A workflow that will use variant calls, expression data and HLA typing to predict Neoantigens
- PCGR : Personal Cancer Genome Reporter, version 2.0.3
- bcftools version 1.9
- Variant Effect Predictor, version 112.0
- pvactools version 4.3.0
- SB_neoantigen_Models
java -jar cromwell.jar run neoantigenPredictor.wdl --inputs inputs.json
| Parameter | Value | Description |
|---|---|---|
HLAFiles |
HLACalls | an array of HLA files and their associated caller names (t1k, optitype) |
DNAVariantCalls |
VariantCalls | the ensemble/combined DNA vcf file, from multiple callers, with the index and the tumourID |
RNAVariantCalls |
VariantCalls | the RNA seq variant calls from Haplotype Caller |
RNAAbundance |
File | the expression data in text format |
outputFilePrefix |
String | a prefix to add to each of the output files, generally identifying the sample being analyzed |
| Parameter | Value | Default | Description |
|---|---|---|---|
reference |
String | "hg38" | the reference build, defaults to hg38 |
| Parameter | Value | Default | Description |
|---|---|---|---|
extractHLAs.modules |
String | "neopipe/1.0.0" | Names and versions of modules |
extractHLAs.jobMemory |
Int | 6 | Memory allocated for task in GB |
extractHLAs.timeout |
Int | 20 | Timeout in hours |
format2pcgr.modules |
String | "neopipe/1.0.0 bcftools/1.9" | Names and versions of modules |
format2pcgr.jobMemory |
Int | 6 | Memory allocated for task in GB |
format2pcgr.timeout |
Int | 20 | Timeout in hours |
PCGR.modules |
String | "pcgr/2.0.3" | Names and versions of modules |
PCGR.jobMemory |
Int | 6 | Memory allocated for task in GB |
PCGR.timeout |
Int | 20 | Timeout in hours |
vepAnnotate.modules |
String | "vep/112.0 pcgr/2.0.3 pvactools/4.3.0 hg38/p12" | Names and versions of modules |
vepAnnotate.jobMemory |
Int | 6 | Memory allocated for task in GB |
vepAnnotate.timeout |
Int | 20 | Timeout in hours |
getPeptides.modules |
String | "pvactools/4.3.0" | Names and versions of modules |
getPeptides.jobMemory |
Int | 6 | Memory allocated for task in GB |
getPeptides.timeout |
Int | 20 | Timeout in hours |
formatCalls.modules |
String | "bcftools/1.9" | Names and versions of modules |
formatCalls.jobMemory |
Int | 6 | Memory allocated for task in GB |
formatCalls.timeout |
Int | 20 | Timeout in hours |
vafDeciles.modules |
String | "neopipe/1.0.0 bcftools/1.9" | Names and versions of modules |
vafDeciles.jobMemory |
Int | 6 | Memory allocated for task in GB |
vafDeciles.timeout |
Int | 20 | Timeout in hours |
ExpressionDeciles.modules |
String | "neopipe/1.0.0 ensembl/104-hg38" | Names and versions of modules |
ExpressionDeciles.jobMemory |
Int | 6 | Memory allocated for task in GB |
ExpressionDeciles.timeout |
Int | 20 | Timeout in hours |
rnaseqVariants.modules |
String | "bcftools/1.9" | Names and versions of modules |
rnaseqVariants.jobMemory |
Int | 6 | Memory allocated for task in GB |
rnaseqVariants.timeout |
Int | 20 | Timeout in hours |
mergePredictorInputs.modules |
String | "neopipe/1.0.0" | Names and versions of modules |
mergePredictorInputs.jobMemory |
Int | 6 | Memory allocated for task in GB |
mergePredictorInputs.timeout |
Int | 20 | Timeout in hours |
chunkPredictorInputFile.modules |
String | "neopipe/1.0.0 sb-neoantigen-models/1.0.0" | Names and versions of modules |
chunkPredictorInputFile.jobMemory |
Int | 6 | Memory allocated for task in GB |
chunkPredictorInputFile.timeout |
Int | 20 | Timeout in hours |
predict.modules |
String | "sb-neoantigen-models/1.0.0" | Names and versions of modules |
predict.jobMemory |
Int | 6 | Memory allocated for task in GB |
predict.timeout |
Int | 20 | Timeout in hours |
mergePredictorOutputs.modules |
String | "neopipe/1.0.0 sb-neoantigen-models/1.0.0" | Names and versions of modules |
mergePredictorOutputs.jobMemory |
Int | 6 | Memory allocated for task in GB |
mergePredictorOutputs.timeout |
Int | 20 | Timeout in hours |
| Output | Type | Description | Labels |
|---|---|---|---|
NeoAntigenPredictions |
File | An xlsx file with worksheets detailing the predictions | vidarr_label: NeoAntigenPredictions |
NeoAntigenNmers |
File | a tsv file with the predictions | vidarr_label: NeoAntigenNmers |
This section lists command(s) run by neoantigenPrediction workflow
- Extraction of HLA Calls to an HLA String as input to sb_neoantigen_prediction
#Inline python code, see WDL task for extractHLAs for details (todo: convert to script)
