CoChat
Sign Up
PreprintCited by 47Genetic Associations and EpidemiologyGenetic and phenotypic traits in livestockGenetic Mapping and Diversity in Plants and Animals

Family-GWAS reveals effects of environment and mating on genetic associations

Open access
Tan, Tammy, Jayashankar, Hariharan, Guan, Junming, Nehzati, Seyed Moeen, Mir, Mahdi, Bennett, Michael, Agerbo, Esben, Ahlskog, Rafael +58 more·2024·medRxiv
DOI 10.1101/2024.10.01.24314703PMC PMC12633112
Open this paper in CoChat — free accountOpen PDF

Abstract

Genome-wide association studies (GWAS) have discovered thousands of replicable genetic associations, guiding drug target discovery and powering genetic prediction of human phenotypes and diseases. However, genetic associations can be affected by gene-environment correlations and non-random mating, which can lead to biased inferences in downstream analyses. Family-based GWAS (FGWAS) uses the natural experiment of random assignment of genotype within families to separate out the contribution of direct genetic effects (DGEs) — causal effects of alleles in an individual on an individual — from other factors contributing to genetic associations. Here, we report results from an FGWAS meta-analysis of 34 phenotypes from 17 cohorts. We found evidence that factors uncorrelated with DGEs make substantial contributions to genetic associations for 27 phenotypes, with population stratification confounding — a form of gene-environment correlation — likely the major cause. By estimating SNP heritability and genetic correlations using DGEs, we found evidence that assortative mating has led to overestimation of SNP heritability for 5 phenotypes and overestimation of the degree of shared genetic effects (pleiotropy) between 22 pairs of phenotypes. Polygenic predictors constructed from DGEs are particularly useful for studying natural selection, assortative mating, and indirect genetic effects (effects of relatives’ genes mediated through the family environment). We validate our meta-analysis results by predicting phenotypes in hold-out samples using polygenic predictors constructed from DGEs, achieving statistically significant out-of-sample prediction for 24 phenotypes with little attenuation of predictive power within-families. We provide FGWAS summary statistics for 34 phenotypes that can be used for downstream analyses. Our study provides both a template for performing FGWAS and an argument for its value for debiasing inferences and understanding the impact of environment and mating patterns.

Verified 2026-10-29 · 2 references not checked yet

AI snapshot

Written from the paper’s full text.

TL;DR

Family-based GWAS meta-analysis of 34 phenotypes across 17 cohorts separates direct genetic effects (DGEs) from confounding due to gene-environment correlation and non-random mating. It finds substantial population-stratification confounding in standard GWAS, assortative-mating inflation of heritability and genetic correlations, and validates DGE-based polygenic predictors in hold-out samples.

Key findings

  • Factors uncorrelated with DGEs make substantial contributions to genetic associations for 27 phenotypes; estimated contributions were 10.7% (S.E.=0.6%) for height, 10.2% (S.E.=1.0%) for BMI, 48.3% (S.E.=3.2%) for educational attainment, and 58.7% (S.E.=7.0%) for depressive symptoms, with population stratification likely the major cause.
  • The genome-wide correlation between DGEs and population effects was below 1 for 27/30 phenotypes (FDR<0.05); LDSC-estimated correlations were significantly below 1 for height, myopia, hypertension, allergic rhinitis, depressive symptoms, neuroticism, self-rated health, HDL cholesterol, and educational attainment.
  • Assortative mating led to overestimation of SNP heritability for 5 phenotypes (age at first birth in women, educational attainment, depression, ever-smoker, height) and overestimation of genetic correlations/pleiotropy for 22 trait pairs; e.g., the EA–BMI genetic correlation fell from −0.288 (S.E. 0.026) using population effects to −0.068 (S.E. 0.045) using DGEs.
  • DGE-based polygenic predictors had statistically significant out-of-sample prediction for 24 phenotypes, with the height DGE PGI achieving partial R² 17.9% (95% C.I. 15.8%−20.1%) and generally little within-family attenuation.

Methods

  • FGWAS meta-analysis of 34 phenotypes from 17 cohorts: 16 predominantly European-ancestry cohorts meta-analyzed together, and the China Kadoorie Biobank analyzed separately due to LD differences. Missing parental genotypes were imputed with snipar; variants required INFO ≥ 0.99, and phenotypes with DGE effective sample size below 5000 were excluded, leaving 30 phenotypes for main analyses.
  • Cohort-level FGWAS estimates of DGEs, non-transmitted coefficients, and population effects were combined using fixed-effect multivariate meta-analysis, yielding joint sampling variance-covariance matrices for downstream analyses.
  • Downstream analyses used LDSC and GenomicSEM for SNP heritability, genetic correlations, and functional enrichment; polygenic prediction used PRS-CS weights and snipar family-based PGI regressions in the Millennium Cohort Study (primary) and UK Biobank (secondary).

Limitations

  • Stated: DGE effective sample sizes remain on the order of 10⁴ to 10⁵, an order of magnitude lower than large GWAS meta-analyses; analyses were restricted to common variants (MAF >1%); China Kadoorie Biobank results need confirmation in additional Chinese and East Asian cohorts; functional enrichment power was limited for educational attainment DGEs; some phenotypes were excluded for low DGE effective sample size.
  • Stated: FGWAS requires parental or sibling genotypes, and future biobanks need family-based sampling to realize its potential.
  • Evident from the text: the study includes only one East Asian cohort, so cross-ancestry generalization is limited; several key comparisons rely on LDSC assumptions and imputed parental genotypes, and the paper notes that confounding may still affect DGE PGIs under assortative mating.

