CoChat
Sign Up
Cited by 422Genetic Associations and EpidemiologyTreatment of Major DepressionNeurotransmitter Receptor Influence on Behavior

Bi-ancestral depression GWAS in the Million Veteran Program and meta-analysis in >1.2 million individuals highlight new therapeutic directions

Open access
Levey, Daniel F., Stein, Murray B., Wendt, Frank R., Pathak, Gita A., Zhou, Hang, Aslan, Mihaela, Quaden, Rachel, Harrington, Kelly M. +12 more·2021·Nature Neuroscience
DOI 10.1038/s41593-021-00860-2PMC PMC8404304

Published version of a preprint

Open this paper in CoChat — free accountOpen PDF

Abstract

Major depressive disorder is the most common neuropsychiatric disorder, affecting 11% of veterans. Here we report results of a large meta-analysis of depression using data from the Million Veteran Program, 23andMe, UK Biobank and FinnGen, including individuals of European ancestry (n = 1,154,267; 340,591 cases) and African ancestry (n = 59,600; 25,843 cases). Transcriptome-wide association study analyses revealed significant associations with expression of NEGR1 in the hypothalamus and DRD2 in the nucleus accumbens, among others. We fine-mapped 178 genomic risk loci, and we identified likely pathogenicity in these variants and overlapping gene expression for 17 genes from our transcriptome-wide association study, including TRAF3. Finally, we were able to show substantial replications of our findings in a large independent cohort (n = 1,342,778) provided by 23andMe. This study sheds light on the genetic architecture of depression and provides new insight into the interrelatedness of complex psychiatric traits. This bi-ancestral genome-wide association study of major depressive disorder (MDD) identified 178 risk variants. The results advance understanding of the biology of MDD and hint at new treatment possibilities.

AI snapshot

Written from the paper’s full text.

TL;DR

This study reports a large bi-ancestral depression GWAS and meta-analysis, combining Million Veteran Program data with prior cohorts to identify many risk loci and prioritize genes and tissues. It also replicates most findings in an independent 23andMe sample and highlights dopaminergic, glutamatergic, and other potential therapeutic directions.

Key findings

  • In European-ancestry meta-analysis (n=1,154,267; 340,591 cases), the authors identified 223 independent significant SNPs across 178 genomic risk loci; a transancestral meta-analysis including African-ancestry MVP participants (n=59,600; 25,843 cases) identified 233 independent SNPs at 183 loci.
  • Replication in an independent 23andMe cohort (n=1,342,778; 455,350 cases) tested 211 variants: 209 had concordant effect directions (99.1%), 192 were nominally significant (91%), 144 survived Bonferroni correction (68%), and 81 were genome-wide significant (38%).
  • TWAS analyses found 153 genes below the Bonferroni threshold across 14 tissues, including NEGR1 in hypothalamus (p=3.19×10−25) and DRD2 in nucleus accumbens (p=1.87×10−20); fine-mapping yielded 1,620 causal-set SNPs, 19 SNPs with CADD >10, and 17 genes overlapping TWAS results.
  • LDSC estimated SNP-based heritability of 11.3% for the primary MDD meta-analysis, and 669 phenotypes were significantly genetically correlated after correction; drug mapping identified 10 annotations with FDR < 0.05, including estrogen-related drugs, nicotine, cocaine, cyclothiazide, felbamate, and riluzole.

Methods

  • GWAS were run in MVP by ancestry and data tranche using PLINK 2.0 with logistic regression for MDD and self-reported depression and linear regression for PHQ-2, adjusting for age, sex, and 10 principal components; meta-analyses used METAL with inverse variance weighting and included PGC/UK Biobank, FinnGen, and 23andMe.
  • Post-GWAS analyses included LDSC, latent causal variable analysis, genomic structural equation modeling, TWAS with MetaXcan/S-MultiXcan using GTEx v8, fine-mapping with CAVIAR, CADD annotation, colocalization, MAGMA/FUMA gene-based tests, and ShinyGO/MATADOR enrichment.
  • Replication used an independent 23andMe GWAS with logistic regression under an additive allelic model, adjusting for age, sex, four principal components, and array platform.

