How long does it usually take to process exome sequencing samples with GATK's UnifiedGenotyper

Question

[idedios]

I'm using a Xeon E5-2650 with Lubuntu 13.10 and GATK version 3.1-1. The computer has 16 GB of 1333 ram and a 7200rpm hard drive with the OS. I set -nct to 16 threads and it's taking about 7 hours to process one sample (this is a TruSight One sample that has already been aligned and processed by MiSeq Reporter). How are most people configuring their servers for processing with GATK?

Does GATK benefit from hyperthreading or should I set it to use the number of cores my CPU has?

Also I tested out the AVX acceleration for HaplotypeCaller and it reduced estimated processing time from 7 hours to 5 hours. I was expecting it to be considerably quicker.

Answer 1

[Zaag]

Try the -nt 16 option instead of -nct 16

[Brett Kennedy]

I find that combining -nt and -nct give the fastest speeds, so I'll usually give each half of the total number of CPUs I want to use.