Variant calling

Germline variant calling compared caller by caller: coverage, bcftools and VCFtools statistics, agreement between callsets, SnpEff consequences and the genes that carry them, next to the pipeline's own MultiQC report.

nf-co.re GitHub

Draft

Template version 3.10.0 latest

The sarek template follows a variant-calling run from the reads to the genes the calls land on, and reads the VCFs themselves, so callers are compared call by call and gene by gene rather than only by how many calls each made:

• Overview: the run in four key figures, live findings and four figures, each linked to the tab that explains it
• Data & QC: read, mapping and recalibration panels from the pipeline's own MultiQC report, and the depth over the intervals with an X against Y sex check
• Variant calls: what each caller called and of what kind, how clean each callset looks, and which calls the callers share
• Annotation: SnpEff's consequences recomputed per call, the genes that carry the calls, and depth and calls at one region on one axis

The persistent Sample filters (status, patient, sample) sit in the left panel, and a pick there reaches every per-sample collection and the MultiQC panels through the sample links.

Concordance is agreement, not truth

Nothing in this template compares a callset against a benchmark truth set: "concordance" means agreement between callers and samples. For a truth-set benchmark, nf-core/variantbenchmarking has its own template.

Structural-variant callers call few SNPs

Manta and TIDDIT report a Ts/Tv of 0.0 in bcftools stats, which is the caller doing its job. Every Ts/Tv card and filter keeps only ts_tv > 0, so they do not drag the medians down.

---

Quick start

Point at a finished runFrom the pipeline itself (v1.10.0+)

depictio ingest /path/to/sarek_results \
  --template nf-core/sarek/latest

The results directory is the only thing you have to pass. The sample hub is built from the CSV manifests sarek writes under csv/ (patient, sex, status and the callers run per sample), so the launch samplesheet is not needed. If the run was aligned to another assembly than hg38, name it so the locus tracks draw the right axis and gene lane:

depictio ingest /path/to/sarek_results \
  --template nf-core/sarek/latest \
  --var GENOME=hg19

Variable	Default	Role
GENOME	hg38	UCSC name of the assembly (hg38, hg19, mm10): the axis and gene lane of every locus track

depictio config nextflow --install     # once per machine
nextflow run nf-core/sarek -r 3.10.0 -profile docker --outdir results

No depictio ingest, and no template named: the pipeline ingests its own output directory when it finishes and resolves this template from its own manifest. See Nextflow trigger.

---

Reference

The template reads the native MultiQC report, the per-caller bcftools stats and VCFtools summaries, mosdepth's coverage files, the called VCFs, and SnpEff's annotated VCFs, CSV statistics and per-gene tables. The caller and the sample live only in the directory path (reports/<tool>/<caller>/<sample>/), so raw scans carry the path into the table and recipes read both off it. sarek runs mosdepth on the duplicate-marked and on the recalibrated CRAM; the coverage recipes keep one pass per sample, so no tile counts a sample twice.

Self-adapting layout

The dashboard adapts to whatever the run actually produced: components bound to pruned or unparsed data collections are hidden, tabs left with no real visualizations are dropped entirely, and the remaining components are re-packed so there are no empty rows. A run without SnpEff in --tools drops the Consequences and Genes tabs instead of failing to ingest.

Template variables

Variables you provide when running the template: DATA_ROOT is the results directory given to depictio ingest, the others are set with --var NAME=value:

Variable	Required	Description
DATA_ROOT	✓	Root directory of the sarek output (csv/, reports/, variant_calling/, annotation/, multiqc/, pipeline_info/).
GENOME	—	UCSC name of the assembly the run was aligned to (hg38, hg19, mm10): the axis and gene lane of every locus track.

Data collections

37 data collections — 17 required 20 optional · 17 direct 20 derived.

Origin tells you whether a collection is real pipeline data or a reshape of it: direct = a pipeline output (scanned, or a recipe that reads raw files); derived = a recipe that reshapes one or more direct collections into the layout a visualization needs (no new measurement). Reads shows what produces it: a recipe .py transform, whose name links to its source on GitHub, or a raw file scanned off disk. (A direct collection can still have a recipe — one that merely parses/cleans the raw file; derived means the recipe reshapes another collection.)