Limitations

  • The paper states that using broad biobank phenotyping to maximize power may reduce specificity for core depression, that the broad FinnGen phenotype showed substantial correlation variance (95% CI 0.72–1.7), and that non-European ancestries remain understudied.
  • Evident from the text: the African-ancestry primary GWAS produced no genome-wide significant findings, and cross-ancestry lookup was limited (206 of 223 European GWS SNPs available; 125 same direction; 20 nominally significant; 1 surviving Bonferroni correction), indicating constrained power and diversity in non-European analyses.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41593-021-00860-2

Cites

55 of 55 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 35 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    Factor 3 represented educational attainment (0.99) and cognitive performance (0.68).

    “Factor 3 represented educational attainment (loading = 0.99 ± 0.03) and cognitive performance (loading = 0.68 ± 0.03).”
    Quote in the paperNot a substantial claim
  • Findingown data

    1620 SNPs in the causal set out of 14,016 GWS hits had high posterior probability for causal relation with MDD-META.

    “1620 SNPs in the causal set out of 14,016 GWS hits have high posterior probability for causal relation with MDD-META”
    Quote in the paperSubstantial claim
  • Findingown data

    Exploratory factor analysis supported three- and four-factor models in which each factor contributed over 10% to cumulative explained variance.

    “Exploratory factor analysis (EFA) was conducted simultaneously on all traits and supported three- (cumulative variance = 0.605) and four-factor models (cumulative variance = 0.624) where each factor contributed over 10% to the cumulative explained variance.”
    Quote in the paperSubstantial claim
  • Findingown data

    After multiple testing correction, 669 phenotypes were significantly genetically correlated with MDD-META.

    “After multiple testing correction (p = 0.05/1,457 trait pairs = 3.43×10−5), 669 phenotypes were significantly genetically correlated with MDD-META”
    Quote in the paperNot a substantial claim
  • Findingown data

    No changes in h2 were observed in the conditioned analyses.

    “We observed no changes in h2.”
    Quote in the paperNot a substantial claim
  • May combine two claims. 31
  • Classified as probably not a substantial claim. 30
  • The quote only partly states this claim. 1

  • May combine two claims.

    Findingown data

    Anorexia nervosa did not load onto any factor during EFA and was excluded from confirmatory factor analyses.

    “Anorexia nervosa did not load onto any factor during EFA and was therefore excluded from confirmatory factor analyses (CFA).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    Confirmatory factor analysis did not converge on a four-factor model because two factors were highly correlated.

    “CFA did not converge on a four-factor model due to high correlation between two factors.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Genetic correlations among depression phenotypes ranged from 0.59 to 1.21, with lower correlations between symptom measures and case-control phenotypes.

    “Genetic correlation between depression phenotypes ranged from 0.59 to 1.21, with lower rg identified between measures of depressive symptoms and case-control phenotypes”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Among 206 European-ancestry GWS SNPs available in the African-ancestry cohort, 61% had the same direction of effect, 20 were nominally significant, and 1 survived Bonferroni correction.

    “Of the 223 GWS SNPs from the primary analysis, 206 were available following QC in the AA cohort. 61% (n=125) of the EUR GWS SNPs had the same direction of effect in AAs, with 20 nominally significant (p<0.05) and 1 surviving Bonferroni correction.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    No brain imaging phenotype met corrected significance for genetic correlation with MDD-META; the most significant was left subcallosal cortex grey matter volume (rg=0.205).

    “No brain imaging phenotype met corrected significance criteria for genetic correlation with MDD-META; the most significantly genetically correlated brain imaging phenotype, using data provided from the Oxford Brain Imaging Genetics (BIG) project,14 relative to MDD-META was left subcallosal cortex grey matter volume (BIG Field ID 0078) rg =0.205±0.061, p=9.00×10-4.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [14] Bycroft, C. et al. The UK Biobank resource with deep phenotyping and genomic data. Nature 562, 203–209 (2018). source

      Not checked in the free run.

  • May combine two claims.

    Findingown data

    MAGMA tissue expression analysis found enrichment across all brain tissues and pituitary, strongest for Brodmann area 9, with no enrichment in non-neuronal tissue.