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

Integrity signals

Registry

Crossref · 10.1101/2024.10.01.24314703

Cites

67 of 76 references verified0 citation errors · 7 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

109 of 128 cited sources checked · 56 supported · 0 flagged · accuracy not yet measured

  • Premisecitation

    Young et al. presented an approach that enables sibling pair data, parent-offspring pair data, and complete parental genotype data to be jointly analyzed through imputation of missing parental genotypes.

    “Young et al. presented an alternative approach that enables sibling pair data, parent-offspring pair data, and complete parental genotype data to be jointly analyzed in a unified analytical framework through imputation of missing parental genotypes3,20.”
    Quote in the paperSubstantial claim2 of 2 sources checked
  • Premisereasoning

    DGEs include causal effects of the focal variant and correlated variants on the same chromosome.

    “As in GWAS, variants are analyzed one at a time, so DGEs include causal effects of the focal variant and correlated variants on the same chromosome14–16.”
    Quote in the paperSubstantial claim3 of 3 sources checked
    • cites [14] source
    • cites [15] Benjamin, D. J. , Cesarini, D. , Turley, P. & Young, A. S. Social-Science Genomics: Progress, Challenges, and Future Directions. (2024). source
    • cites [16] Veller, C. & Coop, G. Interpreting population- and family-based genome-wide association studies in the presence of confounding. PLoS Biol. (2024). source
  • Premisenone given

    FGWAS uses the natural experiment of random assignment of genotype within families to separate out the contribution of direct genetic effects (DGEs) from other factors contributing to genetic associations.

    “Family-based GWAS (FGWAS) uses the natural experiment of random assignment of genotype within families to separate out the contribution of direct genetic effects (DGEs) — causal effects of alleles in an individual on an individual — from other factors contributing to genetic associations.”
    Quote in the paperSubstantial claim5 of 5 sources checked
    • cites [3] source
    • cites [9] source
    • cites [14] source
    • cites [15] Benjamin, D. J. , Cesarini, D. , Turley, P. & Young, A. S. Social-Science Genomics: Progress, Challenges, and Future Directions. (2024). source
    • cites [16] Veller, C. & Coop, G. Interpreting population- and family-based genome-wide association studies in the presence of confounding. PLoS Biol. (2024). source
  • Premisecitation

    Although population structure can introduce bias into imputed parental genotypes, the resulting bias in DGE estimates is negligible for samples of relatively homogeneous genetic ancestry typically used in GWAS.

    “Although population structure can introduce bias into imputed parental genotypes, the resulting bias in DGE estimates is negligible for the samples of relatively homogeneous genetic ancestry typically used in GWAS3,20.”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites [3] source

      The Methods section, “Effect of population structure” (page not available), says population structure biases imputed parental genotypes and can bias direct-effect estimates. It says that for relatively homogeneous samples this bias is *likely* negligible at the individual-SNP level; the claim states negligibility without that uncertainty or scope limitation. The passage also does not establish that such samples are typically used in GWAS. (LLM, conf=0.85)

      Experimental — accuracy not yet measured
    • cites [20] source
  • Premisecitation

    Estimates of DGEs are less precise than estimates of population effects because they only use within-family genotype variation, which is half of the variation in a random-mating population.

    “Furthermore, even when parental genotypes are available, estimates of DGEs are less precise than estimates of population effects because they only use within-family genotype variation, which is half of the variation in a random-mating population3.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [3] source

      The Main section says that random segregation within families enables unbiased estimates of direct genetic effects, and the Results section on genome-wide analyses says power to analyze individual-SNP direct effects was limited (page numbers unavailable). But the paper does not state the claimed comparison with population-effect precision or that within-family genotype variation is half of population variation. Its reported one-half effective sample size compares parent–offspring pairs with complete parental-genotype data, not direct-effect estimates with population-effect estimates. (LLM, conf=0.82)

      Experimental — accuracy not yet measured
    • The authors state this as an assumption: Under random mating, the population effect βl equals the sum of DGE δl and average NTC αl (βl=δl+αl), and deviations from this relationship are negligible for samples with low levels of structure.
  • May combine two claims. 49
  • Classified as probably not a substantial claim. 18
  • The quote only partly states this claim. 2

  • May combine two claims.

    Premisecitation

    These factors include gene-environment correlation due to IGEs and population stratification as well as confounding due to correlations with genetic variants across the genome due to non-random mating.

    “These factors include gene-environment correlation due to IGEs and population stratification as well as confounding due to correlations with genetic variants across the genome due to non-random mating15,16.”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites [15] Benjamin, D. J. , Cesarini, D. , Turley, P. & Young, A. S. Social-Science Genomics: Progress, Challenges, and Future Directions. (2024). source
    • cites [16] Veller, C. & Coop, G. Interpreting population- and family-based genome-wide association studies in the presence of confounding. PLoS Biol. (2024). source

  • May combine two claims.

    Premisecitation

    Genetic associations (including those from GWAS) can be affected by gene-environment correlations and non-random mating, such as indirect genetic effects, assortative mating, and population stratification, leading to biased inferences.

