes before with bowtie2, including human, and haven't had this issue.

Interestingly, when I accidentally aligned mouse reads to hg19 with the same alignment settings, it worked (though obviously gave low alignment rate).

[impeachme2]

I've also tried the same alignment with 5 computers of similar specs, with no success.

[GenoMax]

Interesting. So we can assume that there is nothing wrong with your genome index files.

Have you thought of checking integrity of your sequence file to see if there is something odd there. Simon Andrews had posted a script to fix/check integrity of fastq scores (or files, can't remember) here some time back. Or perhaps something like this: http://genome.sph.umich.edu/wiki/FastQValidator

Found the thread that has Simon's script (Post #8).