    “MAGMA tissue expression analysis identified enrichment across all brain tissues and pituitary using data from GTEX v8, with the strongest findings for Brodmann area 9 (p=7.31×10−16), and no enrichment in non-neuronal tissue”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    MATADOR enrichment for 426 MAGMA genes identified 10 drug annotations with FDR below 0.05, including four estrogen receptor agonists or anti-estrogens and nicotine, cocaine, cyclothiazide, felbamate, and riluzole.

    “This analysis identified 10 drug annotations with FDR < 0.05 including four drugs that are either estrogen receptor agonists (diethylstilbestrol, Implanon [etonogestril implant]), or anti-estrogens (tamoxifen and raloxifene); in addition to nicotine, cocaine, cyclothiazide, felbamate, and riluzole.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [19] Gunther, S. et al. SuperTarget and Matador: resources for exploring drug-target relationships. Nucleic Acids Res. 36, D919–D922 (2008). source

      Not checked in the free run.

  • May combine two claims.

    Findingown data

    Gene ontology analysis identified 219 biological processes with FDR below 0.05, with top findings in nervous system development, synapse assembly, and synapse organization.

    “Gene ontology analysis conducted in ShinyGO18 identified 219 biological processes with FDR < 0.05, with top findings involved in nervous system development (q=1.20×10−10), and synapse assembly (q=9.75×10−9) and organization (q=9.75×10−9)”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [18] Ge, S. X., Jung, D. & Yao, R. ShinyGO: a graphical enrichment tool for animals and plants. Bioinformatics 36, 2628–2629 (2019). source

      Not checked in the free run.

  • May combine two claims.

    Findingown data

    SNP-based heritability was 11.3% in the primary MDD-META analysis, 7.8% in SR-depression, and 5.5% in PHQ-2.

    “Heritability in the primary MDD-META analysis was 11.3% (z= 29.63, sample prevalence 28.6%, population prevalence 20%), while heritability in the secondary analyses of self-reported depression (SR-depression, see online methods) and PHQ-2 were 7.8% (z=28.74, sample prevalence 27.1%, population prevalence 20%) and 5.5% (z=14.0), respectively.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    19 SNPs had CADD scores above 10, representing the top 1% of pathogenic variants across the human genome.

    “There were 19 SNPs with CADD scores >10, representing the top 1% of pathogenic variants across the human genome”
    Quote in the paperNot a substantial claim

  • The quote only partly states this claim.

    Findingown data

    The strongest genetic correlation with MDD-META among tiredness/lethargy measures was with UKB frequency of tiredness/lethargy (rg=0.684).

    “(iv) frequency of tiredness/lethargy in last 2 weeks (UKB Field ID 2080) rg =0.684±0.018, p<1.39×10-300.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The top TWAS gene per tissue included ZKSCAN4 in amygdala, L3MBTL2 in anterior cingulate cortex, ZNF184 in caudate, PGBD1 in cerebellar hemisphere, ZSCAN9 in cerebellum, TMEM161B in cortex, FAM120A in frontal cortex, ZSCAN12 in hippocampus, NEGR1 in hypothalamus, DRD2 in nucleus accumbens, LIN28B-AS1 in putamen, HIST1H1B in spinal cord c-1, RP11-318C24.2 in substantia nigra, and ZNF165 in whole血液.

    “Top genes for each tissue tested were: Amygdala (ZKSCAN4, p=1.65×10−12), anterior cingulate cortex (L3MBTL2, p=1.09×10−14), caudate (ZNF184, p=1.85×10−9), cerebellar hemisphere (PGBD1, p=1.67×10−13), cerebellum (ZSCAN9, p=8.4×10−17), cortex (TMEM161B, p=1.84×10−12), frontal cortex (FAM120A, p=3.25×10−10), hippocampus (ZSCAN12, p=1.14×10−18), hypothalamus (NEGR1, p=3.19×10−25), nucleus accumbens (D”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Five gene-tissue pairs had H4 probability of at least 80%: CCDC71-Amygdala, FADS1-Cerebellar hemisphere, SPPL3-Frontal Cortex, TRAF3-Hypothalamus, and LAMB2-whole blood.

