henhao
Sign in
Genetic & Genomic MedicinemedRxiv
Preprint

FinnGen: Unique genetic insights from combining isolated population and national health register data

Mitja KurkiiD, Juha M KarjalaineniD, Priit PaltaiD, Timo Petteri SipilaiD, Kati KristianssoniD, Kati M. DonneriD, Mary Pat ReeveiD, Hannele M. LaivuoriiD, Mervi AavikkoiD, Mari Anneli KaunistoiD, Anu LoukolaiD, Elisa LahtelaiD, Hannele Mattsson, Päivi LaihoiD, Pietro Della Briotta Parolo, Arto LehistöiD, Masahiro KanaiiD, Nina MarsiD, Joel T RämöiD, Tuomo KiiskineniD, Henrike HeyneiD, Kumar VeerapeniD, Sina RüegeriD, Susanna M. LemmeläiD, Wei ZhouiD, , Marianna Niemi

Abstract

Population isolates such as Finland provide benefits in genetic studies because the allelic spectrum of damaging alleles in any gene is often concentrated on a small number of low-frequency variants (0.1% [≤] minor allele frequency < 5%), which survived the founding bottleneck, as opposed to being distributed over a much larger number of ultra--rare variants. While this advantage is well-- established in Mendelian genetics, its value in common disease genetics has been less explored. FinnGen aims to study the genome and national health register data of 500,000 Finns, already reaching 224,737 genotyped and phenotyped participants. Given the relatively high median age of participants (63 years) and dominance of hospital-based recruitment, FinnGen is enriched for many disease endpoints often underrepresented in population-based studies (e.g., rarer immune-mediated diseases and late onset degenerative and ophthalmologic endpoints). We report here a genome-wide association study (GWAS) of 1,932 clinical endpoints defined from nationwide health registries. We identify genome--wide significant associations at 2,491 independent loci. Among these, finemapping implicates 148 putatively causal coding variants associated with 202 endpoints, 104 with low allele frequency (AF<10%) of which 62 were over two-fold enriched in Finland. We studied a benchmark set of 15 diseases that had previously been investigated in large genome-wide association studies. FinnGen discovery analyses were meta-analysed in Estonian and UK biobanks. We identify 30 novel associations, primarily low-frequency variants strongly enriched, in or specific to, the Finnish population and Uralic language family neighbors in Estonia and Russia. These findings demonstrate the power of bottlenecked populations to find unique entry points into the biology of common diseases through low-frequency, high impact variants. Such high impact variants have a potential to contribute to medical translation including drug discovery.

PDFDOICodeData
Full-text digest ↓

02 /Specification & Classification

Click any tag to browse matching papers
Citations
404
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 1 (6 Mar 2022) · CC BY 4.0

Finnish-enriched low-frequency variants, often coding, underlie many GWAS signals across 1,932 register endpoints in 218,792 Finns; a 15-disease meta-analysis with Estonian and UK biobanks adds 30 unreported associations.

Bottlenecked populations such as Finland carry some damaging alleles at low frequencies that are rare or absent elsewhere, an advantage well known in Mendelian genetics but less tested for common disease. FinnGen pairs imputed genotypes with decades of nationwide health register data in an older, largely hospital-recruited cohort, enriching for endpoints that population-based studies underrepresent.

Key results

  1. ShownA phenome-wide GWAS of 1,932 register-defined endpoints found thousands of independent non-HLA signals, and fine-mapping placed a coding variant in the credible set of a minority of them.

    2,733 independent associations in 2,492 loci across 807 endpoints at p<5×10^-8; 893 signals (771 loci, 247 endpoints) passed Bonferroni correction for 1,932 endpoints (p<2.6×10^-11). A coding variant with posterior inclusion probability (PIP) >0.05 was in 369 associations (13.5%) spanning 202 endpoints; 73 of 493 stringent unique associations (14.8%) had one, 43 with PIP>0.5 and 28 with PIP>0.9. · Coding variant associations; Association analysis and fine-mapping

  2. ShownFinnish enrichment helps discovery: low-frequency association signals are more often Finnish-enriched than frequency-matched null variants, and enriched signals more often contain a coding variant.

