Nextflow Configuration

This README documents the parameters used in the nextflow configuration file. Some default values for configuration variables are initially set up to facilitate the demonstration test-run.

Parameter	Description
large_files	Directory containing large external resources (e.g. gnomAD, MANE, AlphaMissense). See large_files README for instructions on how to source this content
processed_annotations	Output directory for all processed/reformatted data. Cohort/callset independent, to be used across multiple runs.
outputDir	Defaults to processed_annotations, required for the 'workflow outputs' NextFlow functionality. Override for Annotation/Talos workflow using -output-dir <path>
input_vcf_extension	Linked to shards. If a sharded VCF is provided, this argument is the file extension to glob for. By default this is vcf.bgz, but could be altered to vcf.gz.
alphamissense_tsv	Path in large_files to the AlphaMissense raw data, reformatted to be used in Hail annotation
alphamissense_zip	Path in processed_annotations to the reformatted AlphaMissense data, ready to be used by Echtvar for annotation
ensembl_bed	Path in processed_annotations to the Ensembl BED file, one line per gene
ensembl_merged_bed	Path in processed_annotations to the Ensembl BED file, overlapping regions merged
mane_json	Path in processed_annotations to pre-processed MANE data
mitimpact_zip	Path in processed_annotations to the reformatted MitImpact reference data
mitotip_zip	Path in processed_annotations to the reformatted MitoTip reference data
napogee_zip	Path in processed_annotations to the reformatted nAPOGEE reference data
ensembl_gff	Path in large_files to the Ensembl GFF3 file
gnomad_zip	Path in large_files to the echtvar annotation file from Zenodo
mane	Path in large_files to the MANE transcript data
ref_genome	Path in large_files to the reference genome FASTA file
hpo	Path in large_files to the HPO ontology file in .obo form
gen2phen	Path in large_files to the Geno~Phenotype databset in .txt form
phenio_db	Path in large_files to the Monarch Phenotype DB in .db.gz form
mitimpact_tsv	Path in large_files to the raw Mitimpact file
mitotip_tsv	Path in large_files to the raw MitoTip file
napogee_tsv	Path in large_files to the raw Mitochondrial nAPOGEE file
submission_summary	FTP URL for weekly ClinVar submission dump
variant_summary	FTP URL for weekly ClinVar variant dump
clinvar_blacklist	A string, containing quoted, space-delimited entries for all ClinVar submitter sites to blacklist (ignore).
container	Docker image to use
docker.enabled	Parameter for Nextflow to enable usage of Docker

Structural Variant annotation

These parameters are only used by the optional SV annotation workflow. They are ignored unless the input TSV declares an sv column, so an SNV-only run does not need any of the reference data below.

Parameter	Description
svafotate_bed	Path in large_files to the SVAFotate population-frequency BED. Use as downloaded — see below
mane_gtf	Path in large_files to the MANE GTF, used by GATK SVAnnotate for gene consequences
svannotate_noncoding_bed	Path in large_files to the GATK-SV non-coding elements BED
ref_dict	Path in processed_annotations to the sequence dictionary. Generated on the first SV run if absent
sv_overlap_fraction	Reciprocal overlap fraction required to match an SV against the gnomAD reference. Defaults to 0.5
gatk_container	Docker image for GATK SVAnnotate. Pulled from the public registry, not built locally
svafotate_container	Docker image for SVAFotate. Built locally — docker build -f docker/SVAFotate_Dockerfile -t svafotate:0.1.0 .

Providing the input

The joint-called SV VCF is supplied per cohort, as an sv column in the input TSV, alongside the existing mito column. A bgzipped VCF with a matching .tbi is expected. Cohorts with no SV data can leave the column empty, or remove the column from the input file completely.

The input VCF must already carry AC, AF, AN, N_HET, N_HOMALT and array-typed MALE_AF/FEMALE_AF (or AF_MALE/AF_FEMALE). None of these are written by the annotation chain, and RunHailFilteringSv reads them unconditionally — so a VCF missing them fails at the very end of the run, after both expensive annotation steps have completed. A full GATK-SV callset carries all of them; a bare gCNV or caller-native VCF may not.

Trying it out

nextflow/inputs/test_sv.tsv is the SNV test input plus an sv column, pointing at a small adversarial SV VCF. nextflow/inputs/test.tsv deliberately has no sv column, so the default test run never requires the 478 MB of SV reference data. Regenerate the test VCF with:

python nextflow/inputs/generate_sv_test_data.py

Re-runs are skipped automatically

If <outdir>/<cohort>_outputs/<cohort>_sv_annotated.vcf.bgz already exists, that cohort's annotation is skipped entirely and the existing file is reused. Both annotation tools are expensive, so delete that file to force a re-annotation.

Do not add chr prefixes to the SVAFotate BED

The SVAFotate BED uses Ensembl-style contig names (1, not chr1), which looks inconsistent with the rest of Talos. It is correct. SVAFotate strips the chr prefix from the query VCF but not from this reference file, so adding prefixes causes every variant to be annotated with a population frequency of zero, with no error raised — making an entire callset appear rare. Use the file exactly as downloaded.