henhao
Sign in
GenomicsNature
Peer-reviewed journal article

An open resource of structural variation for medical and population genetics

Ryan L. CollinsiD, Harrison BrandiD, Konrad J. KarczewskiiD, Xuefang ZhaoiD, Jessica AlföldiiD, Laurent C. FrancioliiD, Amit V. KheraiD, Chelsea LowtheriD, Laura Doyle Gauthier, Harold Wang, Nicholas A. Watts, Matthew SolomonsoniD, Anne O’Donnell‐LuriaiD, Alexander Baumann, Ruchi Munshi, Mark WalkeriD, Christopher W. WhelaniD, Yongqing HuangiD, Ted Brookings, Ted Sharpe, Matthew R. StoneiD, Elise ValkanasiD, Jack M. FuiD, Grace TiaoiD, Kristen M. LaricchiaiD, , Adolfo CorreaiD

PDFDOIPreprintData
Post to X
Threads on XPost to BlueskyThreads on Bluesky
In-depth full-text digest available · Jump to digest ↓

Abstract

Structural variants (SVs) rearrange large segments of the genome and can have profound consequences for evolution and human diseases. As national biobanks, disease association studies, and clinical genetic testing grow increasingly reliant on genome sequencing, population references such as the Genome Aggregation Database (gnomAD) have become integral for interpreting genetic variation. To date, no large-scale reference maps of SVs exist from high-coverage sequencing comparable to those available for point mutations in protein-coding genes. Here, we constructed a reference atlas of SVs across 14,891 genomes from diverse global populations (54% non-European) as a component of gnomAD. We discovered a rich landscape of 433,371 distinct SVs, including 5,295 multi-breakpoint complex SVs across 11 mutational subclasses, and examples of localized chromosome shattering, as in chromothripsis. The average individual harbored 7,439 SVs, which accounted for 25-29% of all rare protein-truncating events per genome. We found strong correlations between constraint against damaging point mutations and rare SVs that both disrupt and duplicate protein-coding sequence, suggesting intolerance to reciprocal dosage alterations for a subset of tightly regulated genes. We also uncovered modest selection against noncoding SVs in cis-regulatory elements, although selection against protein-truncating SVs was stronger than any effect on noncoding SVs. Finally, we benchmarked carrier rates for medically relevant SVs, finding very large ([≥]1Mb) rare SVs in 3.8% of genomes (~1:26 individuals) and clinically reportable incidental SVs in 0.18% of genomes (~1:556 individuals). These data have been integrated directly into the gnomAD browser (https://gnomad.broadinstitute.org) and will have broad utility for population genetics, disease association, and diagnostic screening.

02 /Specification & Classification

Click any tag to browse matching papers
Citations
1,037
FWCI
86.20
Field percentile
top 1%
OpenAlex Topics

In-depth summary

✦ Multi-agent verified full-text synthesis

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (21 Oct 2019) · CC BY 4.0

Short-read genome sequencing of 14,891 diverse genomes yields 433,371 structural variants (SVs) for gnomAD; loss-constrained genes also resist duplication, and clinical SV carrier rates are benchmarked.

Population references like gnomAD underpin interpretation of point mutations, yet WGS-based SV references were confined to the low-coverage 1000 Genomes Project (N=2,504) or small European-centric cohorts. This work supplies a high-coverage (32×), ancestrally diverse SV map with selection and clinical carrier analyses, openly integrated into the gnomAD browser.

Key results

  1. ShownA whole-genome sequencing (WGS) reference of structural variation spans hundreds of thousands of SVs, including thousands of fully resolved multi-breakpoint complex SVs, most of which contain inverted segments.

    433,371 SVs discovered, 335,470 completely resolved after removing breakends (13.9% of calls); 5,295 complex SVs in 11 subclasses, 3,901 (73.7%) with inversions; 7,439 SVs per genome at ~32× vs 3,441 in 1000 Genomes (~7×), 3,658 in GTEx (~50×), and 24,825 from ~40× long-read WGS. · Abstract; SV discovery and genotyping; Insights into population genetics and genome…; DISCUSSION

  2. ShownThe callset follows Mendelian inheritance, Hardy-Weinberg and linkage expectations and is largely confirmed by long-read sequencing, supporting adequate specificity for most human genomics uses.

    Across 970 trios, Mendelian violation rate 3.8% and heterozygous de novo rate 3.0%; 86% of SVs in Hardy-Weinberg equilibrium; common SVs showed median peak R^2 = 0.85 with nearby SNVs/indels; PacBio data from four individuals confirmed 94.0% of SVs with breakpoint-level read support (92.8% of SVs); 59.8% of breakpoints were within a single nucleotide of long-read calls, 75.9% within ±10bp. · SV discovery and genotyping; Figure legends

  3. ShownRare SVs causing predicted loss-of-function (pLoF), whole-gene copy gain, or exonic duplication are depleted in genes constrained against pLoF point mutations, implying many such genes are intolerant to both dosage loss and gain.

    Spearman rho = 0.90 between SNV pLoF constraint and depletion of rare pLoF SVs; rho = 0.78 for whole-gene copy gain and 0.58 for intragenic exonic duplication. The mean genome had 10.2 genes altered by rare SVs (5.5 pLoF) vs 16.3 rare pLoF SNVs/indels, an estimated 25-29% SV share of rare pLoF events. · Dosage sensitivity of protein-coding genes and…

  4. ShownNoncoding copy-number variants (CNVs) hitting annotated cis-regulatory elements show weak but widespread signals of purifying selection, always below protein-truncating SVs; deletions and CNVs spanning whole elements are under stronger selection.

