henhao
Sign in
GenomicsbioRxiv
Preprint

Whole exome sequencing and characterization of coding variation in 49,960 individuals in the UK Biobank

Cristopher V. Van HoutiD, Ioanna TachmazidouiD, Joshua BackmaniD, Joshua X. Hoffman, Bin YeiD, Ashutosh K. Pandey, Claudia G. Gonzaga-JaureguiiD, Shareef Khalid, Daren LiuiD, Nilanjana BanerjeeiD, Alexander H. LiiD, O’Dushlaine Colm, Anthony MarckettaiD, Jeffrey C. StaplesiD, Claudia SchurmanniD, Alicia C. Hawes, Evan K. MaxwelliD, Leland Barnard, Alexander E. LopeziD, John Penn, Lukas Habegger, Andrew L. BlumenfeldiD, Ashish Kumar YadaviD, Kavita Praveen, Marcus B. JonesiD, , Aris BarasiD

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

Abstract

The UK Biobank is a prospective study of 502,543 individuals, combining extensive phenotypic and genotypic data with streamlined access for researchers around the world. Here we describe the first tranche of large-scale exome sequence data for 49,960 study participants, revealing approximately 4 million coding variants (of which ~98.4% have frequency 10-fold increase compared to imputed sequence for the same participants. Nearly all genes (>97%) had [≥]1 predicted loss of function carrier, and most genes (>69%) had [≥]10 loss of function carriers. We illustrate the power of characterizing loss of function variation in this large population through association analyses across 1,741 phenotypes. In addition to replicating a range of established associations, we discover novel loss of function variants with large effects on disease traits, including PIEZO1 on varicose veins, COL6A1 on corneal resistance, MEPE on bone density, and IQGAP2 and GMPR on blood cell traits. We further demonstrate the value of exome sequencing by surveying the prevalence of pathogenic variants of clinical significance in this population, finding that 2% of the population has a medically actionable variant. Additionally, we leverage the phenotypic data to characterize the relationship between rare BRCA1 and BRCA2 pathogenic variants and cancer risk. Exomes from the first 49,960 participants are now made accessible to the scientific community and highlight the promise offered by genomic sequencing in large-scale population-based studies.

02 /Specification & Classification

Click any tag to browse matching papers
Citations
110
OpenAlex Topics

In-depth summary

✦ Multi-agent verified full-text synthesis

Claude Opus 5.5 · max effort · 8 Nov 2026 · from the full text of version 1 (9 Mar 2019) · CC BY-NC-ND 4.0

Exomes of 49,960 UK Biobank participants reveal 231,631 predicted loss-of-function variants (>10-fold more than imputed data); burden tests implicate PIEZO1, MEPE, and COL6A1 in disease traits.

Array genotyping with imputation captures common variation in UK Biobank but misses much of the rare, protein-altering variation most likely to have large phenotypic effects. Sequencing exomes of 49,960 participants adds directly assayed coding variants, including >10-fold more loss-of-function (LOF) variants than imputed sequence, linked to extensive health records for gene discovery and clinical variant surveys.

Key results

  1. ShownWhole exome sequencing (WES) captures far more rare coding and loss-of-function variation than imputed sequence in the same people, and already finds heterozygous LOF carriers in nearly all genes.

    WES vs imputed in the same 49,797 people: 3,995,794 vs 707,124 coding and 235,915 vs 19,451 LOF variants; only 13.7% of WES coding variants and 5.3% of WES LOFs appeared in imputed data. Genes with heterozygous LOF carriers: 17,751 (WES, 46,827 Europeans) vs 8,724 (imputed info>0.3, 462,427 Europeans); >97% of autosomal genes had at least one carrier. · Abstract; Enhancement of Coding Variation from WES…; Comparison of LOF Variants Ascertained Through…; DISCUSSION

  2. ShownRare LOF burden testing identifies five previously unreported gene-trait associations of large effect: PIEZO1 with varicose veins, MEPE with lower bone density, COL6A1 with reduced corneal resistance, and IQGAP2 and GMPR with blood cell traits.

    PIEZO1, varicose veins: OR=4.8, P=2.7×10^−8; DiscovEHR replication OR=3.8, p=1.5×10^−6. MEPE, heel bone mineral density: −0.50 SD, p=1.4×10^−8; its frameshift rs753138805 gave −0.5 SD (P=2.1×10^−18) and fracture OR=1.4 in HUNT. COL6A1, corneal resistance factor: −1.5 SD, p=3.6×10^−10. IQGAP2 and GMPR replicated in DiscovEHR. · LOF Associations and Novel Gene Discovery

  3. ShownAcross 1,741 traits, LOF gene burden tests recover established disease genes, and most signals exceed both single-variant tests and burden tests using imputed data from all 500k participants.

    In 46,979 European-ancestry people, 25 gene-trait associations reached p<10^−7, 21 stronger than any single LOF; e.g., MLH1-digestive cancer OR=84, PKD1-chronic kidney disease OR=91, TTN-cardiomyopathy OR=11.9. For 19 of 25, WES in 50k was more significant than imputed data in all 500k; equal-size imputed data found none of the five new hits at p<10^−7. · Phenotypic Associations with LOF Variation; LOF Associations and Novel Gene Discovery

  4. ShownAbout 2% of this population-based sample carries a medically actionable pathogenic or likely pathogenic variant in 59 American College of Medical Genetics (ACMG59) genes; BRCA1/2 carriers have raised risk of the five BRCA-linked cancers but not of other cancers combined.