    “However, genetic associations can be affected by gene-environment correlations and non-random mating, which can lead to biased inferences in downstream analyses.”
    • Also stated in: § Introduction · plain
      “However, recent research3–10 has shown that GWASs are susceptible to confounding from indirect genetic effects (IGEs) from relatives — e.g. an effect of parents’ alleles on offspring education mediated through the alleles’ effects on parental education11 — assortative mating (when there are correlations across parents’ phenotypes and genotypes), and population stratification.”
    Quote in the paperSubstantial claim7 of 8 sources checked
    • cites [3] source
    • cites [4] source

      The Materials and methods section discusses fMRI multivariate analysis and motor-sequence performance, not genetic associations. Searches for gene–environment correlation, non-random mating, GWAS, assortative mating, and population stratification found no passage in our copy addressing the claim. (LLM, conf=0.78)

      Experimental — accuracy not yet measured
    • cites [5] source

      The Introduction discusses interpretation of brain activity during motor-sequence learning, and the Materials and methods discusses statistical testing of imaging data (page numbers unavailable). Searches for gene–environment correlation, non-random mating, assortative mating, and population stratification found no passage addressing bias in genetic associations. (LLM, conf=0.78)

      Experimental — accuracy not yet measured
    • cites [6] source

      The Methods section discusses selecting a non-biased peptide–protein structure dataset, and the Results section discusses correlations between docking-model and crystal-structure measurements; neither addresses genetic associations. Searches for gene–environment correlation, non-random mating, GWAS, indirect genetic effects, assortative mating, and population stratification found no passage supporting the claim. Page numbers were unavailable in our copy. (LLM, conf=0.77)

      Experimental — accuracy not yet measured
    • cites [7] 2022, Cross-trait assortative mating is widespread and inflates genetic correlation estimates, Science

      Not checked in the free run.

    • cites [8] source
    • cites [9] source
    • cites [10] source

  • May combine two claims.

    Premisecitation

    Analytical techniques including PCA and linear mixed models reduce population stratification confounding but often leave residual confounding in GWAS results.

    “Analytical techniques including principal component analysis (PCA) and linear mixed models have been developed to reduce population stratification confounding12, but these techniques often leave residual confounding in GWAS results5,8,10.”
    Quote in the paperSubstantial claim4 of 4 sources checked
    • cites [12] source

      The “Mixed model association methods prevent false-positive associations and increase power” section says mixed linear models address confounding from sample structure. The “Pitfall: using a small subset of markers in GRM” section shows that particular implementations can leave stratification insufficiently corrected. But the paper also describes correction as nearly perfect for common variants, so it does not support the broad claim that PCA and linear mixed models often leave residual confounding in GWAS results. (LLM, conf=0.86)

      Experimental — accuracy not yet measured
    • cites [5] source

      The citation is attached to the claim that PCA and linear mixed models reduce population-stratification confounding. The Materials and methods section discusses multivariate analysis of brain-imaging data, not GWAS; searches for PCA, linear mixed models, population stratification, and genetic ancestry found no passage addressing that claim. The later residual-confounding clause carries different citations. (LLM, conf=0.78)

      Experimental — accuracy not yet measured
    • cites [8] source
    • cites [10] source

  • May combine two claims.

    Premisecitation

    FGWAS has been proposed as a solution to confounding in GWAS that also enables deeper investigation of gene-environment correlation and non-random mating.

    “Family-based GWAS (FGWAS) has been proposed as a solution to the problem of confounding in GWAS3,9,14–16 that also enables deeper investigation of the impact of gene-environment correlation — including from IGEs — and non-random mating, including population structure and assortative mating.”
    Quote in the paperSubstantial claim5 of 5 sources checked
    • cites [3] source
    • cites [9] source
    • cites [14] source
    • cites [15] Benjamin, D. J. , Cesarini, D. , Turley, P. & Young, A. S. Social-Science Genomics: Progress, Challenges, and Future Directions. (2024). source
    • cites [16] Veller, C. & Coop, G. Interpreting population- and family-based genome-wide association studies in the presence of confounding. PLoS Biol. (2024). source

      The Introduction and Discussion say family-based GWASs were adopted to reduce confounding and examine how population structure, assortative mating, and indirect genetic effects affect estimates. But the Discussion cautions that family-based estimates retain some confounding and that estimates of indirect genetic effects can absorb confounds, so the paper does not support presenting FGWAS as an unqualified solution that can disentangle those effects. (LLM, conf=0.87)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Premisecitation

    When applied to sibling pair data, the imputation approach increases effective sample size of DGE estimates by up to one third and of average NTCs by up to one half compared to sib-GWAS.

    “When applied to sibling pair data, this approach increases the effective sample size of DGE estimates by up to one third and of average NTCs by up to one half compared to sib-GWAS3.”
    Quote in the paperSubstantial claim1 of 1 sources checked

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

    Premisecitation

    The LDSC intercept has been proposed as a measure of confounding, with deviations above one argued to represent the extent of spurious inflation in GWAS test statistics.

    “The LDSC intercept has been proposed as a measure of confounding, with deviations above one argued to represent the extent of spurious inflation in GWAS test statistics33.”
    Quote in the paperSubstantial claim1 of 1 sources checked

  • May combine two claims.

    Interpretationreasoning

    Since AM is strong for all of these phenotypes, these results suggest that larger LDSC SNP heritability estimates from population effects are primarily due to AM, with the influence of population stratification diminished to the degree that LDSC successfully adjusts for it.

