GATK provides two distinct, best-practices copy-number workflows: a somatic CNV pipeline for tumor samples (read counts, tangent-normalized denoising, joint copy-ratio/allele-fraction segmentation) and a germline GATK-gCNV pipeline for constitutional CNV discovery across a cohort. They share almost no tools. The somatic workflow produces relative copy-ratio segments and +/-/0 calls, not integer allele-specific copy number, purity, or ploidy. The germline workflow produces integer CN genotype VCFs but requires a large, technically matched cohort and QS-based filtering.
conda install -c bioconda gatk4
# gCNV additionally needs the GATK Python (gatkcondaenv) environment installed
Inputs: reference FASTA + .fai + .dict; an interval list (exome targets or WGS bins); a common biallelic-SNP list for allelic counts (somatic); a contig-ploidy priors table (germline). The somatic workflow needs a panel of normals; the germline cohort workflow needs >= ~100 process-matched samples.
Tell the AI agent what to do: - "Run the GATK somatic CNV workflow on this tumor-normal WES pair" - "Build a GATK panel of normals from my tumor-free control BAMs" - "Call rare germline CNVs from my 300-sample exome cohort with GATK-gCNV" - "Filter my raw gCNV calls to high precision for a de novo CNV analysis" - "Explain why a known amplification disappeared after DenoiseReadCounts"
"Run the full GATK somatic CNV pipeline on this tumor-normal WGS pair, including AnnotateIntervals and matched-normal allelic counts, and explain what the CallCopyRatioSegments output can and cannot tell me."
"My GATK panel of normals has only 8 samples. Assess whether tangent normalization is reliable and recommend how to improve it."
"Set up GATK-gCNV cohort mode for my 250-exome neurodevelopmental cohort, including DetermineGermlineContigPloidy and FilterIntervals."
"Apply QS and sample-level filters to my gCNV output so it is suitable for a de novo CNV burden analysis."
"I need integer allele-specific copy number and tumor purity. Explain why the GATK somatic CNV workflow cannot give me that and what to use instead."
"GATK and CNVkit disagree on a focal deletion. Walk through reconciling them."
AnnotateIntervals and pass --annotated-intervals for explicit GC-bias correction.FilterIntervals can drop segdup-flanked loci where genomic-disorder CNVs live; check genes of interest survive filtering.