variantMerging

VariantMerging is a workflow for combining variant calls from SNV analyses done with different callers (such as muTect2, strelka2). The workflow pre-processes input vcf files by removing non-canonical contigs, fixing fields and inferring missing values from available data. It combines calls, annotating them with caller-specific tags which allows identification of consensus variants. The workflow also uses GATK for producing merged results. In this case, all calls appear as-as. Essentially, this is a simple concatenation of the inputs.

Pre-processing

The script used at this step performs the following tasks:

• removes non-canonical contigs
• adds GT and AD fields (dot or calculated based on NT, SGT, if available)
• removes tool-specific header lines

Overview

Dependencies

• tabix 0.2.6
• gatk 4.2.6.1

Usage

Cromwell

java -jar cromwell.jar run variantMerging.wdl --inputs inputs.json

Inputs

Required workflow parameters:

Parameter	Value	Description
reference	String	Reference assmbly id, passed by the respective olive
inputVcfs	Array[Pair[File,String]]	Pairs of vcf files (SNV calls from different callers) and metadata string (producer of calls).
tumorName	String	Tumor id to use in vcf headers
outputFileNamePrefix	String	Output prefix to prefix output file names with.

Optional workflow parameters:

Parameter	Value	Default	Description
normalName	String?	None	Normal id to use in vcf headers, Optional

Optional task parameters:

Parameter	Value	Default	Description
preprocessVcf.preprocessScript	String	"$VARMERGE_SCRIPTS_ROOT/bin/vcfVetting.py"	path to preprocessing script
preprocessVcf.jobMemory	Int	12	memory allocated to preprocessing, in gigabytes
preprocessVcf.timeout	Int	10	timeout in hours
mergeVcfs.timeout	Int	20	timeout in hours
mergeVcfs.jobMemory	Int	12	Allocated memory, in GB
combineVariants.combiningScript	String	"$VARMERGE_SCRIPTS_ROOT/bin/vcfCombine.py"	Path to combining script
combineVariants.jobMemory	Int	12	memory allocated to preprocessing, in GB
combineVariants.timeout	Int	20	timeout in hours
postprocessMerged.postprocessScript	String	"$VARMERGE_SCRIPTS_ROOT/bin/vcfVetting.py"	path to postprocessing script, this is the same script we use for pre-processing
postprocessMerged.jobMemory	Int	12	memory allocated to preprocessing, in gigabytes
postprocessMerged.timeout	Int	10	timeout in hours
postprocessCombined.postprocessScript	String	"$VARMERGE_SCRIPTS_ROOT/bin/vcfVetting.py"	path to postprocessing script, this is the same script we use for pre-processing
postprocessCombined.jobMemory	Int	12	memory allocated to preprocessing, in gigabytes
postprocessCombined.timeout	Int	10	timeout in hours

Outputs

Output	Type	Description	Labels
mergedVcf	File	vcf file containing all variant calls	vidarr_label: mergedVcf
mergedIndex	File	tabix index of the vcf file containing all variant calls	vidarr_label: mergedIndex
combinedVcf	File	combined vcf file containing all variant calls	vidarr_label: combinedVcf
combinedIndex	File	index of combined vcf file containing all variant calls	vidarr_label: combinedIndex

Commands

This section lists command(s) run by variantMerging workflow

Preprocessing

Detect NORMAL/TUMOR swap, impute missing fields (i.e. in case of such callers as strelka)

  python3 PREPROCESSING_SCRIPT VCF_FILE -o VCF_FILE_BASENAME_tmp.vcf -r REFERENCE_ID
  
  bgzip -c VCF_FILE_BASENAME_tmp.vcf > VCF_FILE_BASENAME_tmp.vcf.gz
  gatk SortVcf -I VCF_FILE_BASENAME_tmp.vcf.gz 
               -R REF_FASTA 
               -O VCF_FILE_BASENAME_processed.vcf.gz

Merging vcf files

This is a simple concatenation of input vcfs, there may be duplicate entries for the same call if multiple callers discover the same variant.

  gatk MergeVcfs -I INPUT_VCFS -O PREFIX_mergedVcfs.vcf.gz
  

Combining vcf files

A more complex merging: matching fields will be annotated by caller.

 
   set -euxo pipefail 
   python3 <<CODE
   import sys
   v = "~{sep=' ' inputVcfs}"
   vcfFiles = v.split()
   with open("vcf_list", 'w') as l:
       for v in vcfFiles:
           l.write(v + "\n")
   CODE
 
   python3 COMBINING_SCRIPT vcf_list -c OUTPUT_PREFIX_tmp.vcf -n ~{sep=',' inputNames}
   gatk SortVcf -I OUTPUT_PREFIX_tmp.vcf -R REFERENCE_FASTA -O OUTPUT_PREFIX_combined.vcf.gz
 

Postprocessing (Name injection)

The same script used for preprocessing injects names of samples into the header if argumants are passed

  set -euxo pipefail
  python3 ~{postprocessScript} ~{vcfFile} -o ~{basename(vcfFile, '.vcf.gz')}_tmp.vcf -r ~{referenceId} -t ~{tumorName} ~{"-n " + normalName}
  bgzip -c ~{basename(vcfFile, '.vcf.gz')}_tmp.vcf > ~{basename(vcfFile, '.vcf.gz')}.vcf.gz
  tabix -p vcf ~{basename(vcfFile, '.vcf.gz')}.vcf.gz
 

Support

For support, please file an issue on the Github project or send an email to gsi@oicr.on.ca .

Generated with generate-markdown-readme (https://github.com/oicr-gsi/gsi-wdl-tools/)