CoChat
Sign Up
Cited by 2,514Genetic Associations and EpidemiologyCognitive Abilities and TestingGenetic Mapping and Diversity in Plants and Animals

Gene discovery and polygenic prediction from a genome-wide association study of educational attainment in 1.1 million individuals

Open access
23andMe Research Team, Lee, James J., COGENT (Cognitive Genomics Consortium), Social Science Genetic Association Consortium, Wedow, Robbee, Okbay, Aysu, Kong, Edward, Maghzian, Omeed +72 more·2018·Nature Genetics
DOI 10.1038/s41588-018-0147-3PMC PMC6393768
Open this paper in CoChat — free accountOpen PDF

Abstract

Here we conducted a large-scale genetic association analysis of educational attainment in a sample of approximately 1.1 million individuals and identify 1,271 independent genome-wide-significant SNPs. For the SNPs taken together, we found evidence of heterogeneous effects across environments. The SNPs implicate genes involved in brain-development processes and neuron-to-neuron communication. In a separate analysis of the X chromosome, we identify 10 independent genome-wide-significant SNPs and estimate a SNP heritability of around 0.3% in both men and women, consistent with partial dosage compensation. A joint (multi-phenotype) analysis of educational attainment and three related cognitive phenotypes generates polygenic scores that explain 11-13% of the variance in educational attainment and 7-10% of the variance in cognitive performance. This prediction accuracy substantially increases the utility of polygenic scores as tools in research.

Verified 2026-10-29 · 1 reference not checked yet

AI snapshot

Written from the paper’s full text.

TL;DR

This study expands a GWAS of educational attainment to ~1.1 million individuals, identifying 1,271 autosomal lead SNPs and 10 X-chromosome lead SNPs. Joint multi-phenotype analysis yields polygenic scores explaining 11–13% of educational-attainment variance and 7–10% of cognitive-performance variance.

Key findings

  • The primary EduYears GWAS (N = 1,131,881) identified 1,271 approximately independent genome-wide-significant SNPs (995 at the stricter P < 1×10−8 threshold); median lead-SNP effect size was 1.7 weeks of schooling per allele (5th–95th percentiles: 1.1–2.6 weeks).
  • Lead SNPs showed heterogeneous effects across cohorts: the joint null of homogeneous cohort-level effects was rejected (P = 9.7×10−12), and the inverse-variance-weighted mean cross-cohort genetic correlation was 0.72 (SE = 0.14), distinguishable from one (P = 0.03).
  • X-chromosome analyses (N = 694,894) identified 10 lead SNPs and estimated X-chromosome SNP heritability of ~0.3% in both men and women, consistent with partial dosage compensation.
  • Polygenic prediction: an LDpred score had incremental R² of 12.7% in Add Health and 10.6% in HRS (weighted average 11.4%); MTAG-based scores explained 11–13% of EduYears variance and 7–10% of cognitive-performance variance. In an African-American HRS sample (N = 1,519), the score’s incremental R² was 1.6%, an 85% attenuation.

Methods

  • Sample-size-weighted meta-analysis of 71 quality-controlled cohort-level results files using METAL, restricted to European-ancestry individuals with EduYears measured at age ≥30; ~10 million SNPs from 1000 Genomes phase 3, with LD Score regression intercept used to inflation-adjust test statistics.
  • Additional analyses included within-family association in 22,135 sibling pairs, X-chromosome GWAS in UK Biobank and 23andMe, GWAS of cognitive performance (N = 257,841), math ability (N = 564,698), and highest math (N = 430,445), and MTAG joint analysis of the four phenotypes.
  • Polygenic scores were constructed with LDpred and evaluated in Add Health (N = 4,775), HRS (N = 8,609), and WLS; biological annotation used DEPICT, stratified LD Score regression, and CAVIARBF fine-mapping.

Limitations

  • The paper states that cross-cohort genetic correlation is imperfect for environmentally contingent phenotypes like EA, that within-family effect sizes were roughly 40% smaller than GWAS effect sizes with assortative-mating adjustment explaining at most one third of the deflation, and that GWAS estimates may be biased upward by correlation between EA and rearing environment.
  • The paper states that the score has much lower predictive power in an African-American sample and is expected to have reduced power in other non-European-ancestry groups; it also cautions that the score should not be interpreted as a measure of genetic endowment without controls for environmental amplification.
  • Evident from the text: all primary association analyses were restricted to European-ancestry individuals, and the X-chromosome heritability analysis could not distinguish whether lower heritability reflects smaller per-allele effects or haploidy/partial X-inactivation.

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

Integrity signals

Registry

Crossref · 10.1038/s41588-018-0147-3

Cites

55 of 56 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

20 of 28 cited sources checked · 12 supported · 1 flagged · accuracy not yet measured

  • Findingown data

    Of the 162 lead SNPs identified at genome-wide significance in Okbay et al.'s pooled meta-analysis, 158 pass quality-control filters in our updated meta-analysis.