- Format the Variant Call input so that they are PCGR-ready. This will keep only calls identified by 2 or more callers and filters on Depth and VAF
# Run format2pcgr
format2pcgr \
-i ~{vcfin} \
-o temp.ensemble.somatic.vt.annot.2callers.vcf.gz \
-f 2 \
-t ~{tumorId} \
-v somatic \
> format2pcgr.log 2>&1
# Filter with vcftools
bcftools view -Oz -i 'TDP>=10 && TVAF>=0.05 && NDP>=10 && NVAF<=0.02' \
-o ~{outputFilePrefix}.ensemble.somatic.vt.annot.2callers.vcf.gz \
temp.ensemble.somatic.vt.annot.2callers.vcf.gz
# Index the vcf output
tabix -pvcf ~{outputFilePrefix}.ensemble.somatic.vt.annot.2callers.vcf.gz
- Run PCGR (Personal Cancer Genome Reporter)
# Extract Calls + Tumour VAF with bcftools
set -euxo pipefail
mkdir pcgr
### pcgr is called without reporting, to generate data files from which annotated candidate sites can be extracted
pcgr --vcf2maf \
--force_overwrite \
--vep_buffer_size 500 \
--vep_regulatory \
--exclude_dbsnp_nonsomatic \
--exclude_likely_het_germline \
--exclude_likely_hom_germline \
--exclude_nonexonic \
--maf_gnomad_global 0.002 \
--tumor_site 0 \
--assay WGS \
--tumor_dp_tag TDP --tumor_af_tag TVAF --tumor_dp_min 10 --tumor_af_min 0.05 \
--control_dp_tag NDP --control_af_tag NVAF --control_dp_min 10 --control_af_max 0.02 \
--input_vcf ~{vcf} \
--output_dir pcgr \
--genome_assembly grch38 \
--sample_id ~{outputFilePrefix} \
--vep_dir $VEP_DIR \
--refdata_dir $REFDATA_DIR \
--pcgrr_conda $PCGR_ROOT \
--no_reporting
# Inline R code run pcgr reporting functions to generated tsv and excel files, see WDL task PCGR for details (todo: convert to script)
# Inline python code that filters PCGR tsv output based on EXONIC_STATUS and ACTIONABILITY_TIER to generate candidate sites, see WDL task PCGR for details (todo: convert to script)
# filter the input vcf to only the candidate sites
bcftools view -R ~{outputFilePrefix}.pcgr_filter_sites.txt -o ~{outputFilePrefix}.candidate_sites.vcf ~{vcf}
- Annotate the Candidate sites with VEP to add in consequence information that is needed for Peptide identification
# Extract Calls + Tumour VAF with bcftools
vep \
--input_file ~{vcf} \
--output_file ~{outputFilePrefix}.vep.vcf \
--format vcf --vcf --symbol --terms SO --tsl \
--hgvs --fasta ~{refFasta} \
--offline \
--cache --dir_cache $VEP_DIR --cache_version 112 \
--plugin Frameshift --plugin Wildtype \
--dir_plugins $PVACTOOLS_VEP_PLUGINS --pick --transcript_version \
--force_overwrite
- Generate Peptides for the Candidate sites with pVactools
pvacseq generate_protein_fasta \
-s ~{tumorId} \
-d 12 \
~{vcf} 12 ~{outputFilePrefix}.peptides.fa
- Combined candidate calls with peptide predictions for input to sb_neoantigen_prediction
# Inline Python code, see WDL task formatCalls for details (todo: convert to script)
- Generate Deciles from the PCGR-ready variant calls
# Inline R code to generate Deciles, see WDL task ExpressionDeciles for details (todo: convert to script)
- Generate Deciles from the rnaseq expression data
# Extract Calls + Tumour VAF with bcftools
bcftools query -f '%CHROM\t%POS\t%REF\t%ALT\t%TVAF\n' ~{vcf} > ~{outputFilePrefix}.all_variants_vaf.tsv
# Inline R code to generate Deciles, see WDL task vafDeciles for details (todo: convert to script)
- Extract RNASeq variants to a tsv format
bcftools query -f '%CHROM\t%POS\t%REF\t%ALT\t1\n' ~{vcf} > ~{outputFilePrefix}.rnaseq_variants.tsv
- Combine all sb_neoantigen predictor inputs (candidate calls, vaf deciles, rnaseq deciles, expression calls) to a tsv and xlsx format
# Inline R code to combine data, see WDL task mergePredictoriInputs for details (todo: convert to script)
- process data through sb_neoantigen_predictor
# Input is the xls file with combined input and the hla string
python $SB_NEOANTIGEN_MODELS_ROOT/src/GenerateScores.py ~{xls} ~{hlas}
- scatter/gather for predictor. For improved performance the candidate inputs are separated into smaller subsets and sb_neoantigen_predictor is run in parallel. The output from each subset is then merged into the final output
# subsetting of the data is done with inline python code, see the WDL task chunkPredictorInputFile for details (todo: convert to script)
# merging the subset outputs is done with inline python code, see the WDL task mergePredictorOutputs for details (todo: convert to script)
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/)