    “Since AM is strong for all of these phenotypes31 (AM is likely for age at first birth in women indirectly due to the correlation of age at first birth with EA29,37), these results suggest that larger LDSC SNP heritability estimates from population effects are primarily due to AM6, with the influence of population stratification diminished to the degree that LDSC successfully adjusts for it.”
    Quote in the paperSubstantial claim4 of 4 sources checked
    • cites [31] 2023, Evidence of correlations between human partners based on systematic reviews and meta-analyses of 22 traits and UK Biobank analysis of 133 traits, Nat. Hum. Behav source

      Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

      Experimental — accuracy not yet measured
    • cites [29] source

      Split into 2 atoms — 1 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 1 unverifiable. Expand the claim to see per-atom verdicts.

      Experimental — accuracy not yet measured
    • cites [37] source

      Split into 2 atoms — 0 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 2 unverifiable. Expand the claim to see per-atom verdicts.

      Experimental — accuracy not yet measured
    • cites [6] source

      Split into 2 atoms — 0 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 2 unverifiable. Expand the claim to see per-atom verdicts.

      Experimental — accuracy not yet measured

  • May combine two claims.

    Premisecitation

    Polygenic indices based on direct genetic effects from FGWAS have favorable properties due to the removal of confounding from PGI weights, making them suited to applications that are sensitive to confounding.

    “Polygenic predictors (called polygenic indices, PGIs, or polygenic scores) based on DGE estimates from FGWAS – hereafter, DGE PGIs – have favorable properties due to the removal of confounding from PGI weights, making them suited to applications that are sensitive to confounding4,5,9,13,23.”
    Quote in the paperSubstantial claim5 of 5 sources checked
    • cites [4] source

      The Introduction and Materials and methods discuss motor-sequence consolidation and fMRI pattern analysis, not polygenic indices. Searches for direct genetic effects, FGWAS, polygenic scores, and genetic confounding found no passage addressing the claim in our copy; no page numbers were provided. (LLM, conf=0.78)

      Experimental — accuracy not yet measured
    • cites [5] source

      The Materials and methods section discusses fMRI multivariate pattern analysis of motor sequences, not genetic prediction. Searches for polygenic indices, direct genetic effects, FGWAS, and confounding found no passage addressing the attributed claim in our copy. (LLM, conf=0.78)

      Experimental — accuracy not yet measured
    • cites [9] source

      The Discussion (page unavailable) says family-based estimates are likely to be less biased for phenotypes sensitive to demographic and indirect genetic effects, supporting their use in confounding-sensitive applications. But the Main text’s Fig. 2 explanation says within-sibship estimates can still be affected by siblings’ indirect effects, and the Discussion likewise notes that limitation. The paper therefore supports reduced confounding in the estimates used for PGI weights, not its complete removal. (LLM, conf=0.86)

      Experimental — accuracy not yet measured
    • cites [13] source

      The Results section on Phd1-deficient mice concerns muscle mass and mTORC1 activation, not polygenic prediction. Searches for FGWAS, direct genetic effects, polygenic indices, and confounding returned no passage addressing the claim. The available passages have no page numbers. (LLM, conf=0.78)

      Experimental — accuracy not yet measured
    • cites [23] source

  • May combine two claims.

    Findingown data

    We do not replicate the Howe et al. result for cognitive performance: we find similar SNP heritability from DGEs (0.188, S.E. 0.027) and population effects (0.186, S.E. 0.016), whereas Howe et al. found substantially smaller heritability from DGEs (0.14, S.E. 0.043) than from population effects (0.24, S.E. 0.031).

    “However, we do not replicate the Howe et al. result for cognitive performance: we find similar SNP heritability from DGEs (0.188, S.E. 0.027) and population effects (0.186, S.E. 0.016), whereas Howe et al. found substantially smaller heritability from DGEs (0.14, S.E. 0.043) than from population effects (0.24, S.E. 0.031).”
    Quote in the paperNot a substantial claim1 of 1 sources checked
    • cites [9] source

      Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

      Experimental — accuracy not yet measured

  • May combine two claims.

    Premisecitation

    Beyond uncorrected-for population stratification, other phenomena may lead to LDSC intercepts above one, including violation of LDSC assumptions about genetic architecture, differences in LD between the reference and meta-analysis samples, and cryptic relatedness.

    “Beyond uncorrected-for population stratification, other phenomena may lead to LDSC intercepts above one, including: violation of LDSC assumptions about genetic architecture48, differences in LD between the reference and meta-analysis samples, and cryptic relatedness9.”
    Quote in the paperNot a substantial claim2 of 2 sources checked
    • cites [48] source

      The Results section (page unavailable) reports that relatedness between sample members raised the LDSC intercept, and that a simulated violation of regression linearity could produce an intercept above one without confounding. The passages do not establish that differences in LD between reference and meta-analysis samples raise the intercept; searches for that specific mechanism found no supporting passage. (LLM, conf=0.83)

      Experimental — accuracy not yet measured
    • cites [9] source

  • May combine two claims.

    Study design and conductcitation

    The effective sample size for direct genetic effects is on the order of 10^4 to 10^5, an order of magnitude lower than the most powerful GWAS meta-analyses, which have sample sizes in the millions.

    “the effective sample size for DGEs (on the order of 104 to 105) remains an order of magnitude lower than for the most powerful GWAS meta-analyses, which have sample sizes in the millions17,45.”
    Quote in the paperSubstantial claim1 of 2 sources checked
    • cites [17] source

      Not checked in the free run.