    “Of the 162 SNPs, 158 pass quality-control filters in our updated meta-analysis.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites Okbay et al. [10] Okbay A Genome-wide association study identifies 74 loci associated with educational attainment. Nature 533, 539– 542 (2016). 27225129 PMC4883595 source

      The paper states that its meta-analysis identified 74 approximately independent genome-wide significant loci, with one lead SNP defined for each locus (Abstract; page unavailable). It does not report 162 lead SNPs. The statement that 158 passed filters in the citing authors’ updated meta-analysis describes their own analysis, not a result attributed to this source. (LLM, conf=0.88)

      Checked by a reasoning model — may be wrong.

      Experimental — accuracy not yet measured
  • Findingcitation

    The finding of an imperfect genetic correlation replicates earlier results from smaller samples.

    “Our finding of an imperfect genetic correlation replicates earlier results from smaller samples16,17.”
    Quote in the paperNot a substantial claim2 of 2 sources checked
    • cites [16] de Vlaming R Meta-GWAS Accuracy and Power (MetaGAP) Calculator Shows that Hiding Heritability Is Partially Due to Imperfect Genetic Correlations across Studies. PLOS Genet. 13, e1006495 (2017). 28095416 PMC5240919 source
    • cites [17] Tropf FC Hidden heritability due to heterogeneity across seven populations. Nat. Hum. Behav (2017). doi: 10.1038/s41562-017-0195-1 PMC5642946 29051922 source

      The Discussion reports limited evidence that genetic effects on reproductive behaviour in one country predict those in another, consistent with an imperfect cross-population genetic correlation. But it also says the study lacked the power to estimate genetic correlations directly; the correlations of 0.5–0.8 were used in simulations, not measured as a replicated finding. (LLM, conf=0.83)

      Experimental — accuracy not yet measured
  • Interpretationcitation

    Theoretical projections that failed to consider heterogeneity of effect sizes were optimistic.

    “However, theoretical projections that failed to consider heterogeneity of effect sizes were optimistic4.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [4] Rietveld CA GWAS of 126,559 individuals identifies genetic variants associated with educational attainment. Science (80−.). 340, 1467– 1471 (2013). PMC3751588 23722424 source

      The passage labeled Abstract discusses projections for polygenic-score performance and reports small SNP effects, but it does not evaluate projections that ignored heterogeneity of effect sizes. Searches for that assumption and for “optimistic” found no passage making the attributed point. (LLM, conf=0.70)

      Experimental — accuracy not yet measured
  • Findingown data

    The 162 single-SNP associations from the largest previous study replicate in the subsample of our data (N = 726,808) at a rate closely matching theoretical projections from our Bayesian framework.

    “In the subsample of our data (N = 726,808) that did not contribute to the earlier study’s analyses, the SNPs replicate at a rate that closely matches theoretical projections derived from our Bayesian framework (Supplementary Figure 3).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [10] Okbay A Genome-wide association study identifies 74 loci associated with educational attainment. Nature 533, 539– 542 (2016). 27225129 PMC4883595 source

      The paper’s replication discussion (page unavailable in our copy) reports that SNPs from a previous educational-attainment study replicated in new cohorts, and compares replication of its 74 lead SNPs with expected numbers. Extended Data Figure 4 gives a replication sample of N = 111,349. These passages do not support the claimed 162 associations, N = 726,808 subsample, or projections from the citing authors’ Bayesian framework. (LLM, conf=0.84)

      Experimental — accuracy not yet measured
  • Premisecitation

    The largest previous GWAS's polygenic score had out-of-sample predictive power of 3.2%.

    “reported that a 10-million-SNP linear predictor (hereafter, polygenic score) had an out-of-sample predictive power of 3.2%10.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [10] Okbay A Genome-wide association study identifies 74 loci associated with educational attainment. Nature 533, 539– 542 (2016). 27225129 PMC4883595 source
  • Classified as probably not a substantial claim. 20
  • May combine two claims. 19
  • The quote only partly states this claim. 1

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

    Premisecitation

    CACNA1H encodes the pore-forming subunit of a voltage-gated calcium channel that has been implicated in the trafficking of NMDA-type glutamate receptors.

    “which encodes the pore-forming subunit of a voltage-gated calcium channel that has been implicated in the trafficking of NMDA-type glutamate receptors24.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [24] Wang G CaV3.2 calcium channels control NMDA receptor-mediated transmission: a new mechanism for absence epilepsy. Genes Dev. 29, 1535– 51 (2015). 26220996 PMC4526737 source

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

    Premisecitation

    For height, the analogous estimate for non-transmitted alleles is only 6%.

