henhaobio
Sign in

NatureGenomicsPeer-reviewed journal article

Complex genetic variation in nearly complete human genomes

Glennis A. LogsdoniD, Peter EbertiD, Peter A. AudanoiD, Mark G. LoftusiD, David PorubskỳiD, Jana EbleriD, Feyza YilmaziD, Pille HallastiD, Timofey ProdanoviD, DongAhn YooiD, Carolyn A. PaisieiD, William T. HarveyiD, Xuefang ZhaoiD, Gianni V. MartinoiD, Mir HengliniD, Katherine M. MunsoniD, Keon RabbaniiD, Chen-Shan ChiniD, Bida GuiD, Hufsah AshrafiD, Stephan Scholz, Olanrewaju Austine-OrimoloyeiD, Parithi BalachandraniD, Marc Jan BonderiD, Haoyu ChengiD, , Tobias MarschalliD

PDFDOIPreprintCodeData

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 1 (25 Sept 2024) · CC BY 4.0

Hybrid HiFi/ultra-long ONT assemblies of 65 diverse humans yield 130 haplotypes, 1,246 complete centromeres, and a pangenome panel enabling short-read genome inference at median QV 45.

Long-read assemblies built from one sequencing technology still break at difficult repeats, and gapless hybrid HiFi plus ultra-long ONT assemblies existed for only a few individuals, limiting pangenome references and short-read access to complex structural variants (SVs). Assembling 65 diverse genomes this way closes 92% of earlier HiFi-only gaps, resolves centromeres, MHC, SMN1/SMN2 and amylase haplotypes, and expands the SVs genotypable in 3,202 short-read genomes.

Key results

  1. ShownVerkko hybrid assemblies of all 65 samples, phased with Strand-seq, are highly contiguous and accurate at the base level. They close most gaps left by HiFi-only assemblies and reach telomere-to-telomere (T2T) status for a large share of chromosomes.

    130 haplotypes; median auN 137 Mbp; median QV between 54 and 61; 99% median single-copy gene completeness; 92% of prior HiFi-only gaps closed; 39% of chromosomes T2T; 599 gapless T2T contigs (median 10 per genome) plus 558 single-scaffold chromosomes; MHC resolved in 128/130 haplotypes; 99.8% median trio parental support. · Abstract; Assembly

  2. ShownA Minigraph-Cactus graph of these 65 genomes plus 42 HPRC genomes lets PanGenie genotype many more SVs, especially rare ones, across 3,202 1kGP short-read genomes. Phasing the genotypes rebuilds personal haplotypes more accurately than an existing short-read panel.

    Mean 26,115 SVs per 1kGP sample vs 18,462 (HPRC genotypes) and 9,596 (1kGP high-coverage short-read calls, 1kGP-HC); rare SVs per African sample 1,490 vs 382 and 477; leave-one-out (LOO) concordance up to ∼94% for biallelic SVs; consensus haplotypes median QV 45, or 48 in CHM13-syntenic regions vs 43 for 1kGP-HC; median k-mer completeness 97.4% vs 97.1%. · Genome-wide genotyping with PanGenie; Personal genome reconstruction from an…

  3. ShownPooling Verkko and hifiasm assemblies gives 1,246 complete centromeres validated with reads. They show wide variation in α-satellite higher-order repeat (HOR) array length, new HOR structures, recurrent mobile element insertions (MEIs), and paired hypomethylated centromere dip regions (CDRs).

    822 Verkko and 777 hifiasm centromeres, 28.3% correct in both; mean HOR array 2.31 Mbp, 37.5-fold range on chromosome 10, 5.9 Mbp absolute range on chromosome 18; 4,153 new HOR variants or organizations; two CDRs >80 kbp apart on ∼7% of chromosomes; ∼30% of active arrays carry an MEI (89 unique, 58% L1). · Centromeres; Discussion

  4. ShownAssembly-based PAV calling against T2T-CHM13 gives a larger catalog with fewer Mendelian errors than the earlier 32-assembly resource. It also resolves complex SVs (more than two breakpoints) whose alignments look like simple events.

