mrdetect

Workflow for MRdetect, detection of Minimal Residual Disease from paired tumor-plasma sample

Overview

Dependencies

• mrdetect 1.0
• bcftools 1.9

Usage

Cromwell

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

Inputs

Required workflow parameters:

Parameter	Value	Description
tumorSampleName	String	ID for WGS tumor sample, must match .vcf header
tumorvcf	File	tumor vcf file, bgzip
tumorvcfindex	File	tumor vcf index file
reference	String	genome reference build. Only hg38 supported
instrument	String	sequencing instrument used (Illumina NovaSeq X Plus or Illumina NovaSeq 6000)

Optional workflow parameters:

Parameter	Value	Default	Description
plasmabam	File?	None	plasma input .bam file
plasmabai	File?	None	plasma input .bai file
plasmaSampleName	String?	None	name for plasma sample (from bam)
full_analysis_mode	Boolean	true	Enable full analysis mode with this flag

Optional task parameters:

Parameter	Value	Default	Description
filterVCF.tumorVCFfilter	String	"FILTER~'haplotype'	FILTER~'clustered_events'
filterVCF.tumorVAF	String	"0.1"	Variant Allele Frequency for tumor VCF
filterVCF.jobMemory	Int	64	Memory allocated for this job (GB)
filterVCF.threads	Int	4	Requested CPU threads
filterVCF.timeout	Int	10	Hours before task timeout
parseControls.jobMemory	Int	4	Memory for this task in GB
parseControls.timeout	Int	12	Timeout in hours, needed to override imposed limits
controlPullReads.jobMemory	Int	64	Memory allocated for this job (GB)
controlPullReads.threads	Int	4	Requested CPU threads
controlPullReads.timeout	Int	20	Hours before task timeout
controlPullReads.pullreadsScript	String	"$MRDETECT_ROOT/bin/pull_reads"	pull_reads.py executable
controlQualityScore.jobMemory	Int	64	Memory allocated for this job (GB)
controlQualityScore.threads	Int	4	Requested CPU threads
controlQualityScore.timeout	Int	20	Hours before task timeout
controlQualityScore.pickle	String	"$MRDETECT_ROOT/bin/MRDetectSNV/trained_SVM.pkl"	trained pickle for detecting real tumor reads
controlQualityScore.qualityscoreScript	String	"$MRDETECT_ROOT/bin/quality_score"	quality_score.py executable
controlDetectSNVs.tumorSampleName	String	basename(tumorvcf,".vcf")	name for tumour sample (from vcf)
controlDetectSNVs.jobMemory	Int	64	Memory allocated for this job (GB)
controlDetectSNVs.threads	Int	4	Requested CPU threads
controlDetectSNVs.timeout	Int	20	Hours before task timeout
controlDetectSNVs.blocklist	String	"$PWGS_BLOCKLIST_ROOT/blocklist.vcf.gz"	list of sites to exclude from analysis, gzipped
controlDetectSNVs.filterAndDetectScript	String	"$MRDETECT_ROOT/bin/filterAndDetect"	filterAndDetect.py executable
samplePullReads.jobMemory	Int	64	Memory allocated for this job (GB)
samplePullReads.threads	Int	4	Requested CPU threads
samplePullReads.timeout	Int	20	Hours before task timeout
samplePullReads.pullreadsScript	String	"$MRDETECT_ROOT/bin/pull_reads"	pull_reads.py executable
sampleQualityScore.jobMemory	Int	64	Memory allocated for this job (GB)
sampleQualityScore.threads	Int	4	Requested CPU threads
sampleQualityScore.timeout	Int	20	Hours before task timeout
sampleQualityScore.pickle	String	"$MRDETECT_ROOT/bin/MRDetectSNV/trained_SVM.pkl"	trained pickle for detecting real tumor reads
sampleQualityScore.qualityscoreScript	String	"$MRDETECT_ROOT/bin/quality_score"	quality_score.py executable
sampleDetectSNVs.tumorSampleName	String	basename(tumorvcf,".vcf")	name for tumour sample (from vcf)
sampleDetectSNVs.jobMemory	Int	64	Memory allocated for this job (GB)
sampleDetectSNVs.threads	Int	4	Requested CPU threads
sampleDetectSNVs.timeout	Int	20	Hours before task timeout
sampleDetectSNVs.blocklist	String	"$PWGS_BLOCKLIST_ROOT/blocklist.vcf.gz"	list of sites to exclude from analysis, gzipped
sampleDetectSNVs.filterAndDetectScript	String	"$MRDETECT_ROOT/bin/filterAndDetect"	filterAndDetect.py executable
snvDetectionSummary.pvalue	String	"0.00001"	p-value for HBC error rate
snvDetectionSummary.jobMemory	Int	20	Memory allocated for this job (GB)
snvDetectionSummary.threads	Int	1	Requested CPU threads
snvDetectionSummary.timeout	Int	2	Hours before task timeout
snvDetectionSummary.modules	String	"mrdetect/1.1.1"	Required environment modules
snvDetectionSummary.pwgtestscript	String	"$MRDETECT_ROOT/bin/pwg_test"	executable of pwg_test.R

