Chapter 9 Calculate genotype posteriors

9.1 Brief introduction

After we filtered our callset, we use extra information like pedigree and allele frequencies in relevant populations to refine the genotype assignments.

9.2 Benchmarks

We did benchmarks on the performance of CalculateGenotypePosteriors with different numbers of CPUs and memory. As show in figure 9.1, the runtime was not reduced with increasing threads.

Figure 9.1: Runtime of CalculateGenotypePosteriors as a function of the number of threads

We normally recommend running jobs with 70%-80% efficiency. Based on the efficiency calculated from the benchmarks above (figure 9.2) we recommend not running CalculateGenotypePosteriors with more than 2 threads.

Figure 9.2: Efficiency of CalculateGenotypePosteriors as a function of the number of threads

Increasing memory didn’t improve the performance (figure 9.3).

Figure 9.3: Runtime of CalculateGenotypePosteriors as a function of the number of threads

9.3 Optimized script

#! /bin/bash
set -euo
module load GATK/4.3.0.0
cd data/;
gatk --java-options "-Djava.io.tmpdir=/lscratch/$SLURM_JOBID -Xms2G -Xmx2G -XX:ParallelGCThreads=2" \
   CalculateGenotypePosteriors \
   -V indel.SNP.recalibrated_99.9.vcf.gz \
   -ped trio_pedigree.ped \
   --supporting-callsets af-only-gnomad.hg38.vcf.gz \
   -O trio_refined_99.9.vcf.gz

Job submission:

sbatch --cpus-per-task=2 --mem=2G --gres=lscratch:100 --time=2:00:00  09-GATK_CalculateGenotypePosteriors_99.9.sh

Note:

There are multiple filters could be applied, it depends on your research needs.