Tag	Origin	Type	Reads	Status
samples	direct	Table	recipe nf-core/sarek/samples.py	required
multiqc_data	direct	MultiQC	file (?:.*/)?multiqc(?:/[^/]+)?/multiqc_data/multiqc.parquet$	required
bcftools_stats_raw	direct	Table	file .*.bcftools_stats.txt$	required
bcftools_stats_summary	derived	Table	recipe bcftools/stats_summary.py	required
bcftools_stats_tstv	derived	Table	recipe bcftools/stats_tstv.py	required
bcftools_stats_sections	derived	Table	recipe bcftools/stats_sections.py	required
bcftools_substitution_spectrum	derived	Table	recipe nf-core/sarek/substitution_spectrum.py	required
bcftools_indel_spectrum	derived	Table	recipe nf-core/sarek/indel_spectrum.py	required
mosdepth_regions_raw	direct	Table	file .*.regions.bed.gz$	required
mosdepth_regions	derived	Table	recipe mosdepth/regions.py	required
mosdepth_targets	derived	Table	recipe nf-core/sarek/mosdepth_targets.py	required
mosdepth_windows	derived	Table	recipe nf-core/sarek/mosdepth_windows.py	required
mosdepth_summary_raw	direct	Table	file .*.mosdepth.summary.txt$	required
mosdepth_summary	derived	Table	recipe mosdepth/summary.py	required
mosdepth_xy_sex_check	direct	Table	recipe mosdepth/xy_sex_check.py	optional
vcftools_filter_summary_raw	direct	Table	file .*.FILTER.summary$	optional
vcftools_filter_summary	derived	Table	recipe vcftools/filter_summary.py	optional
vcftools_tstv_qual_raw	direct	Table	file .*.TsTv.qual$	optional
vcftools_tstv_qual	derived	Table	recipe vcftools/tstv_qual.py	optional
vcf_calls_raw	direct	Table	file ^(?!._(?:snpEff|VEP).ann.vcf.gz$)(?!..g(?:enome)?.vcf.gz$).*.vcf.gz$	required
vcf_variants	derived	Table	recipe vcf/variants.py	required
snpeff_ann_vcf_raw	direct	Table	file .*_snpEff.ann.vcf.gz$	optional
snpeff_ann_variants	derived	Table	recipe snpeff/ann_variants.py	optional
snpeff_variant_upset	derived	Table	recipe snpeff/variant_upset.py	optional
snpeff_protein_lollipop	derived	Table	recipe snpeff/protein_lollipop.py	optional
snpeff_csv_raw	direct	Table	file .*_snpEff.csv$	optional
snpeff_csv_stats	derived	Table	recipe snpeff/csv_stats.py	optional
snpeff_genes_raw	direct	Table	file .*_snpEff.genes.txt$	optional
snpeff_genes	derived	Table	recipe snpeff/genes.py	optional
snpeff_gene_upset	derived	Table	recipe snpeff/gene_upset.py	optional
snpeff_gene_heatmap	derived	Table	recipe snpeff/gene_heatmap.py	optional
callset_qc	derived	Table	recipe nf-core/sarek/callset_qc.py	required
snpeff_vcf_files	direct	indexed_file	file ^(?!..(?:manta|tiddit).)._snpEff.ann.vcf.gz$	optional
ascat_segments	direct	Table	file .*.cnvs.txt$	optional
cnvkit_segments	direct	Table	file .*.call.cns$	optional
msisensorpro_summary	direct	Table	file .msisensorpro.(?<!_dis)(?<!_germline)(?<!_somatic)$	optional
ngscheckmate_matches	direct	Table	file .*_all.txt$	optional

Cross-DC links

33 links — selecting a value in the source collection filters the target. The join column is shown after the source.