    “(For height, the analogous estimate is only 6%.)”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [15] Kong A The nature of nurture: effects of parental genotypes. bioRxiv 219261 (2017). doi: 10.1101/219261 29371463 source

      The section ‘The nature of genetic nurture and other polygenic scores’ (page unavailable) reports an association between height and a non-transmitted height polygenic score, but gives no 6% estimate; it says most of the observed effect is estimated to be due to assortative-mating confounding. The height row of Table 1 (page unavailable) uses an educational-attainment polygenic score and does not establish the claimed 6% figure. (LLM, conf=0.78)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Findingown data

    At the standard false discovery rate threshold of 5%, DEPICT prioritizes 1,838 genes, a tenfold increase relative to the earlier GWAS of EduYears.

    “At the standard false discovery rate (FDR) threshold of 5%, the bioinformatics tool DEPICT19 prioritizes 1,838 genes (Supplementary Table 7), a tenfold increase relative to the DEPICT results from an earlier GWAS of EduYears10.”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites [19] Pers TH Biological interpretation of genome-wide association studies using predicted gene functions. Nat. Commun 6, 5890 (2015). 25597830 PMC4420238 source
    • cites [10] Okbay A Genome-wide association study identifies 74 loci associated with educational attainment. Nature 533, 539– 542 (2016). 27225129 PMC4883595 source

      Extended Data Figure 9 documents DEPICT-prioritized genes for EduYears, supporting the identification of this paper as an earlier DEPICT analysis. Our copy does not give the earlier gene count or establish the claimed tenfold comparison; the 1,838 genes are presented as the citing study’s result, not this source’s. (LLM, conf=0.78)

      Experimental — accuracy not yet measured

  • Classified as probably not a substantial claim.

    Premisecitation

    Educational attainment is an important correlate of many social, economic, and health outcomes.

    “Educational attainment (EA) is moderately heritable1 and an important correlate of many social, economic, and health outcomes2,3.”
    Quote in the paperSubstantial claim1 of 2 sources checked
    • cites [2] Conti G, Heckman J & Urzua S The Education-Health Gradient. Am. Econ. Rev 100, 234– 238 (2010). 24741117 PMC3985402 source
    • cites [3] Cutler DM & Lleras-Muney A Education and Health: Evaluating Theories and Evidence in Making Americans Healthier: Social and Economic Policy as Health Policy (eds. House J, Schoeni R, Kaplan G & Pollack H) (Russell Sage Foundation, 2008).

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationown data

    All major conclusions from the largest previous GWAS of EduYears continue to hold but are statistically stronger.

    “all major conclusions from the largest previous GWAS of EduYears10 continue to hold but are statistically stronger.”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [10] Okbay A Genome-wide association study identifies 74 loci associated with educational attainment. Nature 533, 539– 542 (2016). 27225129 PMC4883595 source

  • Classified as probably not a substantial claim.

    Premisecitation

    A recent paper reports that a polygenic score for EduYears based entirely on parents' non-transmitted alleles is approximately 30% as predictive as a polygenic score based on transmitted alleles.

    “a recent paper15 reports that a polygenic score for EduYears based entirely on parents’ non-transmitted alleles is approximately 30% as predictive as a polygenic score based on transmitted alleles.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [15] Kong A The nature of nurture: effects of parental genotypes. bioRxiv 219261 (2017). doi: 10.1101/219261 29371463 source

  • May combine two claims.

    Findingown data

    Stratified LD Score regression estimated weak enrichment of genes highly expressed in glial cells: 1.08-fold for astrocytes (P = 0.07) and 1.09-fold for oligodendrocytes (P = 0.06) versus 1.33-fold for neurons (P = 2.89×10−11).

    “using stratified LD Score regression21, we estimated relatively weak enrichment of genes highly expressed in glial cells (Supplementary Table 9): 1.08-fold for astrocytes (P = 0.07) and 1.09-fold for oligodendrocytes (P = 0.06) versus 1.33-fold for neurons (P = 2.89×10−11).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [21] Finucane HK Partitioning heritability by functional category using GWAS summary statistics. Nat. Genet 47, 1228– 1235 (2015). 26414678 PMC4626285

      Not checked in the free run.

  • May combine two claims.

    Findingown data

    The difference between bottom and top polygenic-score quintiles in college completion is 45 percentage points in Add Health and 36 percentage points in HRS.

    “The difference between the bottom and top quintiles in Add Health and HRS is, respectively, 45 and 36 percentage points”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The polygenic score is a better predictor of EduYears than household income and a worse predictor than mother's or father's education.

    “The score is a better predictor of EduYears than household income and a worse predictor than mother’s or father’s education.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The inverse-variance-weighted mean genetic correlation of EduYears across pairs of cohorts is 0.72 (SE = 0.14), statistically distinguishable from one (P = 0.03).

