dariober / cnv_facets

Somatic copy variant caller (CNV) for next generation sequencing
Other
67 stars 15 forks source link

Run SNP-pileup only [Feature Request] #13

Open not-rs opened 4 years ago

not-rs commented 4 years ago

Hi Dario,

Would it be possible to add an option to run only the SNP-pileup step with --snp-nprocs ? This would only generate the pileup input for option 2.

The current setup on a cluster causes the requesting of excess memory for the SNP pileup and CPUs which are not used in the following steps of cnv_facets.R.

Best wishes, Roshan

dariober commented 4 years ago

The current setup on a cluster causes the requesting of excess memory for the SNP pileup and CPUs which are not used in the following steps of cnv_facets.R.

Hi- I think I see your point. snp-pileup takes many CPUs but little memory while facets needs only 1 CPU but a lot of memory. So splitting the two jobs would make better use of resources.

Adding an option to produce only the pileup is certainly possible - I'll think about it. I should point out, however, that the pileup step doesn't benefit that much of multiple processes. There may be a speed-up of 4-8x relative to a single process so you could run cnv_facets.R with, say, --snp-nprocs 2 or 4 to limit the resource request.