Source · column		Target	Filters
samples · sample_id	→	multiqc_data	Filter every MultiQC panel by the samples picked in the left panel, through sarek's per-tool sample names
samples · sample_id	→	mosdepth_regions	Filter the coverage track by the samples picked in the left panel
samples · sample_id	→	mosdepth_summary	Filter the per-contig coverage summary by the samples picked in the left panel
samples · sample_id	→	mosdepth_xy_sex_check	Filter the sex check by the samples picked in the left panel
samples · sample_id	→	bcftools_stats_sections	Filter the bcftools distributions by the samples picked in the left panel
samples · sample_id	→	vcftools_filter_summary	Filter the FILTER breakdown by the samples picked in the left panel
samples · sample_id	→	vcftools_tstv_qual	Filter the quality sweeps by the samples picked in the left panel
samples · sample_id	→	vcf_variants	Filter the variant calls by the samples picked in the left panel
samples · sample_id	→	snpeff_ann_variants	Filter the annotated calls by the samples picked in the left panel
samples · sample_id	→	snpeff_protein_lollipop	Filter the coding variants by the samples picked in the left panel
samples · sample_id	→	snpeff_csv_stats	Filter the annotation composition by the samples picked in the left panel
samples · sample_id	→	snpeff_genes	Filter the per-gene burden by the samples picked in the left panel
bcftools_stats_summary · caller	→	bcftools_stats_sections	Carry the caller picked on the variant-yield tab over to the bcftools distributions
bcftools_stats_summary · caller	→	vcftools_filter_summary	Carry the caller picked on the variant-yield tab over to the FILTER breakdown
bcftools_stats_summary · caller	→	vcftools_tstv_qual	Carry the caller picked on the variant-yield tab over to the quality sweeps
vcf_variants · caller	→	snpeff_ann_variants	Carry the caller picked on the concordance tab over to the annotated calls
snpeff_csv_stats · caller	→	snpeff_ann_variants	Carry the caller picked on the consequences tab over to the annotated calls
snpeff_genes · caller	→	snpeff_protein_lollipop	Carry the caller picked on the genes tab over to the coding variants
snpeff_genes · gene_name	→	snpeff_protein_lollipop	Open a gene picked in the burden table on the lollipop below it
samples · sample_id	→	bcftools_stats_summary	Filter the per-caller variant counts by the samples picked in the left panel
samples · sample_id	→	bcftools_stats_tstv	Filter the per-caller Ts/Tv ratios by the samples picked in the left panel
bcftools_stats_summary · caller	→	bcftools_stats_tstv	Carry the caller picked on the MultiQC tab over to its Ts/Tv rows
bcftools_stats_summary · sample	→	bcftools_stats_tstv	Carry a sample selected on the variant counts table over to its Ts/Tv rows
mosdepth_windows · chrom	→	mosdepth_targets	A region typed or brushed on the depth navigator narrows the per-target depth track
mosdepth_windows · chrom	→	vcf_variants	A region typed or brushed on the depth navigator narrows the calls track
samples · sample_id	→	mosdepth_windows	Filter the depth navigator by the samples picked in the left panel
samples · sample_id	→	mosdepth_targets	Filter the per-target depth track by the samples picked in the left panel
samples · sample_id	→	bcftools_substitution_spectrum	Filter the substitution spectrum by the samples picked in the left panel
samples · sample_id	→	bcftools_indel_spectrum	Filter the indel length spectrum by the samples picked in the left panel
samples · sample_id	→	callset_qc	Filter the callset QC profile by the samples picked in the left panel
bcftools_stats_summary · caller	→	bcftools_substitution_spectrum	Carry the caller picked on the Variant yield tab over to the substitution spectrum
bcftools_stats_summary · caller	→	bcftools_indel_spectrum	Carry the caller picked on the Variant yield tab over to the indel length spectrum
bcftools_stats_summary · caller	→	callset_qc	Carry the caller picked on the Variant yield tab over to the callset QC profile

Recipes

Each recipe reshapes raw pipeline output into a tidy table. The name links to its source; Output lists the validated OUTPUT_SCHEMA columns.