    “we found that the inverse-variance-weighted mean genetic correlation of EduYears across pairs of cohorts in our sample is 0.72 (SE = 0.14), which is statistically distinguishable from one (P value = 0.03).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    In a sample of African-Americans from HRS (N = 1,519), the score has an incremental R2 of 1.6%, implying an attenuation of 85%.

    “when our score is used to predict EduYears in a sample of African-Americans from the HRS (N = 1,519), the score only has an incremental R2 of 1.6%, implying an attenuation of 85%.”
    • Also stated in: § DISCUSSION · plain
      “we found that our score for EA has much lower predictive power in an African-American sample than in a European-ancestry sample”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The LDpred score was the most predictive in both cohorts, with incremental R2 of 12.7% in Add Health and 10.6% in HRS (sample-size weighted average 11.4%).

    “This score was the most predictive in both cohorts, with an incremental R2 of 12.7% in AddHealth and 10.6% in HRS (and a sample-size weighted average of 11.4%).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Supplementary Table 14 shows moderately high pairwise genetic correlations, ranging from 0.51 to 0.85, which motivate the multivariate analysis.

    “Supplementary Table 14 shows moderately high pairwise genetic correlations, ranging from 0.51 to 0.85, which motivate the multivariate analysis.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    In Add Health, the incremental R2s of the GWAS and MTAG scores for cognitive performance are 5.1% and 6.9%, respectively.

    “In Add Health, where our measure of cognitive performance is the respondent’s score on a test of verbal cognition, the incremental R2s of the GWAS and MTAG scores are 5.1% and 6.9%, respectively.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Within-family sign concordance is substantially greater than expected by chance but weaker than predicted by our Bayesian framework, even after extending the framework to account for assortative mating.

    “The concordance is substantially greater than expected by chance but weaker than predicted by our Bayesian framework, even after we extend the framework to account for inflation in GWAS coefficients due to assortative mating.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The median effect size of the lead SNPs corresponds to 1.7 weeks of schooling per allele; at the 5th and 95th percentiles, 1.1 and 2.6 weeks, respectively.

    “We found that the median effect size of the lead SNPs corresponds to 1.7 weeks of schooling per allele; at the 5th and 95th percentiles, 1.1 and 2.6 weeks, respectively.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Eight of the likely causal SNPs are non-synonymous, and one (rs61734410) is located in CACNA1H.

    “Eight of these are non-synonymous, and one of these (rs61734410) is located in CACNA1H (Supplementary Figure 9)”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingreasoning

    The omnibus Q-statistic for heterogeneity across all 1,271 lead SNPs is 91,830, with corresponding P value equal to 9.68 × 10−12.

    “This gave us an omnibus Q-statistic of 91,830, with corresponding P value equal to 9.68 × 10−12.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites Cochran38

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

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

    Findingown data

    For height, the assortative-mating adjustment fully explains the deflation of within-family effects.

    “For comparison, when we apply the same method to height, we found that the assortative-mating adjustment fully explains the deflation of the within-family effects.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Premisenone given

    The genetic correlation across pairs of cohorts will be correlated across all observations that share one of their cohorts in common.

    “The genetic correlation across pairs of cohorts will be correlated across all observations that share one of their cohorts in common.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    All cohort-level analyses were restricted to European-ancestry individuals that passed the cohort's quality control and whose EduYears was measured at an age of at least 30.

    “All cohort-level analyses were restricted to European-ancestry individuals that passed the cohort’s quality control and whose EduYears was measured at an age of at least 30.”
    Quote in the paperSubstantial claim

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

    Study design and conductnone given

    We dropped a small number of SNPs with male-female allele frequency differences above 0.005 in UKB.

    “Additionally, we dropped a small number of SNPs with male-female allele frequency differences above 0.005 in UKB.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Study design and conductreasoning

    The posterior distribution mean is shrunken toward zero and is not biased by the winner's curse.

    “Relative to the GWAS effect estimate, the mean of the posterior distribution is shrunken toward zero (because zero is the mean of the prior distribution) and is not biased by the winner’s curse.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The marginal effect size of each SNP is drawn from a mixture distribution with a point mass at zero and a normal component for non-null SNPs.

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

    Study design and conductown data

    To adjust the test statistics for bias in the X chromosome analysis, we inflated the standard errors using the LD Score regression intercept estimated from our main autosomal analysis (1.113).

    “To adjust the test statistics for bias, we inflated the standard errors using the LD Score regression intercept estimated from our main autosomal analysis (1.113).”
    Quote in the paperNot a substantial claim

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

    Study design and conductnone given

    To obtain 95% confidence intervals for incremental R2, we bootstrapped the incremental R2's with 1000 repetitions.

    “To obtain 95% confidence intervals, we bootstrapped the incremental R2’s with 1000 repetitions”
    Quote in the paperNot a substantial claim

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

    Interpretationreasoning

    The implicitly-defined Z2 is interpreted as the additional information contained in the new data.