    “These gene-tissue pairs were CCDC71-Amygdala (H4-PP: 93.1%), FADS1-Cerebellar hemisphere (H4-PP: 96.6%), SPPL3-Frontal Cortex (H4-PP: 83.9%), TRAF3-Hypothalamus (H4-PP: 95.2%) and LAMB2-whole blood (H4-PP: 79.9%)”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The genome-wide architecture of depression was robust to shared etiology with other listed comorbid conditions, but conditioning on schizophrenia and anxiety symptoms caused substantial loss of GWS SNPs.

    “Though the genome-wide architecture of depression was robust to shared etiology with all other listed comorbid conditions, shared etiology with schizophrenia and anxiety symptoms resulted in substantial loss of GWS SNPs associated with depression when conditioned upon those traits”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The lead SNP rs7531118 (MAF=0.48, p=8.9×10−29) maps close to NEGR1 and is a brain eQTL for NEGR1.

    “The lead SNP from our primary analysis, rs7531118, (MAF=0.48, p=8.9×10−29) maps close to the neuronal growth regulator 1 gene (NEGR1) and is a brain eQTL for NEGR1.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    DRD2 was a top TWAS finding with significant predicted decreased expression in the nucleus accumbens.

    “The D2 dopamine receptor (DRD2) was another top finding from the TWAS analysis (Figure 3 top), with significant predicted decreased expression in the nucleus accumbens.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    LD score regression showed minimal evidence of inflation or confounding, with 97% of inflation due to depression's high polygenicity.

    “LD-intercept (1.03, SE 0.011) and attenuation ratio (0.0297, SE 0.011) of the LD score regression revealed minimal evidence for inflation or confounding, with 97% of inflation observed due to high polygenicity of depression (Figure S4).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductown data

    All conditioned GWAS had correlation coefficient 1.00 with the unconditioned GWAS, and genomic control factor and intercepts indicated no population substructure.

    “All conditioned GWAS had correlation coefficient = 1.00 with the unconditioned GWAS, and genomic control factor and intercepts consistently indicated a lack of population substructure”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationcitation

    LDSC genetic correlations above 1 are possible and occurred when testing the same sample with similar phenotypes or smaller FinnGen samples against large samples.

    “Some of the genetic correlations from the LD score regression were greater than one; genetic correlation from LDSC does not bound to one,13 and the instances with values higher than one occurred when testing in the same sample with similar phenotype (rg 1.07, SE=0.0343) between MDD and SR-Depression within MVP), or between the somewhat smaller FinnGen sample and the large PGC/UKB broad depression”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [13] Bulik-Sullivan, B. K. et al. LD score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat. Genet. 47, 291–295 (2015). source

      Not checked in the free run.

  • May combine two claims.

    Study design and conductown data

    LD-intercept and attenuation ratio indicated minimal inflation or confounding, with 97% of inflation attributable to polygenicity.

    “LD-intercept (1.03, SE 0.011) and attenuation ratio (0.0297, SE 0.011) of the LD score regression revealed minimal evidence for inflation or confounding, with 97% of inflation observed due to high polygenicity of depression.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Premisecitation

    LAMB2 participates in neuropathic pain and alters gene expression in brain pathways linked to depression.

    “LAMB2 is involved in neuropathic pain and influencing gene expression changes in brain pathways implicated in depression.36”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [36] Descalzi, G. et al. Neuropathic pain promotes adaptive changes in gene expression in brain networks involved in stress and depression. Sci. Signal 10, eaaj1549 (2017). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    There is consistent evidence for an association between depleted omega-3 and increased depression risk, though the role of omega-3 supplementation in treatment remains controversial.

    “There is consistent evidence in the literature for an association with depleted omega-3 and increased depression risk, though a role for omega-3 supplementation in treatment of depression is still controversial.33”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [33] Wani, A. L., Bhat, S. A. & Ara, A. Omega-3 fatty acids and the treatment of depression: a review of scientific evidence. Integr. Med. Res. 4, 132–141 (2015). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Allosteric modulation of glutamatergic receptors has been considered a mechanistic treatment target for depression.