    • cites [45] source

      The ‘GWAS discoveries, sample size and ancestry diversity’ section reports height GWAS sample sizes up to about 5.3 million, supporting the comparison to GWAS meta-analyses with millions of participants. It does not report an effective sample size of 10^4–10^5 for direct genetic effects, so it does not establish the claimed gap. (LLM, conf=0.87)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Premisecitation

    These factors include environmental factors and direct genetic effects of other variants with which the PGI is correlated due to non-random mating.

    “These include environmental factors and DGEs of other variants with which the PGI is correlated due to non-random mating15,23.”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites [15] Benjamin, D. J. , Cesarini, D. , Turley, P. & Young, A. S. Social-Science Genomics: Progress, Challenges, and Future Directions. (2024). source
    • cites [23] source

  • May combine two claims.

    Interpretationcitation

    Inferences from LD score regression applied to GWAS results appear fairly robust to confounding due to population stratification, at least some of which is adjusted out by LDSC.

    “Inferences drawn from the application of LD score regression (LDSC) to GWAS results appear to be fairly robust to confounding due to population stratification, at least some of which is adjusted out by LDSC5,33.”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites [5] source

      The Materials and methods section describes fMRI multivariate analysis and statistical testing, not LD score regression of GWAS results. Searches for LDSC, GWAS, population stratification, and related genetic-association terms found no passage addressing whether LDSC adjusts for stratification. (LLM, conf=0.78)

      Experimental — accuracy not yet measured
    • cites [33] source

  • May combine two claims.

    Premisecitation

    In FPGI analysis, adding parental PGIs as covariates enables estimation of the direct effect of the PGI, which reflects only direct genetic effects.

    “In the FPGI analysis, we add parental PGIs as covariates, enabling estimation of the ‘direct effect’ of the PGI, which reflects only DGEs15,16.”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites [15] Benjamin, D. J. , Cesarini, D. , Turley, P. & Young, A. S. Social-Science Genomics: Progress, Challenges, and Future Directions. (2024). source
    • cites [16] Veller, C. & Coop, G. Interpreting population- and family-based genome-wide association studies in the presence of confounding. PLoS Biol. (2024). source

      The “Polygenic scores and their phenotypic associations” section (page unavailable) describes parent-offspring PGS regressions and says within-family PGS slopes are often interpreted as capturing direct genetic effects. But the Discussion’s “Direct genetic effects” section warns that population-GWAS-derived PGS weights can absorb confounding. The paper does not establish that adding parental PGIs as covariates yields an estimate reflecting only direct genetic effects. (LLM, conf=0.86)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Interpretationcitation

    The primary explanation for positive LDSC estimates of correlation between DGEs and average NTCs for height and EA is likely assortative mating, which is strong for EA and height, although IGEs that are positively correlated with DGEs could also contribute.

    “The primary explanation is likely AM, which is strong for EA and height25,31, although IGEs that are positively correlated with DGEs could also contribute3.”
    Quote in the paperSubstantial claim3 of 3 sources checked
    • cites [25] 2018, Imprint of assortative mating on the human genome, Nat. Hum. Behav source

      Split into 2 atoms — 1 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 1 unverifiable. Expand the claim to see per-atom verdicts.

      Experimental — accuracy not yet measured
    • cites [31] 2023, Evidence of correlations between human partners based on systematic reviews and meta-analyses of 22 traits and UK Biobank analysis of 133 traits, Nat. Hum. Behav source

      The Results (Partner Correlations in the UK Biobank; page unavailable) report substantial partner similarity for educational attainment (r = 0.48) but a smaller correlation for height (r = 0.27), which the paper characterizes as low-to-moderate rather than strong. This citation supports assortative mating for both traits, but not the characterization of height assortment as strong. The sentence cites other sources for the possible IGE contribution; this verdict does not assess that clause. (LLM, conf=0.84)

      Experimental — accuracy not yet measured
    • cites [3] source
    • The authors state this as an assumption: Under assortative mating at equilibrium, βl=δl/(1−rδ), where rδ is the correlation between parents’ DGE components, and thus αl=βl−δl=rδ/(1−rδ)δl.

  • Classified as probably not a substantial claim.

    Premisecitation

    In cases of a large product of sample size and SNP heritability, a normalization of the LDSC intercept called the 'ratio' has been proposed as a more appropriate measure of confounding.

    “In cases of a large product of sample size and SNP heritability, a normalization of the LDSC intercept called the ‘ratio’ has been proposed as a more appropriate measure of confounding49.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [49] Loh, P.-R. , Kichaev, G. , Gazal, S. , Schoech, A. P. & Price, A. L. Mixed-model association for biobank-scale datasets. Nat. Genet. 1 (2018) doi:10/gfxkmc. source

  • May combine two claims.

    Findingcitation

    Our SNP heritability estimates are generally similar to those from Howe et al., although with greatly increased precision for some phenotypes.

    “Our SNP heritability estimates are generally similar to those from Howe et al.9 although with greatly increased precision for some phenotypes.”
    Quote in the paperNot a substantial claim1 of 1 sources checked

  • May combine two claims.

    Interpretationreasoning

    Assortative mating, rather than population stratification, is likely the primary explanation for the differences in SNP heritability, although indirect genetic effects could also contribute.

