henhaobio
Sign in

ScienceGenomicsPeer-reviewed journal article

The complete sequence of a human genome

Sergey NurkiD, Sergey KoreniD, Arang RhieiD, Mikko RautiaineniD, Andrey V. BzikadzeiD, Alla MikheenkoiD, Mitchell Robert VollgeriD, Nicolas AltemoseiD, Lev I. UralskyiD, Ariel GershmaniD, Sergey S. AganezoviD, Savannah J. HoytiD, Mark DiekhansiD, Glennis A. LogsdoniD, Michael AlongeiD, Stylianos E. AntonarakisiD, Matthew BorchersiD, Gerard G. BouffardiD, Shelise Y. BrooksiD, Gina Viviana CaldasiD, Nae-Chyun CheniD, Haoyu ChengiD, Chen-Shan ChiniD, William ChowiD, Leonardo Gomes de LimaiD, , Adam M. PhillippyiD

PDFDOIPreprintData
In-depth full-text digest available · Jump to digest ↓

Abstract

In 2001, Celera Genomics and the International Human Genome Sequencing Consortium published their initial drafts of the human genome, which revolutionized the field of genomics. While these drafts and the updates that followed effectively covered the euchromatic fraction of the genome, the heterochromatin and many other complex regions were left unfinished or erroneous. Addressing this remaining 8% of the genome, the Telomere-to-Telomere (T2T) Consortium has finished the first truly complete 3.055 billion base pair (bp) sequence of a human genome, representing the largest improvement to the human reference genome since its initial release. The new T2T-CHM13 reference includes gapless assemblies for all 22 autosomes plus Chromosome X, corrects numerous errors, and introduces nearly 200 million bp of novel sequence containing 2,226 paralogous gene copies, 115 of which are predicted to be protein coding. The newly completed regions include all centromeric satellite arrays and the short arms of all five acrocentric chromosomes, unlocking these complex regions of the genome to variational and functional studies for the first time.

In-depth summary

Claude Opus 5.5 · max effort · 8 Nov 2026 · from the full text of version 1 (27 May 2021) · CC BY 4.0

A gapless 3.055 billion bp human genome from HiFi and ultra-long nanopore reads resolves all centromeric arrays and acrocentric p-arms, adding 182 Mbp of new sequence.

GRCh38.p13 still contains 151 Mbp of unknown sequence, represents centromeric alpha satellites as computational models, leaves the five acrocentric short arms largely unassembled, and carries false duplications and collapsed repeats from its mosaic BAC-based construction. Pairing accurate HiFi with ultra-long nanopore reads on an effectively haploid cell line closes this remaining 8%, opening these regions to variant calling and functional analysis.

Key results

  1. ShownT2T-CHM13v1.1 gives gapless telomere-to-telomere sequence for all 22 autosomes and Chromosome X, adding or correcting 238 Mbp relative to GRCh38 and repairing falsely duplicated and collapsed GRCh38 regions.

    3,054,815,472 bp nuclear DNA plus a 16,569 bp mitochondrial genome; 238 Mbp is non-syntenic with GRCh38 over 1 Mbp intervals (mostly centromeric satellites 180 Mbp, segmental duplications 68 Mbp, rDNA 10 Mbp; categories overlap), and 182 Mbp is entirely new; removing 1.2 Mbp of false duplication plus other fixes affects at least 388 GRCh38 genes (68 protein coding). · A truly complete genome; Assembly validation and polishing; Abstract

  2. ShownConcordance of HiFi, ONT, Illumina, Strand-seq and Hi-C data with the assembly, plus BAC and ddPCR comparisons, supports high structural and base-level accuracy, with suspected problem regions covering a far smaller share of T2T-CHM13 than of GRCh38.

    Curation corrected 4 large and 993 small variants plus 454 telomeric fixes; 644 of 647 CHM13 BACs matched at >99.99% identity; 99.86% of the assembly lies within three SD of mean coverage (HiFi 34.70 ± 7.03, ONT 116.16 ± 16.96); estimated consensus accuracy Phred Q67 to Q73; flagged issues span 0.3% of length vs 8% for GRCh38. · Assembly validation and polishing

  3. SuggestedProjecting GENCODE v35 with CAT and Liftoff, supplemented by CHM13 Iso-Seq transcripts, yields a draft annotation in which thousands of putative paralogs are exclusive to CHM13, concentrated in pericentromeres and acrocentric p-arms, while few GRCh38 genes lack a CHM13 counterpart.

    63,494 genes and 233,615 transcripts (19,969 and 86,245 protein coding); 3,604 genes (6,693 transcripts) exclusive to CHM13 vs 263 GENCODE genes (448 transcripts) exclusive to GRCh38, at least 24 being presumed GRCh38 errors; 2,226 CHM13-only genes (115 protein coding) sit in sequence with no GRCh38 primary alignment. · A truly complete genome; Abstract

  4. ShownAll five acrocentric short arms are now resolved end to end, sharing a common layout (one rDNA array between distal and proximal satellite and duplication blocks) and high cross-chromosome similarity, with rDNA units more homogeneous within than between chromosomes.

    p-arms span 10.1 to 16.7 Mbp (66.1 Mbp total), each with one rDNA array from 0.7 Mbp (Chr14) to 3.6 Mbp (Chr13); 219 complete rDNA copies total 9.9 Mbp; 5 kbp p-arm windows align between chromosomes at median 98.7% identity; most 45S copies are 100.0% identical within a chromosome, while dominant morphs share 99.4 to 99.7% identity across chromosomes. · Acrocentric chromosomes; rDNA assembly

  5. ShownIn a case study, the assembly resolves FRG1 paralogs absent from GRCh38. On GRCh38, reads from these paralogs mis-map onto a single paralog and inflate its coverage and variant calls; mapping to T2T-CHM13 restores the expected coverage.