    vs T2T-CHM13: 188,500 SVs, 6.3 million indels, 23.9 million SNVs; 1.6× more SV insertions/deletions (177,718 vs 111,679), 3.5× for >10 kbp (4,541 vs 1,299); SV Mendelian error 2.7% (55% decrease); 1,909 complex SVs, mean 55 per sample, 60% leaving deletion-like alignments; no other caller found the NBPF8 structures. · Variant calling; Complex structural polymorphisms

  5. ShownAdding these haplotypes to the reference panel improves targeted Locityper genotyping of 347 complex loci, including MHC genes. The assemblies also fully resolve SMN1/SMN2, amylase and HLA haplotype structures.

    LOO haplotypes at QV ≥30: 80.0% vs 74.6% (HPRC-only panel); MHC gene alleles correct in 81.0% (HPRC-only), 86.3% (LOO), 97.1% (full panel); 101 SMN haplotypes resolved, pooled from this study's and previously published assemblies, with SMN1/2 copy number matching Parascopy and SMNCopyNumberCaller for 31 samples; 35 amylase haplotypes; 99.2% (357/360) HLA alleles match classical typing. · Targeted genotyping of complex polymorphic loci; Major Histocompatibility Complex; Complex structural polymorphisms; Discussion

Methods, data and reproducibility

For 65 samples, mostly from the 1000 Genomes Project (1kGP), the authors generated PacBio HiFi and ultra-long ONT reads plus Strand-seq, Hi-C, optical maps, Iso-Seq and RNA-seq. Verkko assemblies phased with Strand-seq fed PAV-led variant calling and a Minigraph-Cactus graph with 42 HPRC samples; the graph was then used for PanGenie and Locityper genotyping of 3,202 short-read genomes.