    “Because this formula holds when Z1 and Z2 are independent, the implicitly-defined Z2 is interpreted as the additional information contained in the new data.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Z1 and Z2 are independent for the replication formula to interpret Z2 as additional information.

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

    Study design and conductreasoning

    For the cognitive outcomes in HRS, we clustered all standard errors at the person level.

    “we also clustered all standard errors at the person level.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S1 (asserted).

    Interpretationreasoning

    If this hypothesis is correct, some of the predictive power of the polygenic score reflects environmental amplification of the genetic effects.

    “If this hypothesis is correct, some of the predictive power of the polygenic score reflects environmental amplification of the genetic effects.”
    Quote in the paperNot a substantial claim

  • The quote only partly states this claim.

    Interpretationown data

    The lead SNPs implicate genes involved in brain-development processes and neuron-to-neuron communication.

    “The SNPs implicate genes involved in brain-development processes and neuron-to-neuron communication.”
    Quote in the paperSubstantial claim

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

    Premisecitation

    Glial cells are at least as numerous as neurons in the human brain.

    “even though glial cells are at least as numerous as neurons in the human brain20”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [20] Azevedo FAC Equal numbers of neuronal and nonneuronal cells make the human brain an isometrically scaled-up primate brain. J. Comp. Neurol 513, 532– 541 (2009). 19226510 source

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

    Study design and conductown data

    We adjusted the test statistics for Math Ability, Highest Math, and Cognitive Performance using the estimated intercepts from LD Score regressions (1.073, 1.105, and 1.046, respectively).

    “We adjusted the test statistics using the estimated intercepts from LD Score regressions (1.073 for Math Ability, 1.105 for Highest Math, and 1.046 for Cognitive Performance).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationnone given

    As sample sizes get large, thousands of lead SNPs will be identified, and polygenic predictors will attain non-trivial levels of predictive power.

    “as sample sizes get large, thousands of lead SNPs will be identified, and polygenic predictors will attain non-trivial levels of predictive power.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    To obtain correct standard errors for the inverse-variance-weighted mean genetic correlation, we used the node-jackknife variance estimator described by Cameron and Miller40.

    “Therefore, to obtain correct standard errors, we used the node-jackknife variance estimator described by Cameron and Miller40.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Cameron and Miller40

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

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

    Study design and conductnone given

    Both sets of X chromosome association results underwent the same set of quality-control filters as the autosomal analyses prior to meta-analysis.

    “Both sets of association results underwent the same set of quality-control filters as the autosomal analyses prior to meta-analysis.”
    Quote in the paperNot a substantial claim

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

    Interpretationreasoning

    If EA-increasing genotypes are associated with parental EA-increasing genotypes, which are in turn associated with rearing environments that promote EA, then failure to control for rearing environment will bias GWAS estimates.

    “if EA-increasing genotypes are associated with parental EA-increasing genotypes, which are in turn associated with rearing environments that promote EA, then failure to control for rearing environment will bias GWAS estimates.”
    Quote in the paperSubstantial claim

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

    Interpretationreasoning

    If greater parental educational attainment positively influences the rearing environment, then GWAS that control imperfectly for rearing environment will yield inflated estimates.

    “If greater parental EA positively influences the rearing environment, then GWAS that control imperfectly for rearing environment will yield inflated estimates.”
    Quote in the paperSubstantial claim

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

    Study design and conductnone given

    The summary statistics for Math Ability, Highest Math, and Cognitive Performance underwent quality control using the same procedures applied to the EduYears results files.

    “The summary statistics underwent quality control using the same procedures applied to the EduYears results files.”
    Quote in the paperNot a substantial claim
37 of 37

  • The authors state this as an assumption.

    The marginal effect size of each SNP is drawn from a mixture distribution with a point mass at zero and a normal component for non-null SNPs.

    “We assume that the marginal effect size of each SNP is drawn from the following mixture distribution: βj~{N(0,τ2)withprobabilityπ0otherwise, where τ2 is the effect-size variance for non-null SNPs and π is the fraction of non-null SNPs in our data.”

  • The paper relies on this without support.

    Add Health is a representative sample of American adolescents and HRS is a representative sample of Americans over age 50.

    “the National Longitudinal Study of Adolescent to Adult Health (Add Health, N = 4,775), a representative sample of American adolescents; and the Health and Retirement Study (HRS, N = 8,609), a representative sample of Americans over age 50.”

  • The authors state this as an assumption.

    The effective number of SNPs (Meff) is the same in males and females for X chromosome SNP heritability estimation.

    “Meff is the effective number of SNPs (which is assumed to be the same in males and females).”

  • The authors state this as an assumption.

    The fraction of causal variants used in LDpred for polygenic score construction is equal to 1.