    “Allosteric modulation of glutamatergic receptors has been considered a mechanistic treatment target for depression.29”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [29] Alt, A., Nisenbaum, E. S., Bleakman, D. & Witkin, J. M. A role for AMPA receptors in mood disorders. Biochem. Pharmacol. 71, 1273–1288 (2006). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Celf4 mutant mice show aberrations in sodium channel function and increased susceptibility to seizures.

    “Celf4 mutant mice show aberrations in sodium channel function, perhaps through increased NAv 1.6 in the axon initial segment of excitatory neurons, and increased susceptibility to seizures.27”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [27] Sun, W. et al. Aberrant sodium channel activity in the complex seizure disorder of Celf4 mutant mice. J. Physiol. 591, 241–255 (2013). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisenone given

    Larger scale clinical trials have not provided clear evidence to support riluzole's efficacy.

    “However, larger scale clinical trials have not provided clear evidence to support its efficacy.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Riluzole has been identified as a potential antidepressant treatment, with support in rodent models and small clinical studies.

    “Riluzole has already been identified as a potential antidepressant treatment, with support for its antidepressant properties found in rodent models31 and small clinical studies.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [31] Chowdhury, G. M. et al. Transiently increased glutamate cycling in rat PFC is associated with rapid onset of antidepressant-like effects. Mol. Psychiatry 22, 120–126 (2017). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisenone given

    Cyclothiazide is an allosteric modulator of AMPA receptors.

    “Another drug, cyclothiazide, is an allosteric modulator of AMPA (glutamatergic) receptors.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    People treated with interferon face a high risk of developing depressive symptoms.

    “individuals treated with interferon are at high risk to develop depressive symptoms.35”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [35] Chiu, W. C., Su, Y. P., Su, K. P. & Chen, P. C. Recurrence of depressive disorders after interferon-induced depression. Transl. Psychiatry 7, e1026 (2017). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    FADS1 is involved in fatty-acid regulation, and variants in this region have been reported to be associated with depression and substance use disorders.

    “The FADS1 protein product, “Fatty Acid Desaturase 1” is involved in fatty-acid regulation and variants in this region have been reported to be associated with depression and substance use disorders.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisenone given

    The drug screen identified felbamate, which has side effects including increasing depressive symptoms, suicidal ideation, and attempts.

    “This screen also identified an anti-seizure medication, felbamate, which has side effects including increasing depressive symptoms, suicidal ideation, and attempts.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    CELF4 is important in developmental disorders, with deletions of 18q12.2 associated with autism spectrum disorder.

    “This gene is important in developmental disorders, with deletions of the 18q12.2 region which encompass the gene associated with autism spectrum disorder.26”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [26] Gilling, M. et al. A 3.2 Mb deletion on 18q12 in a patient with childhood autism and high-grade myopia. Eur J Hum Genet 16, 312–319 (2008). source

      Not checked in the free run.

  • May combine two claims.

    Premisecitation

    TRAF3 controls type-1 interferon response, and individuals treated with interferon are at high risk to develop depressive symptoms.

    “The TRAF3 protein product, “TNF Receptor Associated Factor 3”, controls type-1 interferon response,34 and it has been reported that individuals treated with interferon are at high risk to develop depressive symptoms.35”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [34] Hacker, H., Tseng, P. H. & Karin, M. Expanding TRAF function: TRAF3 as a tri-faced immune regulator. Nat. Rev. Immunol. 11, 457–468 (2011). source

      Not checked in the free run.

    • cites [35] Chiu, W. C., Su, Y. P., Su, K. P. & Chen, P. C. Recurrence of depressive disorders after interferon-induced depression. Transl. Psychiatry 7, e1026 (2017). source

      Not checked in the free run.

  • May combine two claims.

    Study design and conductnone given

    UK Biobank and MVP are large, non-overlapping samples with consistent phenotypic assessments.

    “UK Biobank and MVP represent large and non-overlapping samples with consistent phenotypic assessments.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Negr-/- mice show irregularities in several brain regions, including reduced hippocampal volume, and abnormalities in social behavior and non-social interest.