    Of 169 lead variants with minor allele frequency (MAF) <10% among 493 Bonferroni-significant unique associations, 87 (51.5%) were over two-fold enriched vs non-Finnish-Swedish-Estonian Europeans (NFSEE) in gnomAD, vs 37% expected from 1,000,000 frequency-matched draws (p 2.8×10^-5). Coding variants sat in credible sets of 34/89 (38.2%) enriched vs 19/80 (23.8%) non-enriched low-frequency associations. · Coding variant associations; Estimation of expected number of enriched…

  3. ShownRegister-defined phenotypes for 15 well-studied diseases matched disease-specific GWAS, and replication meta-analysis with the Estonian Biobank and UK Biobank (UKBB) yielded 30 previously unreported associations, mostly Finnish-enriched low-frequency variants.

    No genetic correlation with published GWAS was significantly below 1 (lowest 0.89, SE 0.07, in age-related macular degeneration). FinnGen gave 275 non-HLA associations; meta-analysis with 136,724 Estonian Biobank participants and UKBB gave 298 genome-wide significant signals, of which 30 were absent from the GWAS Catalog and the largest published meta-analyses; 27 novel leads had MAF<10%. · Genome-wide association studies utilizing…; Estonian Biobank and UKBB replication; Automatic annotation of known GWAS hits

  4. ShownA single intronic TNRC18 variant gives a new inflammatory bowel disease (IBD) signal, replicated in Estonia, with opposite effects across other immune-mediated endpoints.

    Allele frequency (AF) 3.6%, odds ratio (OR) 3.2, p 2.4×10^-61, more significant than the IL23R, NOD2 and MHC signals; 114-fold enriched vs NFSEE (AF 0.04%). Estonian Biobank: AF 1.3%, OR 3.9, p 2.8×10^-6. Also raises risk of ankylosing spondylitis (OR 4.2, p 1.8×10^-34) and psoriasis (OR 1.6, p 1.1×10^-13) but is protective for a combined autoimmune endpoint (OR 0.84, p 6.2×10^-12). · Genome-wide association studies utilizing…

  5. SuggestedFine-mapped coding variants that are rare or absent in non-Finnish Europeans point to candidate genes for venous thromboembolism, sudden hearing loss and coronary revascularization, and prescription-purchase endpoints expose pleiotropy.

    STAB2 p.Ala1996Val with venous thromboembolism (AF 1.6%, OR 1.6, p 3.1×10^-14); TUBA1C p.Ala331Val with sudden idiopathic hearing loss (AF 0.2%, OR 35.2, p 1.4×10^-10); PLTP p.Val70Phe with coronary revascularization (OR 3.0, p 2.1×10^-9); TM6SF2 p.Leu156Pro with fewer statin purchases (OR 0.86, p 3.8×10^-13) and hepatocellular carcinoma (OR 3.7, p 5.9×10^-10). · New disease associations; Coding variants associated with medication use

Methods, data and reproducibility

Array genotypes (154,714 on the custom Axiom FinnGen1 array, 70,023 legacy) were imputed with SISu v3, a panel of 3,775 Finnish genomes sequenced at 25-30x (16,962,023 variants, most imputed at INFO>0.6). SAIGE mixed-model logistic GWAS of register-derived endpoints was followed by SuSiE fine-mapping, eQTL colocalization and inverse-variance meta-analysis with Estonian Biobank and UKBB.

