I am testing out GATK Haplotype caller to make it work with our existing variant calling pipeline. In our current pipeline which is based on GATK best practice, the variants are called globally across the whole population, ie, the VCF file contains genotypes of 1000+ subjects. From my initial testing with Haplotype caller so far, I have noticed that the caller is too slow (remaining time is 80+ hours and growing…), when I am using it with 1000+ samples and a short exome region interval file, 8 threads and SGE with 16 nodes. Do you have suggestions on how to speed up the caller; or how to go about multisample variant calling (1000+) with Haplotype caller in the most efficient manner?
P.S. I am using GATK-2.7