Call copy number variants with the GATK best-practices workflows — the somatic CNV pipeline (CollectReadCounts, DenoiseReadCounts with tangent normalization, ModelSegments, CallCopyRatioSegments) and the germline GATK-gCNV pipeline (DetermineGermlineContigPloidy, GermlineCNVCaller cohort/case mode, PostprocessGermlineCNVCalls). Covers panel-of-normals construction, AnnotateIntervals/FilterIntervals, allelic-count integration, and QS-based filtering. Use when integrating CNV calling into a GATK variant pipeline, calling rare germline CNVs from an exome cohort, deciding between the somatic and germline GATK workflows, or diagnosing why tangent normalization removed a real event or why gCNV output has low precision.
Call copy number variants with the GATK best-practices workflows — the somatic CNV pipeline (CollectReadCounts, DenoiseReadCounts with tangent normalization, ModelSegments, CallCopyRatioSegments) and the germline GATK-gCNV pipeline (DetermineGermlineContigPloidy, GermlineCNVCaller cohort/case mode, PostprocessGermlineCNVCalls). Covers panel-of-normals construction, AnnotateIntervals/FilterIntervals, allelic-count integration, and QS-based filtering. Use when integrating CNV calling into a GATK variant pipeline, calling rare germline CNVs from an exome cohort, deciding between the somatic and germline GATK workflows, or diagnosing why tangent normalization removed a real event or why gCNV output has low precision.
Before using code patterns, verify installed versions match. If versions differ:
CLI: gatk --version then gatk <ToolName> --help to confirm arguments
gCNV requires a working gatkcondaenv (theano/tensorflow stack) — gatk will report if the Python environment is missing
GATK 4.5+ gCNV inference defaults are tuned for whole-exome data; whole-genome runs generally need parameter changes. If a tool reports an unrecognized argument, check the help for that exact GATK version rather than retrying.
GATK CNV Workflows
"Call CNVs the GATK way" -> GATK has two separate CNV workflows that share almost no tools. Picking the wrong one is the most common mistake.
ModelSegments + CallCopyRatioSegments produce copy-ratio segments and a minor-allele fraction per segment, and the "call" is a simple t-test emitting + / - / 0. This is not integer allele-specific copy number, not tumor purity, and not ploidy. Practitioners routinely assume parity with ASCAT/FACETS and there is none. For integer allele-specific CN, purity, ploidy, LOH state, or whole-genome-doubling status, use allele-specific-copy-number (ASCAT, Sequenza, FACETS, or PureCN — PureCN can even reuse the GATK ModelSegments segmentation as input).
AnnotateIntervals (step 1) and supplying --annotated-intervals to the PoN are frequently skipped — they enable explicit GC-bias correction and are recommended.
Case mode (--run-mode CASE) scores a new sample against the cohort *-model shards; it must use the identicalfiltered.interval_list and the same scatter count as the cohort run, or it fails or produces incomparable calls.
Failure Modes
Tangent normalization removes a real CNV
Trigger: PoN is small (< ~20 normals) or contains samples that share a recurrent CNV (e.g. a common germline CNV, or a PoN accidentally built from tumors).
Mechanism:DenoiseReadCounts projects the tumor coverage profile onto the subspace spanned by the PoN's principal components. Any copy-number pattern present in that subspace is treated as "systematic noise" and subtracted. A CNV shared by PoN members is therefore normalized out of the tumor.
Symptom: A known event (recurrent amplification/deletion, or a common germline CNV) is absent from denoisedCR.tsv; denoised profile is suspiciously flat at that locus.
Fix: Build the PoN from >= 20-40 unrelated, tumor-free, process-matched normals. Never put tumors in the PoN. Cross-check against the standardizedCR.tsv (pre-tangent) profile — if the event is there but gone after denoising, the PoN ate it.
Mistaking ModelSegments output for allele-specific integer CN
Trigger: Treating tumor.modelFinal.seg minor-allele fraction as integer minor copy number, or expecting a purity/ploidy field.
Mechanism: GATK somatic CNV models copy ratio and allele fraction but never fits the purity/ploidy grid that converts log-ratio to integer absolute CN.
Symptom: No purity/ploidy in any output; "copy number" is continuous log2; LOH is a low minor-allele fraction, not an explicit CN-LOH state.