Recipe	Transforms	Output
bcftools/stats_sections.py	The distribution blocks of bcftools stats, melted into one long frame.	sample, caller, section, bin, label, count
bcftools/stats_summary.py	Per-sample, per-caller variant record counts from bcftools stats.	sample, caller, n_records, n_no_alts, n_snps, n_mnps, n_indels, n_others, n_multiallelic_sites, n_multiallelic_snp_sites, snp_fraction, log10_n_records
bcftools/stats_tstv.py	Per-sample, per-caller transition/transversion ratio from bcftools stats.	sample, caller, ts, tv, ts_tv, ts_1st_alt, tv_1st_alt, ts_tv_1st_alt
mosdepth/regions.py	Per-target mosdepth depth, binned into a genome-wide coverage track.	chromosome, position, end, value, sample, stage, sample_stage, n_targets
mosdepth/summary.py	Per-contig coverage summary written by mosdepth.	sample, stage, scope, chrom, length, bases, mean_depth, min_depth, max_depth
mosdepth/xy_sex_check.py	X and Y mean depth per library, and the sex the ratio is consistent with.	sample, stage, x_coverage, y_coverage, xy_ratio, inferred_sex
nf-core/sarek/callset_qc.py	One QC profile per callset (sample and caller), for a parallel-coordinates tile.	callset, sample, caller, n_records, snp_fraction, ts_tv, multiallelic_pct, pass_pct, het_hom_ratio, median_dp, median_het_vaf
nf-core/sarek/indel_spectrum.py	Indel length spectrum per callset, from the IDD block of bcftools stats.	sample, caller, callset, length, indel_class, count, fraction
nf-core/sarek/mosdepth_targets.py	Per-target mosdepth depth, unbinned, for the Cohort QC locus section.	chrom, pos, end, depth, sample, stage, sample_stage, target_bp
nf-core/sarek/mosdepth_windows.py	The 1 Mb mosdepth windows under the column names the locus section shares.	chrom, pos, end, depth, sample, stage, sample_stage, n_targets
nf-core/sarek/samples.py	One row per sarek sample, from the CSV manifests sarek publishes under csv/.	sample_id, patient, sex, status, status_label, n_callers
nf-core/sarek/substitution_spectrum.py	SNV substitution spectrum per callset, folded onto the six pyrimidine classes.	sample, caller, callset, substitution_class, mutation_class, count, fraction
snpeff/ann_variants.py	One row per annotated variant call, read from SnpEff's annotated VCF.	sample, caller, chrom, pos, variant_key, ref, alt, variant_type, qual, filter_status, is_pass, gt, dp, vaf, gene, gene_id, impact, consequence, hgvs_p, aa_pos
snpeff/csv_stats.py	The composition sections of SnpEff's CSV summary, melted into one long frame.	sample, caller, section, category, count, percent
snpeff/gene_heatmap.py	Gene by callset burden matrix, for a clustered heatmap.	gene_name
snpeff/gene_upset.py	Gene-level caller overlap: one row per gene, one 0/1 column per caller.	gene_name
snpeff/genes.py	Per-gene variant counts by impact class, from SnpEff's gene effect table.	sample, caller, gene_name, gene_id, biotype, n_high, n_low, n_moderate, n_modifier, n_variants, n_coding
snpeff/protein_lollipop.py	Coding variants positioned along their protein, for a lollipop track.	sample, caller, gene, aa_pos, impact, consequence, hgvs_p, variant_key, vaf
snpeff/variant_upset.py	Variant-level caller overlap: one row per variant, one 0/1 column per caller.	variant_key
vcf/variants.py	One row per variant call, read from the VCFs themselves.	sample, caller, chrom, pos, variant_key, ref, alt, variant_type, qual, filter_status, is_pass, gt, dp, vaf
vcftools/filter_summary.py	How a caller's own FILTER column partitions its calls.	sample, caller, filter_label, n_variants, n_ts, n_tv, ts_tv, is_pass, pct_of_calls
vcftools/tstv_qual.py	Transition/transversion ratio as a function of the quality floor.	sample, caller, series, qual_threshold, n_ts_above, n_tv_above, ts_tv_above

---

Dashboard tabs

One dashboard: the Overview, then eight child tabs in three groups, read as a funnel from the reads to the genes the calls hit and one region of the genome. Each tab below carries the same icon and colour the dashboard gives it, so the page and the app read alike.

Group	Tabs
Data & QC	MultiQC, Coverage
Variant calls	Variant yield, Call quality, Caller concordance
Annotation	Consequences, Genes, Locus

Each child tab opens with a short intro and a strip of four cards, then at most three open sections; tables follow, collapsed. The template has no group column: the sample hub is built from sarek's CSV manifests and carries the patient, sex, status and the callers run per sample. A callset is one caller on one sample. The persistent Sample filters (status, patient, sample) sit in the left panel and narrow every tab through the sample links. The Sample sheet is pinned, collapsed, to the bottom of every child tab.