    “AM (rather than population stratification) is likely to be the primary explanation6, although IGEs could also contribute44.”
    Quote in the paperSubstantial claim2 of 2 sources checked
    • cites [6] source

      The Results section discusses peptide–protein docking and interface hotspots, while the Discussion addresses protein-binding motifs; neither addresses SNP heritability. Searches for assortative mating, population stratification, and indirect genetic effects found no relevant passage in our copy. Page numbers were unavailable. (LLM, conf=0.78)

      Experimental — accuracy not yet measured
    • cites [44] source

      The citation is attached to the claim that assortative mating, rather than population stratification, is the likely primary explanation. Our copy discusses population-stratification bias in the Results (‘Simulation of RDR heritability estimation’) and indirect genetic effects in the Discussion, but does not identify assortative mating as an explanation or compare its importance with stratification. Searches for ‘assortative mating’ and that proposed comparison found no supporting passage; the separate indirect-genetic-effects clause cites another source. (LLM, conf=0.75)

      Experimental — accuracy not yet measured

  • May combine two claims.

    Premisecitation

    SNP heritability can include contributions from IGEs, population stratification, and can be inflated by AM.

    “Thus, SNP heritability can include contributions from IGEs, population stratification (more relevant for GREML than LDSC, which attempts to adjust for population stratification), and can be inflated by AM6.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [6] source

      The Results passages on AF2 peptide–protein docking and the Discussion passage on binding motifs do not address SNP heritability. Searches for indirect genetic effects, population stratification, GREML, LDSC, and assortative mating found no relevant passage in our copy. (LLM, conf=0.78)

      Experimental — accuracy not yet measured

  • Classified as probably not a substantial claim.

    Premisecitation

    GenomicSEM allows users to fit structural equation models based on GWAS summary statistics.

    “GenomicSEM allows users to fit structural equation models based on GWAS summary statistics.”
    Quote in the paperSubstantial claim1 of 1 sources checked
    • cites [39] 2019, Genomic structural equation modelling provides insights into the multivariate genetic architecture of complex traits, Nat. Hum. Behav source

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

    Study design and conductcitation

    To perform quality control on CKB summary statistics, EAS allele frequencies from 1000 Genomes were used.

    “To perform quality control on these summary statistics, we used EAS allele frequencies from 1000 Genomes50.”
    Quote in the paperNot a substantial claim1 of 1 sources checked

  • May combine two claims.

    Findingown data

    Assortative mating biases estimates of SNP heritability and genetic correlations.

    “We find evidence that there is substantial population stratification bias in population-effect estimates across a broad range of phenotypes and that assortative mating biases estimates of SNP heritability and genetic correlations.”
    Quote in the paperSubstantial claim

  • The quote only partly states this claim.

    Findingown data

    Similar patterns of lower cohort-level correlations than meta-analysis correlations were observed for other phenotypes.

    “Similar patterns were observed for other phenotypes (Supplementary Table 3).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Cognitive performance, educational attainment, allergic rhinitis, height, and household income direct genetic effect PGIs had average NTCs showing evidence of being non-zero (two-sided P-value < 0.05), although they were not statistically distinguishable from zero after Benjamini-Hochberg correction.

    “Cognitive performance, educational attainment, allergic rhinitis, height, and household income DGE PGIs had average NTCs that show evidence of being non-zero (two-sided P-value < 0.05), although they were not statistically distinguishable from zero after Benjamini-Hochberg correction.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Height and systolic blood pressure have inflated population effects relative to DGEs, whereas 10 phenotypes (ADHD, eczema, depression, cognitive performance, household income, drinks-per-week, subjective wellbeing, neuroticism, age at first birth in women, and number of children) have deflated population effects (FDR<0.05, two-sided test).

    “Height and systolic blood pressure have inflated population effects, whereas 10 phenotypes (ADHD, eczema, depression, cognitive performance, household income, drinks-per-week, subjective wellbeing, neuroticism, age at first birth in women, and number of children) have deflated population effects (FDR<0.05, two-sided test).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The correlation between DGEs and average NTCs was different from zero (FDR<0.05, two-sided test) for 8 out of 24 phenotypes: below zero for ADHD, allergic rhinitis, eczema, cognitive performance, and household income; and above zero for systolic blood pressure, EA, and height.

    “We found that the correlation between DGEs and average NTCs was different from zero (FDR<0.05, two-sided test) for 8 out of 24 phenotypes (Supplementary Table 3): below zero for ADHD, allergic rhinitis, eczema, cognitive performance, and household income; and above zero for systolic blood pressure, EA, and height (Supplementary Figure 6).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Genetic correlation estimates from population effects are different from those from direct genetic effects for 22 pairs of phenotypes, with some pairs displaying qualitatively different estimates.

    “We demonstrated that genetic correlation estimates from population effects are different from those from DGEs for 22 pairs of phenotypes, with some pairs displaying qualitatively different estimates.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Direct genetic effect PGIs for diastolic and systolic blood pressure, HDL and non-HDL cholesterol, and educational attainment all achieved partial R2 above 1%.

    “with DGE PGIs for diastolic and systolic blood pressure, HDL and non-HDL cholesterol, and EA all achieving partial R² above 1% (Supplementary Table 9).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Population effects on EA are inflated by a factor of 1.131 (S.E.=0.072), but this estimate is not statistically distinguishable from 1.

    “We also estimated that population effects on EA are inflated by a factor of 1.131 (S.E.=0.072), but this estimate is not statistically distinguishable from 1.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    For depression, SNP heritability was estimated at 2.5% (S.E. 1.5%) using direct genetic effects and at 8.2% (S.E. 1.0%) using population effects.