Fix: Accept GATK somatic CNV as a relative caller. For integer ASCN, feed the data to PureCN (segmentationGATK4 reuses GATK segments), FACETS, ASCAT, or Sequenza — see allele-specific-copy-number.
FilterIntervals silently drops intervals containing real variants
Trigger: Aggressive mappability or segmental-duplication cutoffs in FilterIntervals.
Mechanism: A minimum mappability > 0 or a maximum segmental-duplication content < 1 excludes intervals overlapping segdups and low-mappability regions — exactly where many disease-relevant CNVs (e.g. recurrent genomic-disorder loci flanked by segdups) live.
Symptom: Known recurrent CNVs at segdup-mediated loci are never called; the gene of interest has no intervals in filtered.interval_list.
Fix: Inspect filtered.interval_list for genes of interest before calling. Relax mappability/segdup cutoffs for targeted analyses; for genomic-disorder loci, depth-based callers are inherently limited near segdups — confirm with an orthogonal assay.
Raw gCNV output has ~22% precision
Trigger: Using unfiltered GermlineCNVCaller / PostprocessGermlineCNVCalls output for association or de novo analysis.
Mechanism: gCNV is tuned for high recall (~95% of rare coding CNVs >= 2 exons) at the cost of precision; raw calls are dominated by false positives.
Symptom: Implausibly many rare CNVs per sample; de novo CNV rate far above the expected ~0.01-0.02/genome.
Fix: Apply the QS (quality score) filter. QS > 100 is a common starting threshold; QS > 1000 reaches ~96% precision. Also apply sample-level filters (call rate, number of CNVs per sample) per Babadi 2023.
gCNV cohort too small or mismatched
Trigger: Cohort mode with < ~100 samples, or a cohort spanning multiple capture kits / library protocols.
Mechanism: The Bayesian model needs enough technically similar samples to learn coverage bias; mixed protocols are not separable and the model misattributes batch effects to copy number.
Symptom: Unstable calls; many CNVs tracking sequencing batch; model fails to converge.
Fix: Use >= 100 process-matched samples per cohort model; split heterogeneous cohorts by capture kit. For < 100 samples, gCNV case mode against an external compatible model, or a cohort caller like ExomeDepth, is more appropriate — see germline-cnv-interpretation.
Reconciliation: GATK vs Other Callers
Pattern
Likely cause
Action
GATK somatic flat where CNVkit calls a focal event
Tangent normalization absorbed it
Check standardizedCR.tsv; rebuild PoN without the event
GATK and ASCAT disagree on a "deletion"
GATK has no purity model; the event is subclonal or impure
Trust ASCAT/FACETS integer ASCN
gCNV calls a CNV ExomeDepth misses
Different sensitivity profiles; both have poor inter-tool concordance
Require QS filtering + a second caller for rare-CNV claims
gCNV CNV count tracks batch
Cohort mixes protocols
Re-batch by capture kit
Operational rule: GATK somatic CNV output is relative copy ratio — report it as gain/loss/neutral, not absolute CN, unless downstream-fit by an allele-specific tool. gCNV calls are reportable only after QS and sample-level filtering, and rare-CNV or de novo claims need orthogonal confirmation.
Quantitative Thresholds
Threshold
Value
Source / Rationale
PoN size (somatic)
>= 20-40 normals
Larger PoN = stabler tangent subspace; small PoN over-fits
gCNV cohort size
>= ~100 technically matched
Babadi 2023 Nat Genet; model needs coverage-bias signal
gCNV QS for high precision
QS > 1000 -> ~96% precision
Babadi 2023; raw output ~22% precision, ~95% recall
Use a common-SNP list intersected with the capture targets
Expecting purity/ploidy in output
Somatic CNV does not estimate them
Use allele-specific-copy-number
References
GATK Best Practices: Somatic copy number variant discovery (CNV). Broad Institute documentation.
Babadi M et al 2023. GATK-gCNV enables the discovery of rare copy number variants from exome sequencing data. Nat Genet 55:1589
Gao GF, Oh C, Saksena G, Tabak B, Beroukhim R, Getz G et al 2022. Tangent normalization for somatic copy-number inference in cancer genome analysis. Bioinformatics 38:4677 (Tabak is a middle, not first, author).