Overview MultiQC Coverage Variant yield Call quality Caller concordance Consequences Genes Locus

Germline variant calls, from coverage to the genes they hit.

A short hero links the run parameters. About this dashboard says how the two filter levels work, The run lists the samples, the callsets and the callers they come from, the genome and the aligner, and Pipeline walks the six steps from mapping to annotation, each linked to the parameters that drive it and its tab. The findings are live values: they follow the filters, and a run without SnpEff drops the HIGH-impact row and the impact figure.

Filters and components

Filters · the left panel starts collapsed. Key figures and Findings each have their own filter bar (the caller and the sample id): each narrows its own section only.

Section	What it holds
Top	Hero, About this dashboard, The run, Pipeline
Key figures	4 headline cards: variant calls by caller, depth on target, PASS share, Ts/Tv against a 1.8 floor
Findings	Live result rows, then 4 figures: the substitution spectrum, the calls per FILTER value, allele fraction against depth and the impact classes per caller
How to read this dashboard	The tabs by group, each with its question

Data & QC · Did reads, mapping and recalibration work for every sample?

MultiQC panels only, from the run's native report. Open: general statistics, fastp filtered reads, Samtools percent mapped, MarkDuplicates and the mosdepth cumulative coverage. Collapsed: the FastQC panels, the BQSR fit and insert sizes, the bcftools and VCFtools panels with every caller pooled, and the SnpEff and VEP panels. A library carries a lane and a read suffix in the report, so a sample can appear once per read.

Filters and components

Filters · Sample ID, read from the MultiQC report.

Section	What it holds
QC overview	5 MultiQC panels
Read quality (collapsed)	3 MultiQC panels
Alignment details (collapsed)	2 MultiQC panels
Variant-call panels (collapsed)	4 MultiQC panels
Annotation panels (collapsed)	4 MultiQC panels

Data & QC · How deeply were the intervals covered, sample by sample?

mosdepth's depth over the intervals (the capture targets, or the calling intervals of a whole genome) and over each contig. The depth on target per sample, the median depth of an interval against a 20x floor, the depth across contigs and the intervals under 20x, with the contigs that have most. Then the mean depth per contig, one bar per sample and contig, with a bar above it that switches between the intervals and the whole contig (it opens on the intervals), and the X against Y depth scatter, a heuristic sex check. The per-contig and sex-check tables are collapsed.

Filters and components

Filters · a Mean depth range on mosdepth_summary. Depth per contig has its own bar, Intervals or whole contig. There is no contig or scope picker in the left panel: it would empty the Depth on target card, which reads mosdepth's total row of the intervals.

Section	What it holds
Coverage at a glance	4 cards
Depth per contig	Mean depth per contig, with its own filter bar
Sex check	1 advanced visualization
Coverage tables (collapsed)	Per-contig coverage summary, X and Y coverage per sample

Variant calls · How much did each caller call, and of what kind?

Each caller's own bcftools stats report. The median records per callset, ranked by caller (a structural-variant caller writes orders of magnitude fewer), the SNPs and the indels split by caller, and the multiallelic sites. Then SNPs and indels per caller, one bar per sample, and the substitution and indel length spectra, one bar per caller with its callsets' shares averaged. The spectra compare shapes, not yields: a caller whose transversion bars rise is the one with the lower Ts/Tv. The bcftools distribution blocks and the per-caller count table are collapsed.

Filters and components

Filters · Caller on bcftools_stats_summary, and a bcftools block picker on bcftools_stats_sections that chooses the block the collapsed distribution draws (it opens on depth).

Section	What it holds
Yield at a glance	4 cards
Calls per caller	SNPs per caller, Indels per caller
Mutation spectra	Substitution spectrum, Indel length spectrum
bcftools distributions (collapsed)	The chosen block, caller by caller, bcftools stats distributions
Yield table (collapsed)	Per-caller variant counts

Variant calls · Do the callsets look like clean germline calls?