    555 variants (316 reported pathogenic, 239 likely pathogenic) in 1,000 individuals (2.0%) vs 2.76% of 91,514 in DiscovEHR. Pooled BRCA1/2 carriers (114 males, 110 females) vs non-carriers: any of 5 linked cancers 21.0% vs 6.6%, OR = 3.75 (95%CI 2.71,5.18), p = 2×10^−12; other cancers OR = 0.86, p = 0.55; hazard ratio 4.3 for female breast/ovarian, 3.68 for prostate. · Survey of Medically Actionable Pathogenic…

  5. ShownImputed genotypes agree poorly with sequencing at low allele frequencies, and most sampled imputed-only LOF calls in well-covered exome regions lack read support, pointing to imputation error rather than WES misses.

    Per-variant R-squared vs WES: imputed 35% at minor allele frequency (MAF) <0.01% to 96% at MAF >1% (average 54%); array 70% to 99% (average 99%). Of 363 manually reviewed imputed-only LOF sample-sites in consistently covered regions, 76% had no read evidence, ~21% other variation, ~3% clear LOF support. · Enhancement of Coding Variation from WES…; Concordance of WES, Directly Genotyped Array…

Methods, data and reproducibility

Regeneron Genetics Center sequenced exomes captured with a modified IDT xGen v1.0 panel; predicted LOF variants were annotated and compared with array and imputed genotypes, ClinVar/ACMG59 pathogenic variants were surveyed, and rare (MAF<1%) LOF gene burden tests were run across 1,741 traits in European-ancestry participants.

  • UK Biobank exome-sequenced participants (first tranche), selected for richer phenotyping (MRI, enhanced measures, hospital and primary care records) with asthma enrichment · human · n = 49,960 (50,000 selected; 40 excluded)
  • Participants with both WES and imputed sequence, used to compare coding and LOF variant yield · human · n = 49,797
  • UK Biobank array and imputed genotypes compared per variant with WES for concordance (75,334 array and 899,455 imputed variants) · human · n = 46,912 (array) and 46,860 (imputed) individuals
  • European-ancestry WES participants used for LOF burden tests on 1,741 traits (1,073 discrete, 668 quantitative) · human · n = 46,979
  • Full UK Biobank cohort for representativeness comparison, plus imputed sequence in European-ancestry participants for LOF gene counts and burden power · human · n = 502,543 total; 462,427 European-ancestry with imputed data
  • gnomAD exomes, Non-Finnish European subset, re-annotated with the same pipeline to compare numbers of genes carrying LOFs · human · n = 56,885 Non-Finnish European individuals
  • Geisinger-Regeneron DiscovEHR exome cohort, used for ACMG59 prevalence comparison and replication of burden associations · human · n = 91,514 (ACMG59); 1,572 cases/75,704 controls (varicose veins)
  • HUNT study queried for MEPE frameshift rs753138805 against bone mineral density and fracture · human · n = 24,155 fracture cases
Methods & tools
IDT xGen Exome Research Panel v1.0 (modified with supplemental probes), ENSEMBL 85 transcript models for LOF annotation, LOFTEE (high-confidence LOF filtering), CADD deleteriousness scores, Integrative Genomics Viewer (IGV) for manual read review, ClinVar (≥2-star, non-conflicting pathogenic assertions), ACMG SF2.0 59-gene medically actionable list, Gene-level LOF burden tests with leave-one-out and step-wise regression follow-up, Fisher's exact test, Goldilocks (GL) genotype QC filtering, Identity-by-descent relatedness estimation, Genetic principal components for ancestry assignment
Data & accessions
http://ukbiobank.ac.uk/
Reagents & environment
Transcript models from ENSEMBL 85; actionable gene list ACMG SF2.0; no aligner, variant caller, or statistical software versions or containers are stated.
Reproducibility gaps
Exome data are available only to approved researchers via UK Biobank access protocols; no analysis code is shared, and QC, concordance, GL filtering, and ACMG59 variant selection details are deferred to Supplemental Methods.
Limitations (6)
  • Limited ancestral diversity in the cohort constrains its usefulness for judging variant pathogenicity and for characterizing the breadth of disease risk and phenotypic expression. · stated by the authors
  • Targeted capture misses variation outside the design (1,730 imputed-only LOFs, 24.8%, lay in untargeted regions), and LOF calling favors sensitivity: restricting to LOFs affecting all ENSEMBL 85 transcripts cuts the count to 153,903. · stated by the authors
  • Rare-variant burden power was limited for most binary traits at this sample size, and COL6A1 corneal signals could not be replicated because DiscovEHR lacked corneal hysteresis and resistance measures. · stated by the authors
  • Few carriers forced pooling of BRCA1 and BRCA2 pathogenic and likely pathogenic carriers (114 males, 110 females), so BRCA1- versus BRCA2-specific cancer risks were not estimated. · stated by the authors
  • BRCA1/2 cancer odds ratios come from unadjusted Fisher's exact tests on prevalent cases partly defined by self-report, leaving age, family history, and participant selection effects unaddressed. · noted in this summary
  • The MEPE burden signal weakened to p~2×10^−4 after conditioning on six known GWAS variants and was nonsignificant in DiscovEHR (B = −0.19, P = 0.22), so its independence from common-variant signals is only partly established. · noted in this summary

Who should read:Read this if you analyze rare coding or loss-of-function variants in UK Biobank, weigh exome sequencing against imputation, or need population estimates of ACMG59 actionable variants and BRCA1/2 cancer risk.

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
THISJurgens '22Wang '20Szustakowsk '21Karczewski '22Cirulli '20Abul‐Husn '19
74% semantic matchGenetics
Sean Joseph Jurgens, Seung Hoan Choi, Valerie N. Morrill+67 more·Nature Genetics·Feb 17, 2022· 171 citations