    “assuming a fraction of causal variants equal to 1”

  • The paper relies on this without support.

    Z1 and Z2 are independent for the replication formula to interpret Z2 as additional information.

    “Because this formula holds when Z1 and Z2 are independent, the implicitly-defined Z2 is interpreted as the additional information contained in the new data.”
5 of 5

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?

  • The LD score regression intercept for the EA GWAS is 1.66, and confounding accounts for 7% of the inflation.

    “According to the LD score regression8 intercept (1.66), confounding accounts for 7% of the inflation, similar to previous GWAS of EA2”

    From Polygenic prediction of educational attainment within and between families from genome-wide association analyses in 3 million individuals (2022) · cites this paper as [2]

    In Results > Primary GWAS of EduYears (page not provided), the paper reports an LD Score regression intercept of 1.11 and estimates that bias accounts for about 5% of the inflation. Those figures contradict the attributed intercept of 1.66 and 7% estimate. (LLM, conf=0.91)

    Checked by a reasoning model — may be wrong.

    Experimental — accuracy not yet measured

  • All 315 lead SNPs at 102 loci have been previously reported in EUR ancestry GWASs (313 in ref. 14 and rs9257925 and rs7224296 in ref. 10).

    “all of which have been previously reported in EUR ancestry GWASs (313 variants in ref. 14 and rs9257925 and rs7224296 in ref. 10).”

    From Shared genetic architectures of educational attainment in East Asian and European populations (2024) · cites this paper as [14]

    The Results, ‘Primary GWAS of EduYears’ (page unavailable), confirm that this was a European-descent GWAS reporting 1,271 genome-wide-significant SNPs. Our copy does not establish that 313 of the citing paper’s 315 lead SNPs were among them, or verify the stated overlap with 102 loci. The two named SNPs are attributed to a different reference, not this source. (LLM, conf=0.77)

    Experimental — accuracy not yet measured

  • In the EAS population, EduYears-associated SNPs were strongly enriched in the brain, parietal lobe and putamen of the central nervous system at FDR <5%, consistent with previous EUR results.

    “In the EAS population, EduYears-associated SNPs were strongly enriched in the brain, parietal lobe and putamen of the central nervous system at an FDR <5% threshold (Supplementary Fig. 9a and Supplementary Table 10), which is consistent with previously published LDSC-SEG results in EUR14.”

    From Shared genetic architectures of educational attainment in East Asian and European populations (2024) · cites this paper as [14]

    The Results (Biological Annotation; page unavailable) report strong central-nervous-system enrichment for genes near EduYears lead SNPs, based on European-descent samples. Our copy does not report enrichment specifically in the parietal lobe or putamen, or establish the claimed FDR <5% threshold for those regions; its central-nervous-system finding uses DEPICT rather than the cited sentence’s LDSC-SEG comparison. (LLM, conf=0.86)

    Experimental — accuracy not yet measured

  • In the UKBB Chinese samples, both the EUR-based PGS and EAS-based PGS showed lower predictive power than in other EAS cohorts in our study or previous studies.

    “In the UKBB Chinese samples, which are genetically EAS but may be environmentally close to EUR, both the EUR-based PGS and EAS-based PGS showed lower predictive power than in other EAS cohorts in our study or previous studies10,14.”

    From Shared genetic architectures of educational attainment in East Asian and European populations (2024) · cites this paper as [14]

    The Results (Polygenic Prediction; page unavailable) report that a European-ancestry-based score had lower predictive power in an African-American sample than in European-ancestry samples. This supports a general concern about prediction across ancestry groups, but the paper does not establish the cited comparison between UKBB Chinese and other East Asian cohorts, or report predictive power for an East Asian-based score. (LLM, conf=0.82)

    Experimental — accuracy not yet measured

  • The two previously unreported SNPs were located beyond ±500 kb of lead SNPs reported in previous EduYears GWAS.

    “they were located beyond ±500 kb of the lead SNPs reported in previous EduYears GWAS (Supplementary Figs. 10 and 11)10,14.”

    From Shared genetic architectures of educational attainment in East Asian and European populations (2024) · cites this paper as [14]

    The Results section (Primary GWAS of EduYears; page unavailable) reports lead SNPs from this GWAS, supporting its use as a source for the earlier lead-SNP list. It does not identify the citing document’s two SNPs or establish that they lie more than ±500 kb from those lead SNPs. (LLM, conf=0.78)

    Experimental — accuracy not yet measured

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.

  • Previous study's sample size differs by one

    Minor

    The previous study's combined discovery-and-replication sample is reported as 405,072 in one Methods section and 405,073 in another.

  • Conflicting genome-wide significance thresholds

    Minor

    The Results section gives the threshold as 5×12−8, while the Methods section gives it as 5×10−8 for identifying lead SNPs.