    Adjusted proportion of singletons (APS) was elevated for noncoding CNVs overlapping most of 14 element classes (one-tailed t-tests, Bonferroni over 32 comparisons), none matching coding pLoF; CNVs overlapping no element had APS not different from zero; full-element overlap exceeded partial (paired t-test); phastCons conservation correlated with APS. · Dosage sensitivity of protein-coding genes and…; Figure legends

  5. ShownPopulation carrier rates for medically relevant SVs are benchmarked, and WGS-derived genomic disorder CNV frequencies agree with microarray-based estimates.

    3.8% of genomes (95% CI: 3.2-4.6%) carry a rare autosomal SV ≥1Mb; 0.32% carry a very rare (allele frequency < 0.1%) pLoF SV in a clinically reportable gene (0.18% pathogenic/likely pathogenic); 7.22% are heterozygous for rare pLoF SVs in recessive developmental disorder genes; 49 genomic disorder carrier rates matched UK BioBank microarray (R^2 = 0.669; P = 7.38 × 10^−13). · Abstract; Relevance to disease association and clinical…

Methods, data and reproducibility

A cloud-based multi-algorithm pipeline integrating four short-read SV signatures discovered and genotyped canonical and complex SVs across aggregated ~32× genomes; selection was quantified with an Adjusted Proportion of Singletons (APS) metric correcting for SV size, class and genomic context, and per-gene rare-SV depletion was compared with SNV constraint.

  • gnomAD-SV short-read WGS genomes (average 32× coverage) pooled from many sequencing projects; general adult population, 54% non-European · human · n = 14,891 (14,237 post-QC; 12,653 unrelated) · NCBI dbVar nstd166
  • Parent-child trios used for family-based quality assessment (Mendelian violation and apparent de novo rates) · human · n = 970 trios
  • Matched PacBio long-read WGS used for in silico confirmation of short-read SV calls and breakpoint accuracy · human · n = four individuals
  • Unrelated samples with no known neuropsychiatric disease, used for carrier frequencies of 49 genomic disorder CNVs · human · n = 10,047
  • UK BioBank resources: microarray-based genomic disorder carrier frequencies and GWAS results across thousands of phenotypes · human (UK BioBank) · n = 4,203 phenotypes (GWAS)
  • NHGRI-EBI GWAS Catalog associations matched to common SVs in strong LD (R^2 ≥ 0.8) with SNVs/indels · human · n = 15,634 common SVs in strong LD
  • Published WGS SV references used for comparison: 1000 Genomes Project (~7×), GoNL (~13×), GTEx (~50×) · human · n = 2,504 (1000 Genomes Project)
  • Noncoding regulatory element annotations ranging from validated enhancers to computationally predicted elements · human · n = 14 element classes
Methods & tools
Cloud-based multi-algorithm short-read WGS SV discovery and genotyping pipeline, Adjusted Proportion of Singletons (APS) selection metric, Watterson estimator (SV mutation rate), VaPoR long-read realignment for in silico SV confirmation, PacBio long-read WGS, gnomAD SNV/indel pLoF and missense constraint metrics (pLI), phastCons conservation scores, NHGRI-EBI GWAS Catalog, Chromosomal microarray (CMA) carrier frequencies from UK BioBank, NCBI dbVar, gnomAD browser
Data & accessions
https://gnomad.broadinstitute.org/
Reagents & environment
Rearranged sequence coordinates reported relative to the GRCh37 reference; no software versions or containers stated.
Reproducibility gaps
Detailed methods sit in a separate supplementary document not included here; no code repository URL is given for the SV pipeline or APS model.
Limitations (6)
  • Short-read WGS misses a disproportionate share of repeat-mediated SVs, so mutation rates and genomic distributions are underestimated; the 7,439 SVs per genome are below the 24,825 reported from ~40× long-read WGS. · stated by the authors
  • High copy-state MCNVs need specialized algorithms and manual curation, so the 1,055 multiallelic CNVs reported give an incomplete view of extreme copy-number polymorphism. · stated by the authors
  • Current sample sizes are too sparse for precise gene-level SV constraint, sequence-specific mutation rates, or detection of individual constrained noncoding elements; the expected-SV-per-gene model is acknowledged as imperfect. · stated by the authors
  • Whole-gene copy-gain duplications showed no excess selection, possibly from APS overcorrection for size; the Watterson-based estimate of 0.29 de novo SVs per generation (95% CI 0.13-0.44) is an approximation from unrelated individuals. · stated by the authors
  • Several figure-legend values differ from the main text (0.24% vs 0.18% pathogenic carriers, 9.5% vs 7.22% recessive developmental disorder carriers, ~3.1% vs 3.8% with rare ≥1Mb SVs, 40.4% vs 36.9% genes with a pLoF SV), so readers should verify which estimates are final. · noted in this summary
  • Samples were aggregated from many sequencing projects rather than drawn as a population sample, and Latino (N=1,232) and other/admixed (N=170) groups are small, so carrier rates and frequency thresholds may transfer poorly to under-sampled ancestries. · noted in this summary

Who should read:Read this if you interpret structural variants from short-read genomes in research or diagnostics, or study how selection acts on coding and noncoding dosage changes in humans.

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISHehir-Kwa '16Abel '18Sedlazeck '20Jun '23Yu '21Baker '26
69% semantic matchGenetics
Jayne Y. Hehir-Kwa, Tobias Marschall, Wigard P. Kloosterman+26 more·bioRxiv·Jul 2, 2016