    T2T-CHM13 contains 23 FRG1 paralogs on chromosomes 9, 13, 14, 15, 20, 21 and 22 vs 9 in GRCh38; on GRCh38, HiFi reads from missing copies align to FRG1DP (97% identity to some new paralogs), giving excess coverage and non-reference variants, whereas CHM13 mapping gives the expected 34 to 40x coverage. · Analyses and resources; Figure legends

Methods, data and reproducibility

HiFi reads were homopolymer-compressed, error-corrected and assembled into an exact-overlap string graph whose tangles were resolved using coverage-based node multiplicity, manual curation and ONT read alignment; rDNA arrays were built via sparse de Bruijn graphs and ONT-defined morphs, then the assembly was polished and validated with multi-platform data.

  • CHM13hTERT complete hydatidiform mole cell line (46,XX), sequenced with PacBio HiFi, ONT ultra-long, Illumina PCR-free and Illumina/Arima Hi-C · human · n = 30× HiFi; 120× ONT; 100× Illumina; 70× Hi-C · NCBI BioProject PRJNA559484
  • BioNano optical maps and Strand-seq data from CHM13 (Strand-seq used for structural concordance checks) · human
  • Previously sequenced CHM13 BAC clones compared with the assembly as an accuracy check · human · n = 647 BACs
  • CHM13 Iso-Seq, RNA-Seq, PRO-Seq, CUT&RUN and ONT-derived CpG methylation data used for annotation and functional tracks · human
  • GRCh38.p13 reference assembly and GENCODE v35 annotation, serving as comparator and as source for annotation projection · human
  • rDNA reference unit originally from a human Chromosome 21 BAC clone, compared with CHM13 45S and intergenic spacer morphs · human · KY962518.1
  • HiFi reads from three additional human genomes (including HG02145), 1000 Genomes Illumina data and Simons Genome Diversity Panel k-mer copy numbers for FRG1 paralog analysis · human · n = three other human samples
  • HG002 benchmark cell line (46,XY), whose X chromosome was sequenced, assembled and validated for companion studies · human
Methods & tools
Custom HiFi string-graph assembler built from HiCanu and Miniasm components, GraphAligner, Sparse de Bruijn graph assembly for rDNA arrays, Winnowmap2 and BWA, Sniffles, DeepVariant and PEPPER-DeepVariant, Merfin, TandemTools, Comparative Annotation Toolkit (CAT), Liftoff, StringTie2, Droplet digital PCR (ddPCR) and FISH
Reagents & environment
CHM13hTERT complete hydatidiform mole line (hTERT-immortalized; 46,XX with 9qh+; low residual heterozygosity; hemizygous hTERT vector on Chr21, excluded from assembly); gene models projected from GENCODE v35; no software versions or containers stated.
Assay conditions & compute
Sequencing platforms: PacBio HiFi (30×), ONT ultra-long (120×), Illumina PCR-free (100×), Illumina/Arima Hi-C (70×), BioNano optical maps, Strand-seq; plus ddPCR and FISH; no computational hardware, runtime or memory reported.
Reproducibility gaps
The main text links no source code for the purpose-built HiFi string-graph or rDNA assembly workflows (details deferred to supplementary notes; many graph walks curated manually); the CHM13hTERT line was still being banked at Coriell.
Limitations (6)
  • rDNA arrays only approximate true copy number: interiors of the chromosome 13, 15 and 21 arrays were ordered artificially and are model sequences, and because ONT reads rarely spanned entire arrays, internal copy-number variation cannot be excluded. · stated by the authors
  • CHM13 lacks a Y chromosome, so the reference covers only 22 autosomes, Chromosome X and the mitochondrial genome; a complete Y is being assembled separately from the 46,XY HG002 benchmark line. · stated by the authors
  • Annotation of the newly completed regions is an initial draft (GENCODE v35 projected via CAT and Liftoff plus Iso-Seq transcripts) needing further validation; most new genes are putative paralogs whose protein-coding status is inferred from GENCODE relatives. · stated by the authors
  • Q67 to Q73 consensus accuracy is a genome-wide average: low-HiFi-coverage regions are enriched for possible consensus errors, and satellite-specific coverage anomalies (HiFi excess over HSat2/HSat3; HiFi and ONT depletion over HSat1) lack confirmed mechanisms. · stated by the authors
  • As a single, essentially haploid genome, CHM13 cannot reveal which acrocentric p-arm features are conserved across humans, and extending telomere-to-telomere assembly to heterozygous diploid genomes, with automation, remains a difficult open problem. · stated by the authors
  • As one individual's haplotype, the reference embeds idiosyncratic structures such as the multi-megabase 9qh+ HSat3 duplication, and heterozygous bubbles were resolved by keeping the longer allele; how such choices affect analyses of other genomes is not quantified here. · noted in this summary

Read this if you map reads, call variants, or study centromeres, rDNA, acrocentric short arms or segmental duplications and are deciding whether to move from GRCh38 to T2T-CHM13.

Citations
3,136
FWCI
218.36
Field percentile
top 1%

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
THISRhie '23Miga '20He '23Sarashetti '25Yang '22Aganezov '22
73% semantic matchGenomics
Arang Rhie, Sergey Nurk, Monika Čechová+84 more·Nature·Aug 23, 2023· 476 citations