Multiprocessor samtools

[rstarke]

Hi folks -

I'm wondering if anyone knows of a way to make samtools utilize all the processors and cores on a machine. I usually run merge, index, mpileup, and pass the results to bcftools on a machine with 16 cores, but top shows that these programs use only 1 core. For large bam files (~10GB merged), this can mean hours of extra compute time. And if you're paying for compute time by the hour, this means lots of extra $. Has anyone figured out a way to make samtools more "parallel"? Barring a straightforward way to do that, does anyone know of other tools that will merge, index, and do variant discovery/genotyping in a parallel manner from bam files? Thanks.

[ashkot]

I am looking for the same answer, have you figured it out yet?

[colindaven]

We just normally run many processes at once. I haven't heard of anyline parallelising Samtools.

[krobison]

Most of the time, SAMTools is probably waiting on disk (IO-bound) & multithreading might offer little advantage.

[colindaven]

Krobison : if samtools is IO limited do you know of any performance related settings to put more into memory ? As far as I know it does quite a lot of calculation eg local realignment of reads as well.

[jcgrenier]

One way to do it could be to run it per chromosome. You'll gain some time by this way.