2 of 2

Cites

Flagged first, then by citations

Research works printed but not linked: 1

  1. Henmon, V. A. C. & Nelson, M. J. Henmon–Nelson Tests of Mental Ability, High School Examination—Grades 7 to 12—Forms A, B, and C. Teacher’s Manual. (Houghton-Mifflin, Boston, 1946).

Cited by

Sorted by citations

Related

Authors
23andMe Research Team, Lee, J. J., COGENT (Cognitive Genomics Consortium), Social Science Genetic Association Consortium, Wedow, R., Okbay, A., Kong, E., Maghzian, O., Zacher, M., Nguyen-Viet, T. A., Bowers, P., Sidorenko, J., Karlsson Linnér, R., Fontana, M. A., Kundu, T., Lee, C., Li, H., Li, R., Royer, R., Timshel, P. N., Walters, R. K., Willoughby, E. A., Yengo, L., Alver, M., Bao, Y., Clark, D. W., Day, F. R., Furlotte, N. A., Joshi, P. K., Kemper, K. E., Kleinman, A., Langenberg, C., Mägi, R., Trampush, J. W., Verma, S. S., Wu, Y., Lam, M., Zhao, J. H., Zheng, Z., Boardman, J. D., Campbell, H., Freese, J., Harris, K. M., Hayward, C., Herd, P., Kumari, M., Lencz, T., Luan, J., Malhotra, A. K., Metspalu, A., Milani, L., Ong, K. K., Perry, J. R. B., Porteous, D. J., Ritchie, M. D., Smart, M. C., Smith, B. H., Tung, J. Y., Wareham, N. J., Wilson, J. F., Beauchamp, J. P., Conley, D. C., Esko, T., Lehrer, S. F., Magnusson, P. K. E., Oskarsson, S., Pers, T. H., Robinson, M. R., Thom, K., Watson, C., Chabris, C. F., Meyer, M. N., Laibson, D. I., Yang, J., Johannesson, M., Koellinger, P. D., Turley, P., Visscher, P. M., Benjamin, D. J., & Cesarini, D.
Affiliations
  • University of Minnesota (US)
  • University of Colorado Boulder (US)
  • Vrije Universiteit Amsterdam (NL)
  • Harvard University (US)
  • University of Southern California (US)
  • The University of Queensland (AU)
  • Statens Serum Institut (DK)
  • Broad Institute (US)
  • University of Tartu (EE)
  • University of Essex (GB)
  • University of Edinburgh (GB)
  • University of Cambridge (GB)
  • 23andMe (United States) (US)
  • Centre Hospitalier Universitaire Vaudois (CH)
  • Geisinger Medical Center (US)
  • Institute of Mental Health (SG)
  • University of North Carolina at Chapel Hill (US)
  • Institute of Genetics and Cancer (GB)
  • Feinstein Institute for Medical Research (US)
  • University of Dundee (GB)
  • University of Toronto (CA)
  • Princeton University (US)
  • National Bureau of Economic Research (US)
  • Karolinska Institutet (SE)
  • Uppsala University (SE)
  • New York University (US)
  • Autism & Developmental Medicine Institute (US)
  • Stockholm School of Economics (SE)
Corresponding author
  • Okbay, Aysu
  • Visscher, Peter M.
  • Benjamin, Daniel J.
ORCID
Venue
Nature Genetics
Publisher
Springer Science and Business Media LLC
Published
2018-07-23
Volume
50
Issue
8
Pages
1112-1121
ISSN
  • 1061-4036 (ISSN-L)
  • 1546-1718
DOI
10.1038/s41588-018-0147-3
PMC
PMC6393768
OpenAlex
W2883573930
Semantic Scholar
d49aaa352d4d714f564ea6d8f79b51f558f42ef9
Cited by
2,470
References
50
Citations by year
2026: 155 · 2025: 271 · 2024: 392 · 2023: 328 · 2022: 375 · 2021: 363 · 2020: 292 · 2019: 244 · 2018: 43 · 2017: 5 · 2016: 1 · 2012: 1
FWCI
132.81
Type
journal article
Language
en
Open access
Licence
www.springer.com/tdm (tdm)
Locations
  • Nature Genetics (publishedVersion)
  • Research at the University of Copenhagen (University of Copenhagen) (publishedVersion, other-oa)
  • PubMed (publishedVersion)
  • Discovery Research Portal (University of Dundee) (publishedVersion, other-oa)
  • Amsterdam UMC (VU Amsterdam) - Institutional Repository (publishedVersion, other-oa)
  • EUR Research Repository (Erasmus University Rotterdam) (publishedVersion, other-oa)
  • University of Groningen research database (University of Groningen / Centre for Information Technology) (publishedVersion, other-oa)
  • SERVAL (Université de Lausanne) (publishedVersion, cc-by)
  • Research Publications (Maastricht University) (acceptedVersion, cc-by)
  • Warwick Research Archive Portal (University of Warwick) (acceptedVersion)
  • Open Access at Essex (University of Essex) (acceptedVersion)
  • IRIS (submittedVersion, cc-by)
  • Radboud Repository (Radboud University) (submittedVersion)