    “Negr −/− mice have shown irregularities in several brain regions, including reduced brain volume in the hippocampus and have also shown abnormalities in social behavior and non-social interest.20”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [20] Singh, K. et al. Neural cell adhesion molecule Negr1 deficiency in mouse results in structural brain endophenotypes and behavioral deviations related to psychiatric disorders. Sci. Rep. 9, 5457 (2019). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Variants in SPPL3 were reported to be associated with risk to major depression by Hyde and colleagues.

    “Variants in SPPL3, encoding “Signal Peptide Peptidase Like 3”, were reported to be associated with risk to major depression by Hyde and colleagues.9”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [9] Hyde, C. L. et al. Identification of 15 genetic loci associated with risk of major depression in individuals of European descent. Nat. Genet. 48, 1031–1036 (2016). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    The mesolimbic dopamine reward circuit, including the nucleus accumbens, has long been implicated in depression.

    “The mesolimbic dopamine reward circuit, of which nucleus accumbens is a critical part, has long been implicated in depression.22”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [22] Nestler, E. J. & Carlezon, W. A. Jr. The mesolimbic dopamine reward circuit in depression. Biol. Psychiatry 59, 1151–1159 (2006). source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    Bonferroni correction (p=1.79×10−7) was applied to all gene-tissue pairs tested.

    “We applied Bonferroni correction (corrected p-value threshold = 1.79×10−7) for all gene-tissue pair tested.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductreasoning

    Some depression GWS SNPs were not found in the conditioned depression GWAS due to SNP-matching procedures.

    “Due to SNP-matching procedures to condition depression with other phenotypes, some GWS SNPs for depression were not found in the conditioned depression GWAS.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    UK Biobank PHQ2 GWAS also adjusted for age, sex, and the first 10 principal components.

    “A similar GWAS was performed using linear regression in the UK Biobank samples, also using age, sex, and the first 10 PCs for PHQ2.”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: The data distribution was assumed to be normal but was not formally tested.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    For each conditioned depression GWAS, a subset of SNPs could not be matched directly or by proxy.

    “For each conditioned version of the depression GWS a subset of SNPs could not be matched using direct or proxy SNP matching.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    The gene-based GWS threshold was p=0.05/17,927=2.79×10−6.

    “The GWS threshold for the gene-based test was therefore determined to be p = 0.05/17,927 = 2.79×10-6.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    The MASHR model is biologically informed, uses DAP-G fine-mapped variables, and is recommended by its developers.

    “The MASHR model is biologically informed, with Deterministic Approximation of Posteriors (DAP-G) based fine mapped variables and recommended by the developers.45”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [45] Barbeira, A. N. et al. Exploring the phenotypic consequences of tissue specific gene expression variation inferred from GWAS summary statistics. Nat. Commun. 9, 1825 (2018). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    Genetic correlations were summarized after applying 5% false discovery rate and Bonferroni thresholds.

    “Genetic correlations are summarized by total phenotypes tested, nominally significant (p<0.05), and after application of 5% false discovery rate and Bonferroni thresholds (Figure 2 Lower Panel).”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductcitation

    Genetic causality proportions were interpreted only when both traits had heritability z-scores ≥ 7.

    “Genetic causality proportions (gĉp) were interpreted only when the heritability z-score of both traits was ≥ 7, as determined by LCV, not LDSC.43”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [43] O’Connor, L. J. & Price, A. L. Distinguishing genetic correlation from causation across 52 diseases and complex traits. Nat. Genet. 50, 1728–1734 (2018). source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductown data

    MVP GWAS analyses adjusted for age, sex, and the first 10 principal components.

    “GWAS analysis was carried out in the MVP cohorts by logistic regression for MDD and SR-Depression and by linear regression for PHQ2 within each ancestry group and tranche using PLINK 2.0 on dosage data, covarying for age, sex, and the first 10 PCs.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The data distribution was assumed to be normal but was not formally tested.

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    At least 2 genome-wide significant SNPs after Heidi outlier testing were required to estimate causality between phenotypes.

