henhao
Sign in
GeneticsThe Lancet Neurology
Peer-reviewed journal article

Expanding Parkinson’s disease genetics: novel risk loci, genomic context, causal insights and heritable risk

Mike A. NallsiD, Cornelis BlauwendraatiD, Costanza Ludovica VallergaiD, Karl HeilbroniD, Sara Bandrés‐CigaiD, Diana ChangiD, Manuela M. X. TaniD, Demis A. Kia, Alastair John NoyceiD, Angli XueiD, José BrásiD, Emily YoungiD, Rainer von CoellniD, Javier Simón-Sánchez, Claudia SchulteiD, Manu SharmaiD, Lynne KrohniD, Lasse PihlstrømiD, Ari SiitoneniD, Hirotaka IwakiiD, Hampton L. LeonardiD, Faraz FaghriiD, Jesse Raphael GibbsiD, Dena G. HernandeziD, Sonja W. ScholziD, , Ciara García

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

Abstract

We performed the largest genome-wide association study of PD to date, involving the analysis of 7.8M SNPs in 37.7K cases, 18.6K UK Biobank proxy-cases, and 1.4M controls. We identified 90 independent genome-wide significant signals across 78 loci, including 38 independent risk signals in 37 novel loci. These variants explained 26-36% of the heritable risk of PD. Tests of causality within a Mendelian randomization framework identified putatively causal genes for 70 risk signals. Tissue expression enrichment analysis suggested that signatures of PD loci were heavily brain-enriched, consistent with specific neuronal cell types being implicated from single cell expression data. We found significant genetic correlations with brain volumes, smoking status, and educational attainment. In sum, these data provide the most comprehensive understanding of the genetic architecture of PD to date by revealing many additional PD risk loci, providing a biological context for these risk factors, and demonstrating that a considerable genetic component of this disease remains unidentified.

02 /Specification & Classification

Click any tag to browse matching papers
Citations
2,451
FWCI
30.82
Field percentile
top 1%
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 3 (4 Mar 2019) · CC0 1.0

Across 37,688 Parkinson's cases, 18,618 proxy-cases and 1.4M controls, 90 risk signals map to 78 loci (37 new); an 1805-variant polygenic score explains 26-36% of heritability.

Parkinson's disease (PD) has no therapy that delays or alters its course, and while early genetics focused on rare familial mutations, sporadic PD also carries a substantial genetic component. Relative to earlier genome-wide association studies (GWAS), this work nearly doubles known risk variants, nominates putatively causal genes via QTL-based Mendelian randomization, and estimates that roughly 99 K total cases would be needed to detect typical sub-significant variants.

Key results

  1. ShownSingle-stage meta-analysis yielded 90 independent genome-wide significant signals at 78 loci, including 38 new signals in 37 previously unreported loci; 10 loci harbor multiple independent signals, one of them new (UBTF/GRN plus FAM171A2).

    7,784,415 SNPs in 37,688 cases, 18,618 UK Biobank (UKB) proxy-cases and 1,417,791 controls; new signals lie >1MB from Chang et al. 2017 loci; multi-signal loci hold 22 risk SNPs (e.g., GBA, SNCA, LRRK2; new: rs2269906, rs850738); clinician-ascertained cases show genetic correlation (rG) 0.85 (SE = 0.06) with 23andMe self-report and 0.84 (SE = 0.134) with UKB proxy-cases. · Abstract; Novel loci and multiple signals in known loci…

  2. ShownPolygenic risk scores (PRS) discriminate PD modestly: an 1805-variant PRS reaches a higher validation AUC than one built from 88 genome-wide significant variants and explains about 26% of heritability at 0.5% prevalence (36% at 2%), versus about 16% for the 88-variant score.

    Harvard Biomarker Study (HBS; 527 cases, 472 controls): AUC 0.692 (95% CI 0.660-0.725) with 1805 variants vs 0.651 (0.617-0.684) with 88; 88-variant PRS beat Chang et al. 2017 PRS (AUC 0.624, DeLong P < 0.002); top vs bottom quartile odds ratio 3.74 (3.35-4.18) in NeuroX-dbGaP, 6.25 (4.26-9.28) in HBS; LD score regression liability-scale h2 0.22 (95% CI 0.18-0.26). · Refining heritability estimates and determining…; DISCUSSION

  3. SuggestedSummary-data Mendelian randomization (SMR) across four brain and blood QTL resources nominated putatively causal genes for most risk signals, often the gene nearest the lead SNP; rare coding variant burden tests independently supported seven genes, including LRRK2 and NFKB2.

    Of 305 genes in the 78 loci, 237 were testable and 151 (63.7%) showed significant expression or methylation effects on PD risk; 70 of 90 signals lie in loci with such a gene, the nearest gene qualifying for 53 of 70 (75.7%); of 113 burden-tested genes, 7 passed Bonferroni, with LRRK2 (P = 2.15E-10) and NFKB2 (P = 4.02E-07) at MAF < 1% surviving ~20,000-gene correction. · Functional causal inferences via Quantitative…; Rare coding variant burden analysis

  4. ShownGenes at PD loci are enriched for expression only in brain tissues and neuronal cell types, strongest in substantia nigra neurons, and for vacuolar/lysosomal pathways and protein-interaction networks tied to chemical signaling and stressor responses, led by interferon-gamma response.

    13 of 53 tissues enriched, all brain; seven of 88 DropViz brain cell types survive false discovery rate (FDR) correction, all neuronal (substantia nigra neurons P = 1.0E-06; others P < 5.0E-4); 10 of 10,651 pathways at FDR P < 0.05 (four vacuolar), lysosomal and dopamine pathways at FDR P < 0.1; six protein-interaction networks at FDR P < 0.1, led by interferon-gamma response. · Tissue and cell specific expression enrichment…; DISCUSSION

  5. SuggestedPD genetically correlates positively with intracranial and putamen volume and negatively with current smoking and vocational qualifications; Mendelian randomization (MR) supports causal effects of cognitive performance and educational attainment on PD, with no causal link to current smoking.