  • PacBio HiFi (∼47-fold) and ONT (∼56-fold, ∼36-fold ultra-long) reads from Coriell lymphoblast lines: 63 1kGP samples plus NA21487 (HapMap) and HG002 (GIAB) · human · n = 65 samples (30 male, 35 female); 130 haplotype assemblies · PRJEB58376, PRJEB75216, PRJEB77558, PRJEB75190, PRJNA698480 and others (Data availability)
  • Single-cell template strand sequencing (Strand-seq) used for assembly-graph phasing and inversion re-genotyping · human · PRJEB39750, PRJEB12849
  • Bionano Genomics DLE-1 optical maps (Saphyr), also used to support amylase haplotypes · human · PRJNA339722, PRJEB41077, PRJEB58376, PRJEB77842
  • Hi-C (Proximo kit v4.0, NovaSeq 6000) used to test SV-associated chromatin contact changes · human · PRJEB39684, PRJEB75193, PRJEB58376
  • PacBio Iso-Seq and Illumina RNA-seq of EBV-transformed B-lymphocyte lines for isoform and differential expression analysis · human · n = 12 samples · Iso-Seq PRJEB75191; RNA-seq PRJEB75192, PRJEB58376
  • Human Pangenome Reference Consortium (HPRC) haplotype-resolved assemblies added to the Minigraph-Cactus graph and Locityper panels · human · n = 42 samples (graph); 90 haplotypes (HPRC-only Locityper panel)
  • 1kGP Illumina short-read WGS genotyped with PanGenie and Locityper, plus an external rare SNP/indel callset (70,174,243 variants) · human · n = 3,202 samples (602 trios)
  • GWAS Catalog associations (gwas_catalog_v1.0.2, e111_r2024-04-16) tested for linkage disequilibrium with SVs · human · n = 149,081 autosomal SNPs
Methods & tools
Verkko v1.4.1 (primary hybrid assembler), hifiasm (UL) v0.19.6 (centromeres, Yq12), Graphasing v0.3.1-alpha (Strand-seq phasing), Flagger, NucFreq, Merqury, Inspector (assembly QC), PAV v2.4.1 with SV-Pop; complex-SV dev version (10.5281/zenodo.13800981), Minigraph-Cactus v2.7.2, PanGenie v3.1.0, SHAPEIT5 v5.1.1, Locityper v0.15.1, ArbiGent and NAHRwhals (inversion validation), L1ME-AID, PALMER2, MELT-LRA (mobile element insertion calling), CDR-Finder, Hum-AS-HMMER, StainedGlass (centromere analysis)
Software & environment
Verkko v1.4.1, hifiasm (UL) v0.19.6, Graphasing v0.3.1-alpha, PAV v2.4.1 (v2.4.0.1 in Results), Minigraph-Cactus v2.7.2, PanGenie v3.1.0, SHAPEIT5 v5.1.1, Locityper v0.15.1, minimap2 v2.26; PAV also distributed as Docker and Singularity containers.
Reproducibility gaps
Assembly and evaluation workflows sit on prototype branches, and complex-SV calling used a PAV development version. The Globus working-directory URL is malformed, and the long-read accession 'RJEB75739' looks truncated.
Limitations (6)
  • Acrocentric short arms are still unassembled. About 25.6 Mbp of segmental duplications (SDs) per haploid genome remain unresolved relative to T2T-CHM13; 21.2 Mbp of this lies on the p-arms of chromosomes 13, 14, 15, 21 and 22. · stated by the authors
  • Centromeres and Yq12 stayed hard for Verkko to assemble and evaluate. hifiasm (UL) showed lower contiguity and spurious duplications, so its output was used only for those regions after manual curation. · stated by the authors
  • Putative di-kinetochores (two CDRs >80 kbp apart on ∼7% of chromosomes) and the link between CDRs and homogenizing HORs rest on methylation and sequence identity. CENP-A localization and further analyses are needed to confirm them, and effects on chromosome segregation remain untested. · stated by the authors
  • Gains in short-read genotyping cannot be traced to one cause: assembly quality, panel size, newer Minigraph-Cactus and PanGenie versions, and the switch to T2T-CHM13 all changed together. Locityper accuracy is also bounded by the haplotypes present in the panel. · stated by the authors
  • SV effects on isoforms, expression and chromatin come from only 12 EBV-transformed B-cell lines. The 122 SV-expression associations and 29 Hi-C contact changes are therefore correlative and limited to one cell type. · noted in this summary
  • Manual inspection estimated a 12.33% false discovery rate for inversion calls, and Strand-seq re-genotyping could assess only 252 of the 276 T2T-CHM13 inversions, so individual inversion calls warrant caution. · noted in this summary

Read this if you build pangenome references, genotype structural variants from short reads, or study centromeres, MHC, SMN1/SMN2 or other complex human loci using phased near-T2T assemblies.

Citations
138
FWCI
78.95
Field percentile
top 1%

Abstract

Diverse sets of complete human genomes are required to construct a pangenome reference and to understand the extent of complex structural variation. Here, we sequence 65 diverse human genomes and build 130 haplotype-resolved assemblies (130 Mbp median continuity), closing 92% of all previous assembly gaps1,2 and reaching telomere-to-telomere (T2T) status for 39% of the chromosomes. We highlight complete sequence continuity of complex loci, including the major histocompatibility complex (MHC), SMN1/SMN2, NBPF8, and AMY1/AMY2, and fully resolve 1,852 complex structural variants (SVs). In addition, we completely assemble and validate 1,246 human centromeres. We find up to 30-fold variation in -satellite high-order repeat (HOR) array length and characterize the pattern of mobile element insertions into -satellite HOR arrays. While most centromeres predict a single site of kinetochore attachment, epigenetic analysis suggests the presence of two hypomethylated regions for 7% of centromeres. Combining our data with the draft pangenome reference1 significantly enhances genotyping accuracy from short-read data, enabling whole-genome inference3 to a median quality value (QV) of 45. Using this approach, 26,115 SVs per sample are detected, substantially increasing the number of SVs now amenable to downstream disease association studies.

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
THISEbert '21Gao '24Sun '26Eizenga '26Logsdon '24Wu '26
70% semantic matchGenetics
Peter Ebert, Peter A. Audano, Qihui Zhu+62 more·Science·Feb 25, 2021· 701 citations