Speed up GATK joint genotype calling?

Question

[jtwalker]

I'm using GATK's GenotypeGVCFs tool to jointly genotype ~1000 samples. It's my understanding that because of the genome wide annotations that are calculated, I can't speed things up by using CombineVCFs on smaller jointly called groups. Am I correct?

Is there some way to speed up my joint genotyping with GATK?

Thanks!

edit: The reference genome we're using is only composed of scaffolds. We're worried that because our genome is only composed of scaffolds that this might be slowing things down. Is this thinking correct, or does GATK not care?

[Pierre Lindenbaum]

you can also parallelize things using each contig and the option -L

Answer 1

[Medhat]

Parallelism options

This tool can be run in multi-threaded mode using this option.

TreeReducible (-nt)
  

for more details -nt

[jtwalker]

I had considered this, but the gatk doc says that we'd need to have the same amount of ram available for each thread. As our ram usage is already a concern, would this still be a viable option?

[Medhat]

The idea of threading is that all threads shares the same amount of memory (in contrast to multiprocessing), So my expectation that it should be Ok.