    Of 757 LD Hub traits, rG with intracranial volume 0.351 (P = 4.64E-06), putamen volume 0.248 (P = 9.55E-05), current smoking −0.134 (P = 7.92E-05), NVQ/HND/HNC qualifications −0.169 (P = 2.00E-04); MR effect of cognitive performance on PD 0.213 (SE 0.041, Bonferroni P = 8.00E-07), educational attainment 0.162 (P = 2.06E-04); current smoking to PD −0.069 (P = 0.125). · Genetic correlations and Mendelian randomization…; DISCUSSION

Methods, data and reproducibility

Single-stage GWAS meta-analysis with GCTA-COJO and 23andMe conditional analyses; LD score regression heritability; two-stage polygenic risk scoring; SMR using brain and blood expression and methylation QTLs; rare-variant burden tests; FUMA/webgestaltR enrichment; and cross-trait genetic correlation plus bi-directional Mendelian randomization.

  • European-ancestry PD GWAS meta-analysis pooling clinically ascertained, 23andMe self-reported and UK Biobank proxy-case series over 7,784,415 SNPs · human · n = 37,688 cases; 18,618 proxy-cases; 1,417,791 controls
  • The 11 clinically ascertained GWAS datasets, meta-analyzed for LD score regression heritability · human · n = 11 datasets
  • NeuroX-dbGaP cohort (NeuroX array) used for PRS variant selection, permutation-based P-threshold tuning and training · human · n = 5,851 cases; 5,866 controls
  • Harvard Biomarker Study (HBS) cohort (NeuroX array) used to validate the PRS · human · n = 527 cases; 472 controls
  • Four QTL resources for SMR: brain mRNA expression meta-analysis, substantia nigra expression, blood expression and blood methylation · human
  • DropViz single-cell RNA-seq reference spanning 88 brain cell types for cell-type expression enrichment
  • GWAS summary statistics curated by LD Hub for cross-trait genetic correlation · human · n = 757 GWAS datasets
  • Rare coding variant data from a subset of the GWAS studies, used for kernel-based burden tests of 113 genes in PD loci · human
Methods & tools
Fixed- and random-effects GWAS meta-analysis, GCTA-COJO conditional and joint analysis, LD score regression (LDSC) for heritability and genetic correlation, LD Hub (757 GWAS datasets), SMR with HEIDI filtering (HEIDI P > 0.01), GSMR for bi-directional Mendelian randomization, FUMA tissue and cell-type expression enrichment, webgestaltR pathway and protein-protein interaction analysis, DropViz single-cell brain expression reference, Polygenic risk scoring with permutation-based P-threshold selection, DeLong's test for AUC comparison, Kernel-based rare coding variant burden tests
Code
http://cnsgenomics.com/software/gcta/
Data & accessions
https://drive.google.com/file/d/1FZ9UL99LAqyWnyNBxxlx6qOUlfAnublN/view?usp=sharinghttp://dropviz.org/
Software & environment
No software versions or containers are stated; named analysis packages include GCTA-COJO, LD score regression, SMR (HEIDI P > 0.01), GSMR, FUMA and webgestaltR, and the PRS cohorts were genotyped on the PD-focused NeuroX array.
Reproducibility gaps
Full summary statistics need a 23andMe agreement (research.23andme.com/collaborate/#publication); public Drive file omits Nalls et al. 2014, post-Chang 23andMe and PDWBS data; IPDGC participant-level data by collaboration request; no study code linked.
Limitations (6)
  • No dedicated external replication cohort was used; 16 of the 90 risk variants fall between P = 5E-08 and 5E-9, a range simulations flag as needing caution without replication, including known signals near WNT3 and BIN3. · stated by the authors
  • Only individuals of European ancestry were analyzed; the authors note that non-European cohorts and ancestry-specific LD reference panels (e.g., Ashkenazi Jewish) are needed to sharpen fine-mapping and conditional analyses at loci such as GBA and LRRK2. · stated by the authors
  • Candidate-gene nomination depends on QTL datasets that are often limited by small sample sizes and sparse cis-SNP coverage; the authors call for larger QTL studies and PD-specific network screens to make functional inference more robust. · stated by the authors
  • The 1805-variant PRS alone cannot support population-wide screening, giving roughly 14 false positives per true positive at 0.5% prevalence; the authors expect that non-genetic inputs such as smell tests, family history, age and sex would be needed. · stated by the authors
  • The authors caution that the bi-directional MR link between smoking initiation and PD needs follow-up, and that the PD-smoking genetic correlation may not be independent of education, as smoking status and years of education are correlated (rG = −0.361). · stated by the authors
  • SMR flagged 151 of 237 testable genes (63.7%) and most loci contained several putatively causal genes, so single-gene resolution is limited; no experimental validation of nominated genes is reported, and sample sizes for the rare-variant burden subset are not given in the main text. · noted in this summary

Who should read:Read this if you study Parkinson's disease genetics and want the 90-signal risk-variant set, PRS performance benchmarks, QTL-based causal gene nominations, or genetic links to brain volume, smoking and education.

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
THIS(GP2) '25Blauwendraa '19(GP2) '26Kim '23(GP2) '26Andrews '23
82% semantic matchGenetic & Genomic Medicine
The Global Parkinson’s Genetics Program (GP2), Hampton L. Leonard·medRxiv·Mar 14, 2025· 61 citations