Outputs

Output	Type	Description
snvDetectionResult	File?	{'description': 'Result from SNV detection incl sample HBCs', 'vidarr_label': 'snvDetectionResult'}
pWGS_svg	File?	{'description': 'pWGS svg', 'vidarr_label': 'pWGS_svg'}
snpcount	File	{'description': 'number of SNPs in vcf after filtering', 'vidarr_label': 'snpcount'}
snvDetectionVAF	File?	{'description': 'VAF from SNV detection for sample', 'vidarr_label': 'snvDetectionVAF'}
final_call	File?	{'description': 'Final file of mrdetect results', 'vidarr_label': 'final_call'}
filteredvcf	File?	Filtered vcf

Commands

This section lists commands run by the MRDetect workflow.

filterVCF

Performs vcf Filtering, followed by processing of individual MRDetect calls. Filters include removing difficult regions (optional), splitting multiallelic loci into one allele per line, removing indels, removing loci by quality metrics (set by tumorVCFfilter) and finally removing SNPs by VAF (set by tumorVAF).

<<< set -euo pipefail

	$BCFTOOLS_ROOT/bin/bcftools view -s ~{tumorSampleName} --regions-file ~{difficultRegions} ~{tumorvcf} |\
	$BCFTOOLS_ROOT/bin/bcftools norm --multiallelics - --fasta-ref ~{genome} |\
	$BCFTOOLS_ROOT/bin/bcftools filter -i "TYPE='snps'" |\
	$BCFTOOLS_ROOT/bin/bcftools filter -e "~{tumorVCFfilter}" |\
	$BCFTOOLS_ROOT/bin/bcftools filter -i "(FORMAT/AD[0:1])/(FORMAT/AD[0:0]+FORMAT/AD[0:1]) >= ~{tumorVAF}" > ~{tumorSampleName}.SNP.vcf

	awk '$1 !~ "#" {print}' ~{tumorSampleName}.SNP.vcf | wc -l > ~{tumorSampleName}.SNP.count.txt

>>>

MRDetect proceed across three steps. These tasks are run through for the sample and for all the controls.

pullReads

1- pull_reads takes any reads in the plasma .bam that corresponds to a SNP in the solid-tumour .vcf.

<<< set -euo pipefail

	~{pullreadsScript} \
		--bam ~{plasmabam} \
		--vcf ~{tumorvcf} \
		--out PLASMA_VS_TUMOR.tsv

>>>

calculateQualityScore

2- quality_score assesses the likelihood that any read is plasma based on the quality score and the trained pickle.

<<< set -euo pipefail

	~{qualityscoreScript} \
		--pickle-name ~{pickle} \
		--detections ~{snvDetectionReads} \
		--output_file PLASMA_VS_TUMOR.svm.tsv

>>>

detectSNVs

3- filterAndDetect keeps reads with high plasma likehood and removed those for which SNPs are in the blocklist.

<<< set -euo pipefail

	~{filterAndDetectScript} \
		--vcfid ~{tumorSampleName} --bamid ~{plasmaSampleName} \
		--svm ~{snvDetectionReadsScored} \
		--vcf ~{tumorvcf} \
		--output ./ \
		--blocklist ~{blocklist} \
>>>

parseControls

This command processes the list of control files as paired bam/bai files and prints them out for detection.

<<< python <<CODE import os

	with open(os.environ.get("~{controlFileList}")) as f:
		for line in f:
			line = line.rstrip()
			tmp = line.split("\t")
			r = tmp[0] + "\t" + tmp[1]
			print(r)
	f.close()
	CODE
>>>

snvDetectionSummary

Finally, pwg_test.R will process the controls and the sample to make a final call.

<<< set -euo pipefail

	cat ~{sep=' ' controlCalls} | awk '$1 !~ "BAM" {print}' > HBCs.csv

	cat ~{sampleCalls} HBCs.csv > ~{plasmaSampleName}.HBCs.csv

	~{pwgtestscript} \
		--sampleName ~{plasmaSampleName} \
		--results ~{plasmaSampleName}.HBCs.csv \
		--candidateSNVsCountFile ~{snpcount} \
		--vafFile ~{vafFile} \
		--pval ~{pvalue} 

>>>

Support

For support, please file an issue on the Github project or send an email to [email protected] .

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