Workflow for MRdetect, detection of Minimal Residual Disease from paired tumor-plasma sample
java -jar cromwell.jar run mrdetect.wdl --inputs inputs.json
| 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) |
| 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 |
| 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 |
detectControl.tumorSampleName |
String | basename(tumorvcf,".vcf") | name for tumour sample (from vcf) |
detectControl.jobMemory |
Int | 64 | Memory allocated for this job (GB) |
detectControl.threads |
Int | 4 | Requested CPU threads |
detectControl.timeout |
Int | 20 | Hours before task timeout |
detectControl.pickle |
String | "$MRDETECT_ROOT/bin/MRDetectSNV/trained_SVM.pkl" | trained pickle for detecting real tumor reads |
detectControl.blocklist |
String | "$PWGS_BLOCKLIST_ROOT/blocklist.vcf.gz" | list of sites to exclude from analysis, gzipped |
detectControl.pullreadsScript |
String | "$MRDETECT_ROOT/bin/pull_reads" | pull_reads.py executable |
detectControl.qualityscoreScript |
String | "$MRDETECT_ROOT/bin/quality_score" | quality_score.py executable |
detectControl.filterAndDetectScript |
String | "$MRDETECT_ROOT/bin/filterAndDetect" | filterAndDetect.py executable |
detectSample.tumorSampleName |
String | basename(tumorvcf,".vcf") | name for tumour sample (from vcf) |
detectSample.jobMemory |
Int | 64 | Memory allocated for this job (GB) |
detectSample.threads |
Int | 4 | Requested CPU threads |
detectSample.timeout |
Int | 20 | Hours before task timeout |
detectSample.pickle |
String | "$MRDETECT_ROOT/bin/MRDetectSNV/trained_SVM.pkl" | trained pickle for detecting real tumor reads |
detectSample.blocklist |
String | "$PWGS_BLOCKLIST_ROOT/blocklist.vcf.gz" | list of sites to exclude from analysis, gzipped |
detectSample.pullreadsScript |
String | "$MRDETECT_ROOT/bin/pull_reads" | pull_reads.py executable |
detectSample.qualityscoreScript |
String | "$MRDETECT_ROOT/bin/quality_score" | quality_score.py executable |
detectSample.filterAndDetectScript |
String | "$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 |
| Output | Type | Description | Labels |
|---|---|---|---|
snvDetectionResult |
File? | Result from SNV detection incl sample HBCs | vidarr_label: snvDetectionResult |
pWGS_svg |
File? | pWGS svg | vidarr_label: pWGS_svg |
snpcount |
File | number of SNPs in vcf after filtering | vidarr_label: snpcount |
snvDetectionVAF |
File? | VAF from SNV detection for sample | vidarr_label: snvDetectionVAF |
final_call |
File? | Final file of mrdetect results | vidarr_label: final_call |
filteredvcf |
File? | Filtered vcf |
This section lists commands run by the MRDetect workflow.
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
This command processes the list of control files as paired bam/bai files and prints them out for detection.
MRDetect proceed across three steps. This task is run through for the sample and for all the controls.
1- pull_reads takes any reads in the plasma .bam that corresponds to a SNP in the solid-tumour .vcf.
2- quality_score assesses the likelihood that any read is plasma based on the quality score and the trained pickle.
3- filterAndDetect keeps reads with high plasma likehood and removed those for which SNPs are in the blocklist.
set -euo pipefail
~{pullreadsScript} \
--bam ~{plasmabam} \
--vcf ~{tumorvcf} \
--out PLASMA_VS_TUMOR.tsv
~{qualityscoreScript} \
--pickle-name ~{pickle} \
--detections PLASMA_VS_TUMOR.tsv \
--output_file PLASMA_VS_TUMOR.svm.tsv
~{filterAndDetectScript} \
--vcfid ~{tumorSampleName} --bamid ~{plasmaSampleName} \
--svm PLASMA_VS_TUMOR.svm.tsv \
--vcf ~{tumorvcf} \
--output ./ \
--blocklist ~{blocklist} \
--troubleshoot
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
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}
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/)