    “including depression, for which SNP heritability was estimated at 2.5% (S.E. 1.5%) when using DGEs and at 8.2% (S.E. 1.0%) when using population effects.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Estimates of the contribution to population effects from factors uncorrelated with DGEs were 10.7% (S.E.=0.6%) for height and 10.2% (S.E.=1.0%) for BMI but reached 48.3% (S.E.=3.2%) for EA and 58.7% (S.E.=7.0%) for depressive symptoms.

    “Estimates of the contribution to population effects from factors uncorrelated with DGEs were 10.7% (S.E.=0.6%) for height and 10.2% (S.E.=1.0%) for BMI but reached 48.3% (S.E.=3.2%) for EA and 58.7% (S.E.=7.0%) for depressive symptoms.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingreasoning

    For EA, the meta-analysis estimate of the cohort-level correlation between DGEs and population effects was 0.547 (S.E.=0.052), lower than from the meta-analysis summary statistics, 0.689 (S.E.=0.023).

    “For EA, the meta-analysis estimate of the cohort-level correlations was 0.547 (S.E.=0.052), lower than from the meta-analysis summary statistics, 0.689 (S.E.=0.023).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    Among the 18 phenotypes with an acceptable product of DGE sample size and SNP heritability, the average DGE-based LDSC ratio was 0.104, smaller than their average population-based ratio of 0.306.

    “Overall, the 18 phenotypes with an acceptable product of DGE sample size and SNP heritability showed an average DGE-based LDSC ratio of 0.104, smaller than their average population-based ratio of 0.306—consistent with a successful removal of confounding bias.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisenone given

    FGWAS requires parental genotypes, which are often not available in typical GWAS samples.

    “While FGWAS has favorable theoretical properties, it requires parental genotypes, which are often not available in typical GWAS samples.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationcitation

    Howe et al. used variants with INFO score greater than 0.3, implying many low-quality imputed variants were analysed, which may have affected the results and conclusions drawn from their study.

    “In this regard, our study differs from the sib-GWAS performed by Howe et al., which used variants with INFO score greater than 0.3, implying many low-quality imputed variants were analysed, which may have affected the results and conclusions drawn from their study.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Howe et al.

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

  • May combine two claims.

    Premisereasoning

    FGWAS estimates DGEs by adding parents' genotypes to regressions, using random assignment of genotype within-family due to Mendelian segregation.

    “FGWAS adds the parents’ genotypes to the regressions performed in GWAS (Methods), thereby using the natural experiment of random assignment of genotype within-family — due to Mendelian segregation during meiosis — to estimate ‘direct genetic effects’ (DGEs).”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    We imposed a stringent imputation quality threshold (INFO score at least 0.99) for our analysis plan.

    “We therefore imposed a stringent imputation quality threshold (INFO score at least 0.99) for our analysis plan (Supplementary File 1).”
    Quote in the paperSubstantial claim

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

    Study design and conductcitation

    We included genetic principal components as covariates in cohort-level analyses to ensure that population effects derived from our analyses are comparable to those derived from standard GWAS.

    “Although not necessary to remove confounding from estimates of DGEs, we included genetic principal components as covariates in cohort-level analyses to ensure that population effects derived from our analyses are comparable to those derived from standard GWAS3,20.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data

    We found evidence supporting Border et al.'s hypothesis that genetic correlation estimates have been inflated by cross-trait assortative mating, but other factors may also contribute, such as indirect genetic effects.

    “We found evidence supporting the hypothesis put forward by Border et al.7 that genetic correlation estimates have been inflated by cross-trait AM, but other factors may also contribute, such as IGEs.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Border et al. [7] 2022, Cross-trait assortative mating is widespread and inflates genetic correlation estimates, Science

      Not checked in the free run.

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

    Premisenone given

    The predominance of GWAS in human genetics has led to study designs that prioritize sampling the maximum number of unrelated individuals.

    “The predominance of GWAS in human genetics has led to study designs that prioritize sampling the maximum number of unrelated individuals.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). The quote only partly states this claim.

    Interpretationcitation

    Within-family contrast effects, where family members differentiate from each other, inducing IGEs in the opposite direction to DGEs, could contribute to negative correlations between DGEs and average NTCs.

    “Another phenomenon that could contribute is within-family contrast effects, where family members (e.g. siblings) differentiate from each other, inducing IGEs in the opposite direction to DGEs30.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [30] source

  • May combine two claims.

    Premisereasoning

    This sampling strategy maximizes power to discover genotype-phenotype associations but often results in datasets without many first-degree relative pairs.

    “This sampling strategy maximizes power to discover genotype-phenotype associations but often results in datasets without many first-degree relative pairs.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Negative correlations between DGEs and average NTCs - and concomitant deflation of population effects - could be due to collider bias induced by biased sampling with respect to phenotype values.

    “Young et al. showed that negative correlations between DGEs and average NTCs — and concomitant deflation of population effects — could be due to collider bias induced by biased sampling with respect to phenotype values3.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Young et al.

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

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    Multivariate meta-analysis may introduce some bias into direct genetic effect estimates due to heterogeneity in the parameter vector across cohorts.

