henhao
Sign in
GeneticsNature Neuroscience
Peer-reviewed journal article

Genome-wide meta-analysis of depression identifies 102 independent variants and highlights the importance of the prefrontal brain regions.

David M. HowardiD, Mark J. AdamsiD, Toni‐Kim ClarkeiD, Jonathan D. HaffertyiD, Jude GibsoniD, Masoud ShiraliiD, Jonathan R. I. ColemaniD, Saskia P. HagenaarsiD, Joey WardiD, Eleanor M. WigmoreiD, Clara AllozaiD, Xueyi SheniD, Miruna C. BarbuiD, Eileen Y. XuiD, Heather Clare WhalleyiD, Riccardo Emilio MarioniiD, David John PorteousiD, Gail DaviesiD, Ian J. DearyiD, Gibran HemaniiD, Klaus BergeriD, Henning TeismanniD, Rajesh RawaliD, Volker AroltiD, Bernhard Theodor BauneiD, , Andrew Mark McIntoshiD

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

Abstract

Major depression is a debilitating psychiatric illness that is typically associated with low mood, anhedonia and a range of comorbidities. Depression has a heritable component that has remained difficult to elucidate with current sample sizes due to the polygenic nature of the disorder. To maximise sample size, we meta-analysed data on 807,553 individuals (246,363 cases and 561,190 controls) from the three largest genome-wide association studies of depression. We identified 102 independent variants, 269 genes, and 15 gene-sets associated with depression, including both genes and gene-pathways associated with synaptic structure and neurotransmission. Further evidence of the importance of prefrontal brain regions in depression was provided by an enrichment analysis. In an independent replication sample of 1,306,354 individuals (414,055 cases and 892,299 controls), 87 of the 102 associated variants were significant following multiple testing correction. Based on the putative genes associated with depression this work also highlights several potential drug repositioning opportunities. These findings advance our understanding of the complex genetic architecture of depression and provide several future avenues for understanding aetiology and developing new treatment approaches.

02 /Specification & Classification

Click any tag to browse matching papers
Citations
2,565
FWCI
32.65
Field percentile
top 1%
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 2 (8 Jan 2019) · CC BY 4.0

A meta-analysis of 807,553 people from 23andMe, UK Biobank and PGC finds 102 depression variants. 87 of them replicate in 1,306,354 individuals, and heritability is enriched in prefrontal brain regions.

Depression heritability is spread across many small-effect variants, so finding them needs very large samples, but clinical diagnosis at that scale is slow and costly; the prior largest meta-analysis (461,134 people) reported 44 loci. Pooling the three largest non-overlapping depression genome-wide association studies (GWAS), with mixed clinical and broad self-reported phenotypes, yields 102 independent variants plus leads on genes, pathways and drug interactions.

Key results

  1. ShownCombining three non-overlapping depression GWAS yields 102 independent genome-wide significant variants in 101 loci. The inflation reflects polygenic signal rather than population stratification, and the three cohorts share most of their genetic architecture.

    8,098,588 variants tested; 9,744 at P < 5 × 10^-8; λGC 1.63 with LDSC intercept 1.015 (0.011); pairwise cohort rG from 0.85 (s.e. 0.03) to 0.87 (s.e. 0.04) across UK Biobank, PGC_139k and 23andMe_307k; 246,363 cases and 561,190 controls. · 102 independent genetic variants associated with…; Genetic correlations with depression; Meta-analysis of genome-wide association study…

  2. ShownAll 102 lead variants have the same direction of effect in an independent 23andMe sample, and most replicate. A polygenic risk score (PRS) built from the meta-analysis explains more clinical MDD variance than a PGC-based score in three target cohorts.

    Replication n = 1,306,354 (414,055 cases, 892,299 controls): 97 of 102 at P < 0.05 and 87 after Bonferroni (P < 4.90 × 10^-4). The meta-analysis PRS beat the Wray PRS in Generation Scotland (975 cases/5,971 controls), Münster (960/834) and BiDirect (811/469); R^2 values appear only in a table. · 102 independent genetic variants associated with…; 23andMe replication sample; Polygenic risk score analysis

  3. ShownSNP heritability of depression is concentrated in conserved genomic regions and central nervous system (CNS) tissue. Brain-region and cell-type analyses point to the anterior cingulate cortex, frontal cortex and neurons.

    Liability-scale SNP h2 0.089 (0.003). Enrichment: conserved regions 17.49× (s.e. 1.68) vs intronic 1.16× and H3K4me1 1.41×. CNS and skeletal muscle enriched (Pcorrected < 0.05). LDSC-SEG across 13 GTEx brain regions was significant for anterior cingulate cortex, frontal cortex and cortex. · Partitioning of the heritability component of…; Partitioned heritability analyses

  4. ShownGene-level tests implicate 269 genes, led by SORCS3, plus Gene Ontology sets for synapses and behaviour. Drug-gene mapping flags DRD2 and psycholeptics as repositioning leads; classic serotonergic genes are not among the hits.

    MAGMA on 17,842 genes, P < 2.80 × 10^-6; SORCS3 P = 2.27 × 10^-19, TCF4 P = 4.55 × 10^-16; 14 GO sets (Pcorrected < 0.05) plus somatosensory pyramidal neurons (P = 3.87 × 10^-5); DGIdb v3.0: 560 interactions across 57 genes and 514 drugs, 47 of them between DRD2 and N05 psycholeptics. · Genes and gene-sets putatively associated with…; Drug - gene interactions; Drug - gene interactions; Gene and gene-set analyses

  5. SuggestedDepression is genetically correlated with 41 traits spanning psychiatric, cardiometabolic and reproductive-timing traits. Mendelian randomization supports a causal effect of neuroticism on depression more robustly than the reverse, and suggests depression increases smoking.