    “We required at least 2 genome-wide significant SNPs after Heidi outlier testing with which to estimate causality between phenotypes.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductcitation

    Phenotypes with heritability z-scores ≥ 4 were considered suitable for genetic correlation against MDD-META.

    “Phenotypes with heritability z-scores ≥ 4 were considered suitable for genetic correlation against MDD-META. 13”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [13] Bulik-Sullivan, B. K. et al. LD score regression distinguishes confounding from polygenicity in genome-wide association studies. Nat. Genet. 47, 291–295 (2015). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductreasoning

    All data were collected independently, and analysts were blinded to the analysis conditions.

    “All data were collected independently and therefore the analysts were blinded to the conditions of the analysis.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    EFAs were performed until adding factor N contributed less than 10% explained variance.

    “EFAs were performed for 1 through N factors until the addition of factor N contributed less than 10% explained variance to the model.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    No randomization was performed.

    “No randomization was performed.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductcitation

    No statistical methods were used to pre-determine sample sizes, but the sample sizes were similar to those in previous publications.

    “No statistical methods were used to pre-determine sample sizes but our sample sizes are similar to those reported in previous publications. 8,10”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [8] Howard, D. M. et al. Genome-wide association study of depression phenotypes in UK Biobank identifies variants in excitatory synaptic pathways. Nat. Commun. 9, 1470 (2018). source

      Not checked in the free run.

    • cites [10] Howard, D. M. et al. Genome-wide meta-analysis of depression identifies 102 independent variants and highlights the importance of the prefrontal brain regions. Nat. Neurosci. 22, 343–352 (2019). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    The 23andMe replication GWAS adjusted for age, sex, 4 principal components, and array platform.

    “An independent GWAS was run at 23andMe using logistic regression assuming an additive model for allelic effects, while covarying for age, sex, 4 principal components and array platform”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: The data distribution was assumed to be normal but was not formally tested.

  • Classified as probably not a substantial claim. May combine two claims.

    Study design and conductcitation

    Fully causal relationships were deduced for significant trait pairs with gĉp ≥ 0.70; otherwise partial causality was considered.

    “Fully causal relationships were deduced for significant trait pairs with gĉp estimates ≥ 0.70; otherwise gĉp estimates were considered evidence for partial causality.43”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [43] O’Connor, L. J. & Price, A. L. Distinguishing genetic correlation from causation across 52 diseases and complex traits. Nat. Genet. 50, 1728–1734 (2018). source

      Not checked in the free run.

53 of 53

  • The paper relies on this without support.

    The EHR-derived ICD-code definition of MDD in MVP accurately identifies major depressive disorder cases for GWAS.

    “For the ICD code definition of major depressive disorder (see Online Methods for detailed diagnosis definitions), the phenotype with the most available data for the MVP cohort, we conducted a GWAS on 250,215 European individuals (83,810 cases).”

  • The paper relies on this without support.

    The depression phenotype in the 23andMe replication cohort is comparable to the MDD-META phenotype.

    “We performed replication analysis in 1,342,778 independent samples provided by 23andMe, including 455,350 depression cases.”
    “99% of our findings showed concordant direction of effect between these two very large and independent cohorts.”

  • The paper relies on this without support.

    Other ancestries remain understudied relative to Europeans.

    “other ancestries remain understudied in relation to Europeans”

  • The authors state this as an assumption.

    The data distribution was assumed to be normal but was not formally tested.

    “Data distribution was assumed to be normal but this was not formally tested.”
4 of 4

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • GWAS have discovered and successfully replicated genome-wide associations of common gene variants for schizophrenia and major depression.

    “large consortia that have discovered and successfully replicated genome-wide associations of common gene variants (single nucleotide polymorphisms, SNPs) for several disorders including schizophrenia [2] and major depression [3–6].”

    From Neandertal introgression partitions the genetic landscape of neuropsychiatric disorders and associated behavioral phenotypes (2022) · cites this paper as [5]

  • Epidemiological surveys have estimated the workplace costs of depression on low work performance while on the job; in the US, the annual salary-equivalent human capital value of these losses has been estimated in the range of $30·1 billion to $51·5 billion.

