Project Name: UMGC_CNV_PIPELINE
Run name: 603989-20260611_1726
Sample: GIAB_NA24385
Sex: Male
Deletions: 1,337
Duplications: 788
Total CNV calls: 2,125
CNV Report Guide: Link to Guide
Report Generated: 2026-07-20 07:02

Genome Overview — click a chromosome row to zoom in

This Genome Overview places each CNV call along its chromosome (deletions in red, duplications in blue); click a chromosome row to zoom into a detailed, pannable view.

Deletions Duplications

CNV Confidence Score Distribution

Each CNV call is scored from 0 to 100 by a machine-learning classifier (the “tree score”); calls scoring 31 or above are high-quality (PASS). This plot shows the score distribution for all calls — those kept and those filtered out — with the dashed line marking the PASS threshold.

CNV Size Distribution

Counts of the high-quality deletions and duplications grouped into six size bins, showing how large the calls are.

Data — files in the output folder

bed/SampleName.cnv.bed

The high-quality CNV call set in BED format. Each row is a deletion or duplication with its genomic coordinates, type, ML score, copy number, and calling source. This is the recommended file for downstream analysis.

cnv_calls/SampleName.cnv_calls.tsv

Tab-delimited table of the high-quality CNV calls, with one column each for sample, chrom, start, end, svtype, filter, tree_score, svlen, source, and cn. Convenient for spreadsheets or scripting.

report/SampleName.cnv_report.html

This interactive HTML report.

vcf/SampleName.mrg.vcf.gz

The high-quality CNV calls as a bgzip-compressed VCF with full INFO annotations; suitable for genome browsers and standard VCF tools.

raw/SampleName.raw_dump.vcf.gz

The complete, unfiltered call set (high-quality and low-quality) with all INFO annotations. Filter to high-quality calls with bcftools view -f PASS to reproduce the files above. Use this when you need the full results, including the calls the quality filter removed.

Methods

Copy-number variants (CNVs) were called from the aligned reads using the Ultima Genomics germline CNV pipeline (protocol) on a HPC cluster. Per-base read coverage was normalized against a reference cohort and segmented with cn.mops (v1.55.0), and read-depth CNVs were independently called with CNVpytor (v1.3.1) at 500 bp and 2500 bp resolution. Calls from both callers were merged, breakpoints were refined by local re-alignment of split reads (jalign, ugbio_cnv v1.24.0), and each candidate CNV was scored by a machine-learning decision-tree classifier; calls with a TREE_SCORE ≥ 31 are flagged PASS. Redundant overlapping calls were then collapsed to produce the final call set reported.