FilterMutectCalls crashes when vcf is empty (when running on a sge cluster run)

[waemm]

Hi Brad et al,

When running bcbio on an sge cluster I get the following error. It looks like this is due to the VCF file being empty for one of the small chromosome scaffolds "chr14_GL000009v2_random". This issue is similar to #2829 but I'm currently running on the hg38 build with mostly default parameters.

Details below, thanks again for a fantastic tool!
Warren

command:
java -Dsamjdk.use_async_io_read_samtools=false -Dsamjdk.use_async_io_write_samtools=true -Dsamjdk.use_async_io_write_tribble=false -Dsamjdk.compression_level=2 -Xms681m -Xmx6818m -XX:+UseSerialGC -Djava.io.tmpdir=/shared/pipeline-user/test_data/cll_5_tumor/work/bcbiotx/tmpl2aen8h4 -jar /shared/pipeline-user/bcbio/anaconda/share/gatk4-4.1.2.0-1/gatk-package-4.1.2.0-local.jar FilterMutectCalls --reference /shared/pipeline-user/bcbio/genomes/Hsapiens/hg38/seq/hg38.fa --variant /shared/pipeline-user/test_data/cll_5_tumor/work/bcbiotx/tmpl2aen8h4/CLL006_tumor-chr14_GL000009v2_random_0_122919-raw.vcf.gz --output /shared/pipeline-user/test_data/cll_5_tumor/work/bcbiotx/tmpl2aen8h4/CLL006_tumor-chr14_GL000009v2_random_0_122919-raw-filt.vcf.gz

some of the output:
12:40:05.745 INFO FilterMutectCalls - Done initializing engine 12:40:05.907 INFO ProgressMeter - Starting traversal 12:40:05.907 INFO ProgressMeter - Current Locus Elapsed Minutes Variants Processed Variants/Minute 12:40:05.909 INFO FilterMutectCalls - Starting pass 0 through the variants 12:40:05.980 INFO FilterMutectCalls - Finished pass 0 through the variants 12:40:06.005 INFO FilterMutectCalls - Starting pass 1 through the variants 12:40:06.017 INFO FilterMutectCalls - Shutting down engine [May 24, 2019 12:40:06 PM UTC] org.broadinstitute.hellbender.tools.walkers.mutect.filtering.FilterMutectCalls done. Elapsed time: 0.02 minutes. Runtime.totalMemory()=691404800 java.lang.IllegalArgumentException: log10 p: Values must be non-infinite and non-NAN

YAML:
`details:

• algorithm:
  aligner: bwa
  ensemble:
  numpass: 2
  aligner: bwa
  ensemble:
  numpass: 2
  mark_duplicates: true
  recalibrate: true
  remove_lcr: true
  svcaller:
  • cnvkit
  • lumpy
  • manta
    tools_on:
  • gemini
  • damage_filter
    variant_regions: /shared/pipeline-user/bcbio/genomes/Hsapiens/hg38/coverage/capture_regions/Exome-NGv3.bed
    variantcaller:
  • mutect2
  • vardict
    vcfanno:
  • gemini
  • somatic
    analysis: variant2
    description: CLL006_tumor
    files:
  • /shared/pipeline-user/test_data/input/ERR315865_1.fastq.gz
  • /shared/pipeline-user/test_data/input/ERR315865_2.fastq.gz
    genome_build: hg38
    metadata:
    phenotype: tumor`

[roryk]

Thanks, I know this is stupid but do you think you could pass on the empty VCF file so I can test and fix this? It is helpful to have the actual failing file so I can definitely reproduce.

[waemm]

CLL006_tumor-chr14_GL000009v2_random_0_122919-raw.vcf.gz

Hi @roryk , sorry for multiple edits, is there a way around this bug ? Will using "--force-single" prevent the pipeline from splitting jobs into chromosomes ? Thanks.

Here is the file and the two commands I used:

mkdir -p /shared/pipeline-user/test_data/cll_5/work/bcbiotx/tmpjgip15hg

java -Dsamjdk.use_async_io_read_samtools=false -Dsamjdk.use_async_io_write_samtools=true -Dsamjdk.use_async_io_write_tribble=false -Dsamjdk.compression_level=2 -Xms681m -Xmx6818m -XX:+UseSerialGC -Djava.io.tmpdir=/shared/pipeline-user/test_data/cll_5/work/bcbiotx/tmpjgip15hg -jar /shared/pipeline-user/bcbio/anaconda/share/gatk4-4.1.2.0-1/gatk-package-4.1.2.0-local.jar Mutect2 --annotation ClippingRankSumTest --annotation DepthPerSampleHC --reference /shared/pipeline-user/bcbio/genomes/Hsapiens/hg38/seq/hg38.fa --annotation MappingQualityRankSumTest --annotation MappingQualityZero --annotation QualByDepth --annotation ReadPosRankSumTest --annotation RMSMappingQuality --annotation FisherStrand --annotation MappingQuality --annotation DepthPerAlleleBySample --annotation Coverage --read-validation-stringency LENIENT -I /shared/pipeline-user/test_data/cll_5/work/align/CLL006_tumor/CLL006_tumor-sort-recal.bam --tumor-sample CLL006_tumor -L /shared/pipeline-user/test_data/cll_5/work/mutect2/chr14_GL000009v2_random/CLL006_tumor-chr14_GL000009v2_random_0_122919-regions-nolcr.bed --interval-set-rule INTERSECTION -O /shared/pipeline-user/test_data/cll_5/work/bcbiotx/tmpjgip15hg/CLL006_tumor-chr14_GL000009v2_random_0_122919-raw.vcf.gz

java -Dsamjdk.use_async_io_read_samtools=false -Dsamjdk.use_async_io_write_samtools=true -Dsamjdk.use_async_io_write_tribble=false -Dsamjdk.compression_level=2 -Xms681m -Xmx6818m -XX:+UseSerialGC -Djava.io.tmpdir=/shared/pipeline-user/test_data/cll_5/work/bcbiotx/tmpjgip15hg -jar /shared/pipeline-user/bcbio/anaconda/share/gatk4-4.1.2.0-1/gatk-package-4.1.2.0-local.jar FilterMutectCalls --reference /shared/pipeline-user/bcbio/genomes/Hsapiens/hg38/seq/hg38.fa --variant /shared/pipeline-user/test_data/cll_5/work/bcbiotx/tmpjgip15hg/CLL006_tumor-chr14_GL000009v2_random_0_122919-raw.vcf.gz --output /shared/pipeline-user/test_data/cll_5/work/bcbiotx/tmpjgip15hg/CLL006_tumor-chr14_GL000009v2_random_0_122919-raw-filt.vcf.gz

Also, I have had some similar issues with other samples in this project, sometimes they seem to work when I run them locally but in this case and a few others they always crash the pipeline. Weird bug!

[chapmanb]

Warren;
Thanks much for this report and apologies about the issue. This looks related to #2829 and we were able to reproduce when the MuTect2 associated .stats file has 0 callable reads. The latest development has a workaround which avoid this by flooring this stat at 1. It also leaves the intermediate files in the work directory (instead of transactional directory) to make it easier to debug if it still fails. If you still have issues after updating please share the input *.vcf.gz and *.vcf.gz.stats files to FilterMutectCalls and we can investigate more. Thanks again and hope this gets your analysis done.

[waemm]

Much appreciated @chapmanb and @roryk for your work on this! Run completed with no problems!