    “Epidemiological surveys have estimated the workplace costs of depression on low work performance while on the job; 232 in the US, the annual salary-equivalent human capital value of these losses has been estimated in the range of $30·1 billion 233 to $51·5 billion. 234”

    From Time for united action on depression: a Lancet–World Psychiatric Association Commission (2022) · cites this paper as [232]

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • MAGMA Bonferroni correction uses conflicting gene counts

    Major

    The results describe correction for 16,038 genes, whereas the methods specify a threshold calculated from 17,927 genes. Those denominators imply different significance thresholds.

  • Listed colocalization falls below the stated cutoff

    Minor

    The text says the listed gene-tissue pairs met an H4 probability cutoff of at least 80%, but LAMB2 is listed at 79.9%.

  • Results describe conditioning on an unlisted trait

    Minor

    The results attribute a loss of significant SNPs to conditioning on anxiety symptoms, but the methods' complete list of eight conditioning analyses contains no anxiety-symptoms analysis.

  • Replication methods and results give different discovery-SNP counts

    Minor

    The methods call the discovery set 221 independent genome-wide-significant SNPs, while the primary results report 223. The text does not explain the two-SNP difference before describing availability for replication.

4 of 4

Cites

Flagged first, then by citations

Cited by

Sorted by citations

Related

Authors
Levey, D. F., Stein, M. B., Wendt, F. R., Pathak, G. A., Zhou, H., Aslan, M., Quaden, R., Harrington, K. M., Nuñez, Y. Z., Overstreet, C., Radhakrishnan, K., Sanacora, G., McIntosh, A. M., Shi, J., Shringarpure, S. S., 23andMe Research Team, Program, T. M. V., Concato, J., Polimanti, R., & Gelernter, J.
Affiliations
  • Yale University (US)
  • University of California San Diego (US)
  • VA Boston Healthcare System (US)
  • Boston University (US)
  • University of Kentucky (US)
  • United States Department of Veterans Affairs (US)
  • Royal Edinburgh Hospital (GB)
  • 23andMe (United States) (US)
  • United States Food and Drug Administration (US)
Corresponding author
  • Stein, Murray B.
  • Gelernter, Joel
ORCID
Venue
Nature Neuroscience
Publisher
Springer Science and Business Media LLC
Published
2021-05-27
Volume
24
Issue
7
Pages
954-963
ISSN
  • 1097-6256 (ISSN-L)
  • 1546-1726
DOI
10.1038/s41593-021-00860-2
PMC
PMC8404304
OpenAlex
W3164229919
Semantic Scholar
6444689c844a037a21e97c79935d167fc4a23f75
Cited by
422
References
55
Citations by year
2026: 47 · 2025: 97 · 2024: 84 · 2023: 107 · 2022: 65 · 2021: 21 · 2020: 1
FWCI
10.24
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Neuroscience (publishedVersion)
  • medRxiv (acceptedVersion)
  • PubMed (publishedVersion)
  • PubMed Central (acceptedVersion)
  • eScholarship (California Digital Library) (submittedVersion)
Funders
  • U.S. Department of Veterans Affairs (I01CX001849, I01 BX006482)
  • Office of Research and Development
  • National Alliance for Research on Schizophrenia and Depression
  • National Institutes of Health (UL1TR001863)
  • Medical Research Council (MC_PC_17228)
  • National Institute of Mental Health (T32MH014276)
  • National Center for Advancing Translational Sciences
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genetic Associations and Epidemiology
  • Health Sciences › Medicine › Pharmacology › Treatment of Major Depression
  • Life Sciences › Neuroscience › Cellular and Molecular Neuroscience › Neurotransmitter Receptor Influence on Behavior
Keywords
major depressive disorder, depression genetics, African ancestry, transcriptome-wide association study, NEGR1, DRD2, TRAF3, hypothalamus, nucleus accumbens, fine mapping, psychiatric traits, veterans, therapeutic target
MeSH terms
Major Depressive Disorder, Female, Humans, Male, Veterans, Genetic Predisposition to Disease, Genome-Wide Association Study

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-06.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-06.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review