QC numbers computed from each callset's VCF. The median Ts/Tv against a 1.8 floor, the PASS share, the het to hom ratio and the allele fraction of heterozygous calls. Then the callset QC profile, six numbers per callset on parallel axes coloured by caller, the calls per FILTER value per caller, with PASS in green, and the Ts/Tv above a rising quality floor on a log axis, because callers write QUAL on scales a thousand-fold apart. A clean germline SNP callset has a Ts/Tv near 2 on a genome, higher on an exome, and its heterozygous calls sit near an allele fraction of 0.5. The FILTER breakdown and the Ts/Tv per callset are collapsed.

Filters and components

Filters · Caller and a Ts/Tv ratio range on callset_qc, and FILTER value on vcftools_filter_summary.

Section	What it holds
Quality at a glance	4 cards
Callset QC profile	1 advanced visualization
FILTER partitions	Calls per FILTER value
Quality calibration	1 advanced visualization
Quality tables (collapsed)	FILTER breakdown per caller, Ts/Tv per callset

Without VCFtools

A run that skipped VCFtools has no FILTER partitions, no quality sweep and no FILTER value filter here, and the Overview drops its FILTER figure.

Variant calls · Which calls do the callers share, and where do they differ?

One row per call, read from the VCFs. The calls ranked by caller, the calls by variant type, the allele fraction and the depth at the call. Then the UpSet of PASS calls shared between callsets, matched on locus and allele (fixed: the pickers on this tab do not narrow it), allele fraction against depth as a density beside its histogram per caller (a clean diploid callset forms ridges at 0.5 and 1), and the rainfall of the calls along the genome, each at its distance to the previous call. The call table is collapsed. Agreement between callers is not truth.

Filters and components

Filters · Caller, Variant type, Contig and a Depth at the call range on vcf_variants.

Section	What it holds
Concordance at a glance	4 cards
Shared calls	1 advanced visualization
Depth sensitivity	1 advanced visualization, Allele fraction per caller
Calls along the genome	1 advanced visualization
Call table (collapsed)	Variant calls

Annotation · What do the calls do to the transcripts?

SnpEff gives each call its most severe effect on a transcript, in four impact classes. The annotated calls by impact class, the HIGH-impact calls by variant type, the genes with a HIGH or MODERATE call and the depth at HIGH-impact calls. Then SnpEff's own counts for the summary section picked in the left panel (opening on impact), caller by caller, and the impact recomputed on the calls: the impact classes per caller as shares (MODIFIER left out), and the allele fraction against depth scatter coloured by impact, with the record of the picked call beside it. The SnpEff summary and the annotated calls are collapsed.

Filters and components

Filters · Caller on snpeff_ann_variants and SnpEff section on snpeff_csv_stats. Impact per call has its own bar, Impact class and Consequence, rather than the left panel, because three cards pin an impact class.

Section	What it holds
Consequences at a glance	4 cards
What kind of variants	The chosen section, caller by caller
Impact per call	Impact classes per caller, then 2 advanced visualizations: the impact scatter and the Variant record, with its own filter bar
Consequence tables (collapsed)	SnpEff summary sections, Annotated variant calls

Annotation · Which genes carry the calls, and where on the protein?

SnpEff's per-gene counts. The genes with a call by biotype, the HIGH-impact variants by caller, the top gene burden (the most HIGH or MODERATE variants on one gene, the callers' maxima) and the protein changes by impact class. Then the gene by callset burden heatmap, coloured by log(1 + variants) so one long gene does not wash out the rest, the UpSet of genes the callers share (fixed), and the coding variants along the protein, one lane per gene. A high burden is often a long, repetitive or poorly mappable gene before it is a biological signal. The per-gene burden and protein position tables are collapsed.

Filters and components

Filters · Biotype and Caller on snpeff_genes, and Impact class on snpeff_protein_lollipop.

Section	What it holds
Genes at a glance	4 cards
Burden across callsets	1 advanced visualization
Genes the callers share	1 advanced visualization
Along the protein	1 advanced visualization
Gene tables (collapsed)	Per-gene variant burden, Coding variants with a protein position

Annotation · What do depth and calls show at one region?

The locus tab stacks four tracks on one {GENOME} axis. The depth navigator, the mean depth in 1 Mb windows, drives the others: a typed locus or gene, or a brush on its axis, moves the depth per interval, the calls over the gene lane (one lane per caller) and the annotated VCFs, range-read from their files below 2 Mb. It opens on chr1:1,000,000-2,000,000. The four cards (calls by caller, depth per interval, allele fraction, depth at the call) are recounted on the region in view.