  • FinnGen Release 5 biobank participants from nine Finnish biobanks, genotyped and linked to national health registers · human · n = 224,737 genotyped after QC; 218,792 with phenotype data used in analyses
  • Nationwide Finnish health registers (HILMO, AvoHILMO, Causes of Death, KELA medication purchases and reimbursements, Finnish Cancer Registry) used to build clinical endpoints · human · n = 1,932 endpoints analysed after clustering; 2803 endpoints in full public results
  • SISu v3 population-specific imputation reference panel, high-coverage (25-30x) whole-genome sequencing · human (Finnish individuals) · n = 3,775
  • Estonian Biobank replication cohort, association analysis of the 15 benchmark phenotypes · human · n = 136,724
  • Pan UKBB European-subset association summary statistics used for replication · human (UK Biobank)
  • 1000 Genomes Project phase 3 reference samples for ancestry principal component analysis · human (1000 Genomes) · n = 2504
  • gnomAD v2.0.1 allele frequencies, with non-Finnish-Swedish-Estonian Europeans (NFSEE) as reference · human (gnomAD)
  • Fine-mapped eQTL data from GTEx and the eQTL Catalogue, plus 36 continuous endpoints and 57 biomarkers from UKBB, for colocalization · human (GTEx, UK Biobank)
Methods & tools
Axiom FinnGen1 array, GenCall, zCall and AxiomGT1 genotype calling, Eagle 2.3.5, Beagle 4.1, GATK HaplotypeCaller, Hail, SAIGE 0.35.8.8, SuSiE, LDStore2, eCAVIAR colocalization model (CLPP), Variant Effect Predictor (VEP), GWAS Catalog, ClinVar, gnomAD
Code
https://github.com/FINNGEN/finemapping-pipelinehttps://github.com/FINNGEN/META_ANALYSIShttps://github.com/FINNGEN/pca_kinship
Data & accessions
https://www.finngen.fi/en/access_resultshttps://r5.finngen.fi/https://r5.risteys.finngen.fi/https://www.finngen.fi/en/researchers/clinical-endpointshttps://pan.ukbb.broadinstitute.org/
Software & environment
SAIGE 0.35.8.8; Eagle 2.3.5; Beagle 4.1 (08Jun17.d8b); Hail v0.1; GRCh38 coordinates; GWAS Catalog release of 21 Apr 2021 with dbSNP build 153; liftover and imputation protocols at dx.doi.org/10.17504/protocols.io.nqtddwn and dx.doi.org/10.17504/protocols.io.nmndc5e.
Reproducibility gaps
Public release covers summary statistics (new releases after a one-year embargo); the text gives no access route for individual-level genotypes or register data. No container or dependency files are described, and endpoint and QC details are in supplementary material not included here.
Limitations (6)
  • Recruitment is mostly hospital-based plus legacy disease cohorts, so the authors state FinnGen is not epidemiologically representative of the Finnish population. · stated by the authors
  • The authors note fine-mapping may sometimes misassign causality to a coding variant, and defer calibration and replicability to companion papers. · stated by the authors
  • eQTL colocalization was found for only 5 of 27 novel loci, which the authors attribute to 25 of the 27 variants being rare in eQTL resources such as GTEx, whose samples are mostly not of Finnish or Estonian ancestry. · stated by the authors
  • Some Finnish-enriched hits could not be replicated because the variants are rare or absent elsewhere; the rare prostate cancer variant near SNORA40/STAT3 still needs an independent cohort. · stated by the authors
  • Most phenome-wide coding-variant associations outside the 15 benchmark diseases (e.g., TUBA1C, STAB2, PLTP, TM6SF2) come from FinnGen alone, with no external replication reported in this text. · noted in this summary
  • Some counts disagree across sections (2,491 vs 2,492 loci; 87 vs 89 enriched low-frequency leads; a variant total printed as 116 387 711), so figures should be checked against the supplementary tables before reuse. · noted in this summary

Who should read:Read this if you use FinnGen summary statistics or want evidence that a founder population with register-based phenotypes can surface low-frequency, high-effect coding variants for common diseases.

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
THISChheda '16Locke '19Heyne '21Yin '22Kanai '21Martin '18
73% semantic matchGenetics
Himanshu Chheda, Priit Palta, Matti Pirinen+9 more·bioRxiv·Jul 12, 2016