We've moved!
This site is now read-only. You can find our new documentation site and support forum for posting questions here.
Be sure to read our welcome blog!

speeding up HaplotyperCaller


I've started to call variants on 200 diploid individuals with a genome size of 160Mb, using HaplotypeCaller and 20 cores. In about 10 hours it's covered the first 600,000 bases, so it looks like completion will take several months. I could potentially use up to 64 cores although this would inhibit other users.

My input parameters are as below. They are probably pretty loose. I plan on hard-filtering variants after this.

Do you have any advice on how to speed this up?



                    --heterozygosity 0.01 \
                    --indel_heterozygosity 0.001 \
                    -stand_call_conf 31 \
                    -stand_emit_conf 31 \
                    -mbq 10 \
                    -gt_mode DISCOVERY \
                                            -L chr2L -L chr2R -L chr3L -L chr3R -L chrX -L chr4 -L chrM \

Best Answer


  • Geraldine_VdAuweraGeraldine_VdAuwera Cambridge, MAMember, Administrator, Broadie admin

    Please show the full command line. Are you running HC on all the bam files together?

  • BlueBlue Member

    Hi Gera, yes, I'm running HC on all 222 bam files together.


    $ -N HAPCALL

    $ -pe openmp 20

    $ -S /bin/sh

    $ -cwd

    $ -j y

    $ -q bioinf.q

    . /etc/profile.d/modules.sh
    module load gatk/3.2.2 jre/1.7.0_25
    GenomeAnalysisTK -nct 20 --analysis_type HaplotypeCaller \
    --reference_sequence ../../reference_sequences/dmel/v6.0/dm6.fa \
    --input_file lhm_rg_bams.list \
    --heterozygosity 0.01 \
    --indel_heterozygosity 0.001 \
    -stand_call_conf 31 \
    -stand_emit_conf 31 \
    -mbq 10 \
    -gt_mode DISCOVERY \
    -L chr2L -L chr2R -L chr3L -L chr3R -L chrX -L chr4 -L chrM \
    --out LHm_RG_raw.vcf

    and a bit from the log.

    Executing as [email protected] on Linux 2.6.32-358.14.1.el6.x86_64 amd64; Java HotSpot(TM) 64-Bit Server VM 1.7.0_25-b15.

  • pdexheimerpdexheimer Member ✭✭✭✭

    If I recall correctly, I could only ever run HC over about 70 or 80 exomes (genome size ~50MB) at a time. This is one of the reasons the gVCF workflow is so great - it dodges this kind of issue altogether.

    It's been a couple of years, but I think my case at the time was ~150 exomes that I had to split into 2 batches in order to successfully run with HC, and it still took about a week. When GATK3 came out with gVCFs, my calling time dropped to something like 8 hours

Sign In or Register to comment.