  • Europe PMC (PubMed Central) (submittedVersion)
  • PubMed Central (submittedVersion)
  • EUR Research Repository (Erasmus University Rotterdam) (submittedVersion)
  • ArTS Archivio della ricerca di Trieste (University of Trieste https://www.units.it/) (submittedVersion)
  • UCL Discovery (University College London) (submittedVersion)
  • Max Planck Digital Library (submittedVersion)
  • Lirias (submittedVersion)
  • Site cant be reached (submittedVersion, other-oa)
  • Publikationen an der Universität Bielefeld (Universität Bielefeld) (submittedVersion)
  • Research Explorer (The University of Manchester) (submittedVersion, other-oa)
  • Edinburgh Research Explorer (University of Edinburgh) (submittedVersion, other-oa)
  • MPG.PuRe (Max Planck Society) (submittedVersion)
  • MPG.PuRe (Max Planck Society) (submittedVersion, other-oa)
  • VU Research Portal (submittedVersion, other-oa)
  • OPUS (Augsburg University) (submittedVersion)
  • STM:n Hallinnonalan avoin julkaisuarkisto (Julkari) (submittedVersion)
  • Apollo (University of Cambridge) (submittedVersion)
  • Radboud Repository (Radboud University) (submittedVersion, other-oa)
  • University of Groningen research database (University of Groningen / Centre for Information Technology) (submittedVersion, other-oa)
  • NOVA (University of Newcastle Australia) (submittedVersion)
  • Data Archiving and Networked Services (DANS) (submittedVersion, other-oa)
  • Apollo
Funders
  • Brigham and Women's Hospital
  • Rush University
  • Queen Mary University of London
  • Wellcome Trust (213422/Z/18/Z)
  • Broad Institute
  • Open Philanthropy Project (2016-152872)
  • National Institute for Health and Care Research (NF-SI-0507-10228, NF-SI-0514-10027, NF-SI-0512-10135, NF-SI-0617-10149, NIHR-RP-R3-12-013)
  • University of Warwick
  • University of Oxford
  • University of Manchester
  • University of Aberdeen
  • Rijksuniversiteit Groningen
  • Lundbeckfonden (R190 2014-3904)
  • Karolinska Institutet (2017-00641)
  • King Abdulaziz University
  • Vetenskapsrådet (421-2013-1061, 2017-00641, 00641)
  • Westfälische Wilhelms-Universität Münster
  • Universitair Medisch Centrum Groningen
  • Universität Regensburg
  • Rheinische Friedrich-Wilhelms-Universität Bonn
  • Novo Nordisk Fonden (NNF16OC0021496)
  • National Institutes of Health (2017-00641, 5r01ag015819-21, 5t32ag000186-35, 5p2chd050924-20, R01AG042568, AG042568, 5p01ag005842-36)
  • Medical Research Council (1811434, MR/N01104X/1, MC_PC_13048, MR/K026992/1, MC_UU_12015/2, MR/N01104X/2, MC_UU_12015/1, MC_UU_00007/10)
  • Biotechnology and Biological Sciences Research Council (BB/F022441/1, BB/F019394/1)
  • Economic and Social Research Council (ES/N00812X/1, ES/K005774/1, ES/M008592/1, ES/S008349/1)
  • Warwick Medical School
  • Diamantina Institute, University of Queensland
  • National Institute on Aging (T32–AG000186, T32-AG000186-23, R01 AG042568, P30 AG012810, AG005842, P30AG010161, R01 AG055406, P01-AG005842, P01-AG005842-20S2, R01-AG015819)
  • National Institute of Mental Health (L30 MH104879)
  • Eunice Kennedy Shriver National Institute of Child Health and Human Development ([P2C HD050924], R01HD073342, P2CHD066613, P2CHD047879)
  • H2020 European Research Council (647648)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genetic Associations and Epidemiology
  • Social Sciences › Psychology › Experimental and Cognitive Psychology › Cognitive Abilities and Testing
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genetic Mapping and Diversity in Plants and Animals
Keywords
educational attainment, polygenic risk score, genetic association, single nucleotide polymorphisms, cognitive performance, brain development, X chromosome, dosage compensation, heritability, gene discovery
MeSH terms
Adult, Aged, Aged, 80 and over, Educational Status, Female, Humans, Male, Middle Aged, Phenotype, Cohort Studies, Multifactorial Inheritance, Polymorphism, Single Nucleotide, 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 2026-10-29.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-29.

Cited sources are checked by gpt-6-sol.

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