splitting haplotyepecaller onto multiple threads

[lre1234]

Hi all,

Quick question, I am doing some whole genome and am using the GATK haplotypecaller. To speed things up, I would like to break the genome into smaller chuncks and run each chromosome separately on its own thread (using -L options with a bed file for each chromosome). This will produce multiple gvcf files for 1 sample. If I use the genotypeGVCFs function, will it treat the gvcf's with the same sample name as a single file? Would the combineGVCFs do the same thing and that should be done first?

Thanks

[dpryan]

If you're just going to split things to run on individual threads, then why not just use the -nt option? What you suggest would only make sense if you're using a cluster and want to split individual samples across nodes.

[lre1234]

Sorry, I wasn't clear, but I am going to be splitting the samples across nodes on a cluster. also on the GATK documentation page, there is a caveat listed:

Many users have reported issues running HaplotypeCaller with the -nct argument, so we recommend using Queue to parallelize HaplotypeCaller instead of multithreading

I assume that means that the -nt options shouldn't be used, otherwise it would be a better way to do things.

[dpryan]

Ah, guess I've just never run into issues with it.