Filters and components

Filters · Caller and Variant type on vcf_variants, for the call track and the cards.

Section	What it holds
Region at a glance	4 cards
One region, four tracks	4 advanced visualizations: the depth navigator, the depth per interval, the calls over the genes and the annotated VCFs

Tables and point views select on their entity column: the sample sheet on sample_id; the contig, sex-check, FILTER and distribution tables and the sex-check scatter on sample; the bcftools and SnpEff summary tables on caller; the rainfall plot, the impact scatter and the call tables on variant_key; the per-gene table on gene_name, which reaches the protein lollipop. A pick narrows the other tiles of its collection and follows the project links to the collections downstream of it. The variant record on Consequences shows the call picked in the scatter beside it.

---

Running the pipeline

Depictio reads the output of nf-core/sarek, it does not run the pipeline. Run the pipeline first, with the callers you want to compare and SnpEff among the tools:

nextflow run nf-core/sarek -r 3.10.0 \
  --input samplesheet.csv \
  --genome GATK.GRCh38 \
  --tools haplotypecaller,deepvariant,strelka,freebayes,manta,snpeff \
  --outdir results -profile docker

Then point Depictio at the results. sarek 3.10.0 ships a MultiQC parquet, so no reprocess step is needed:

depictio ingest results/ --template nf-core/sarek/latest

See nf-co.re/sarek/usage for full pipeline documentation.

---

Required data structure

Point depictio ingest at the directory holding the pipeline output. Depictio scans recursively and matches on file name; the caller and sample directories under reports/, variant_calling/ and annotation/ are read off the path.

<DATA_ROOT>/
├── csv/*.csv                                  # the hub: sarek's resume manifests
├── pipeline_info/
│   ├── params_*.json
│   └── *software*versions*.yml
├── multiqc/multiqc_data/multiqc.parquet       # native, no reprocess
├── reports/
│   ├── bcftools/<caller>/<sample>/*.bcftools_stats.txt
│   ├── vcftools/<caller>/<sample>/*.{TsTv.qual,FILTER.summary}   # optional
│   ├── mosdepth/<sample>/*.{mosdepth.summary.txt,regions.bed.gz}
│   └── snpeff/<caller>/<sample>/*_snpEff.{csv,genes.txt}         # optional
├── variant_calling/<caller>/<sample>/*.vcf.gz # one VCF per sample and caller
└── annotation/<caller>/<sample>/
    └── *_snpEff.ann.vcf.gz{,.tbi}             # optional; the index feeds the locus track

gVCFs are not read. The somatic outputs (ASCAT, CNVkit, MSIsensor-pro, NGSCheckMate) are declared optional so a somatic run ingests, but no tile binds them yet.

---

Validation runs

The template was validated against the AWS megatest of the 3.10.0 release, results-8ccac7ad37b05dd792447763bf9671b719824587, on its test_full_germline_ncbench_agilent/ profile: one NA12878 exome sequenced at two depths, five germline callers (DeepVariant, FreeBayes, HaplotypeCaller, Manta, Strelka) and both annotators. The screenshots above come from that run. megatest.yaml lists the tables-only subset the template needs, plus the per-caller VCFs and their SnpEff twins:

bash depictio/projects/nf-core/sarek/3.10.0/download_test_data.sh /tmp/sarek_test
depictio ingest /tmp/sarek_test --template nf-core/sarek/latest

Do not pass --project when ingesting: the dashboard is attached to the project by name, so renaming it breaks a later depictio dashboard import. Re-ingesting accumulates dashboards, so delete the project before repeating a run.

---

Additional resources

• nf-co.re/sarek: official pipeline documentation
• nf-co.re/sarek/3.10.0/results: AWS test results
• Template System Reference: YAML format, variables, conditionals
• Recipes: how to read, test, and write recipes

---

Authorship

Developers Wrote the template, its recipes and its dashboards. weber8thomas

Reviewers Nobody has run it on their own data and signed it off yet, which is what keeps it a Draft. Open

Maintainers Keep it working as nf-core/sarek releases. weber8thomas

Reviewing a template on your own data, or taking over a role here, is a contribution in itself: the contributing guide says what each one involves.