    41 of 234 LD Hub traits at PFDR < 0.01 (schizophrenia rG = 0.32, bipolar 0.33, age at menopause −0.11, Crohn's 0.09). IVW neuroticism on depression: beta = 0.366, PFDR = 2.63 × 10^-21; 0.289 after removing 15 overlapping instruments. Depression on neuroticism: 0.146 falls to 0.112 (PFDR = 0.037). Depression on ever smoked: beta = 0.285. · Genetic correlations with depression; Causal relationships between depression and…; Discussion; Mendelian randomization

Methods, data and reproducibility

Inverse-variance-weighted meta-analysis (Metal) of UK Biobank, 23andMe and PGC summary statistics, then clumping and conditional analysis, 23andMe replication and polygenic risk scoring. Downstream: LDSC genetic correlation and partitioned heritability, two-sample Mendelian randomization, MAGMA gene and gene-set tests, and DGIdb drug-gene mapping.

  • UK Biobank 'broad depression' (ever saw a GP or psychiatrist for nerves, anxiety, tension or depression), reanalysed in a genetically homogeneous subgroup · human (UK Biobank cohort) · n = 127,552 cases, 233,763 controls (n = 361,315)
  • 23andMe_307k discovery cohort: web-survey self-report of clinical diagnosis or treatment for depression (Hyde et al.) · human (23andMe research participants) · n = 75,607 cases, 231,747 controls (n = 307,354)
  • PGC_139k: European-ancestry PGC MDD meta-analysis, excluding the 23andMe and earlier UK Biobank cohorts · human (European ancestry) · n = 43,204 cases, 95,680 controls (n = 138,884)
  • Independent 23andMe replication sample of unrelated Europeans using the same self-reported phenotype as 23andMe_307k · human (European ancestry) · n = 414,055 cases, 892,299 controls (n = 1,306,354)
  • Generation Scotland unrelated subset with SCID-diagnosed MDD, used as a PRS target · human (Generation Scotland cohort) · n = 975 cases, 5,971 controls
  • Münster clinically ascertained MDD inpatients and screened controls, used for out-of-sample PRS · human · n = 960 cases, 834 controls
  • BiDirect cohort: acute depression cases (aged 35 to 65, in treatment) and random local controls, used for PRS · human · n = 811 cases, 469 controls
  • GTEx brain-region expression used for LDSC-SEG enrichment of specifically expressed genes · human (GTEx) · n = 13 brain regions
Methods & tools
Metal (inverse-variance-weighted meta-analysis), BGENIE v1.1, PLINK v1.90b4, LDSC regression (stratified LDSC, LDSC-SEG), LD Hub, MR-Base TwoSampleMR v0.4.9 (R), MAGMA, Gene Ontology / MSigDB v5.2 gene sets, Drug Gene Interaction Database (DGIdb) v3.0, KEGG and WHO ATC drug classification, 1000 Genomes phase 3 European LD reference, circlize v0.4.1
Data & accessions
http://dx.doi.org/10.7488/ds/2458http://www.ukbiobank.ac.uk/https://research.23andme.com/collaborate/http://www.med.unc.edu/pgc/statgenhttp://ldsc.broadinstitute.org/http://geneontology.org/
Reagents & environment
BGENIE v1.1; Plink v1.90b4; MR-Base TwoSampleMR v0.4.9 in R; DGIdb v3.0; MSigDB v5.2; circlize v0.4.1; gene positions on NCBI 37.3 and variants on GRCh37; 1000 Genomes phase 3 European LD reference.
Reproducibility gaps
Only 10,000 variants from the full three-cohort meta-analysis are openly released; full statistics need a 23andMe data transfer agreement. UK Biobank raw data are under licence, and PGC data required secondary analysis proposals.
Limitations (6)
  • The cohorts define depression differently (help-seeking, self-reported diagnosis, mostly clinical MDD). The authors note that clinically ascertained MDD cohorts are still needed to validate large, broadly phenotyped samples. · stated by the authors
  • Some Mendelian randomization tests were not run owing to sample overlap or fewer than 30 instruments (e.g. smoking on depression). Shared LD-linked instruments could confound the bidirectional depression-neuroticism result; once they were removed, only the neuroticism-on-depression effect stayed significant. · stated by the authors
  • Because depression is causally heterogeneous, the drug-repositioning leads are tentative; the authors expect that stratifying patients into subtypes will be needed to separate targets and treatments. · stated by the authors
  • The MHC region is represented only by its single most significant variant, a reporting choice the authors attribute to the complexity of that region. · stated by the authors
  • PGC_139k and the replication sample are European-ancestry and UK Biobank was restricted to a genetically homogeneous subgroup, so whether the loci and PRS transfer to other ancestries is untested here. · noted in this summary
  • Replication uses the same 23andMe self-reported phenotype as one discovery cohort, so it does not independently test clinically defined MDD. Drug-gene links are database lookups with no functional or clinical validation. · noted in this summary

Who should read:Read this if you work on psychiatric genetics and want the 102-variant depression GWAS with replication, brain-region heritability enrichment, Mendelian randomization with neuroticism, or drug-gene repositioning leads.

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
THISLi '25eQTLGen '18Howard '18Grimes '25McIntosh '24Levey '21
77% semantic matchGenetic & Genomic Medicine
Yifan Li, Xinglun Dang, Rui Chen+13 more·Nature Human Behaviour·Feb 24, 2025· 30 citations