    “multivariate meta-analysis may introduce some bias into DGE estimates due to heterogeneity in the parameter vector across cohorts.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    LDSC-estimated correlations between DGEs and population effects below 1 could indicate the influence of confounding factors other than population stratification in standard GWAS - such as IGEs - or imperfect control for population stratification by LDSC.

    “LDSC-estimated correlations between DGEs and population effects below 1 could therefore indicate the influence of confounding factors other than population stratification in standard GWAS — such as IGEs — or imperfect control for population stratification by LDSC5.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationown data

    Confounding factors uncorrelated with DGEs make a relatively small but non-negligible contribution to GWAS of traits such as height and BMI but comprise the majority of population effects for some phenotypes.

    “These results indicate that confounding factors uncorrelated with DGEs make a relatively small but non-negligible contribution to GWAS of traits such as height and BMI but comprise the majority of population effects for some phenotypes.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    Samples were restricted to homogeneous European ancestry subsamples for the 16 cohorts of predominantly European ancestries.

    “Samples were restricted to homogeneous European ancestry subsamples for the 16 cohorts of predominantly European ancestries with the exact procedure varying depending upon cohort.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premiseown data

    These phenotypes are not closely related to education or cognitive ability, the only previous phenotypes shown to have correlations below one.

    “phenotypes not closely related to education or cognitive ability, the only previous phenotypes shown to have correlations below one.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationcitation

    Howe et al.'s analysis only examined SNPs that were strongly associated (P<5×10−8 or P<1×10−5) based on standard GWAS in a UK Biobank subsample and weighted the analysis towards SNPs with stronger population effects, which may have contributed toward their analysis finding more deflation than there is genome-wide.

    “However, their analysis only examined SNPs that were strongly associated (P<5×10−8 or P<1×10−5) based on standard GWAS in a UK Biobank subsample and weighted the analysis towards SNPs with stronger population effects, which may have contributed toward their analysis finding more deflation than there is genome-wide.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites Howe et al.

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

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    Our analyses were restricted to common variants (minor allele frequency >1%).

    “our analyses were restricted to common variants (minor allele frequency >1%)”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationown data

    Our analyses showed the value of FGWAS in debiasing inferences drawn from GWAS and in investigating the impact of gene-environment correlation and non-random mating.

    “Our analyses showed the value of FGWAS in debiasing inferences drawn from GWAS and in investigating the impact of gene-environment correlation and non-random mating.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationown data

    In the China Kadoorie Biobank, the observed results are consistent with strong ascertainment bias and/or confounding due to negative gene-environment correlations.

    “We observed results consistent with strong ascertainment bias and/or confounding due to negative gene-environment correlations (Supplementary Table 4)”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Study design and conductreasoning

    We excluded from further analysis phenotypes where our meta-analysis effective sample size for DGEs was below 5000: ever-cannabis, COPD, alcohol use disorder, and extraversion, leaving 30 phenotypes.

    “To avoid under-powered analyses, we excluded from further analysis phenotypes where our meta-analysis effective sample size for DGEs was below 5000: ever-cannabis, chronic obstructive pulmonary disease (COPD), alcohol use disorder, and extraversion. This left 30 phenotypes for the subsequent analyses that we report on below (Table 1).”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    Phenotypes were adjusted for standard covariates: age, sex, and genetic principal components.

    “Phenotypes were adjusted for standard covariates: age, sex, and genetic principal components.”
    Quote in the paperSubstantial claim

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

    Interpretationreasoning

    The lack of susceptibility of functional enrichment estimates to confounding may be due to adjustments for stratification made by LDSC and/or because stratification may affect all loci similarly irrespective of functional annotation.

    “potentially due to adjustments for stratification made by LDSC and/or because stratification may affect all loci similarly irrespective of functional annotation.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationcitation

    It is plausible that the deflation of population effects we observed for 9 phenotypes is due in part to ascertainment bias, although natural selection may also contribute through negative LD induced between sign-concordant causal alleles, consistent with the Bulmer effect.

    “It is therefore plausible that the deflation of population effects we observed for 9 phenotypes is due in part to ascertainment bias, although natural selection (specifically, directional or stabilizing selection) may also contribute through negative LD induced between sign-concordant causal alleles, consistent with the Bulmer effect28.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [28] source

  • May combine two claims.

    Interpretationown data

    Most of the correlations estimated by LDSC were close to 1, indicating that much of the confounding in GWAS is likely due to uncorrected population stratification.

    “However, most of the correlations estimated by LDSC were close to 1, indicating that much of the confounding in GWAS is likely due to uncorrected population stratification.”
    Quote in the paperSubstantial claim

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

    Interpretationreasoning

    Confounding at the variant level varies somewhat randomly across cohorts and is thus partially cancelled out in meta-analysis estimates of population effects, leading to increased correlation with DGEs.

    “A plausible explanation for this is that confounding at the variant level varies somewhat randomly across cohorts and is thus partially cancelled out in meta-analysis estimates of population effects, leading to increased correlation with DGEs.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationreasoning

    This pattern is consistent with deflation of population effects for eczema, cognitive performance and household income and inflation for height and systolic blood pressure.

    “This is consistent with deflation of population effects for eczema, cognitive performance and household income and inflation for height and systolic blood pressure.”
    Quote in the paperSubstantial claim

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

    Study design and conductown data

    Summary statistics provided by each cohort were passed through a quality control pipeline before meta-analysis was performed.

    “Summary statistics provided by each cohort were passed through a quality control pipeline, described in Supplementary Note Section 1, before meta-analysis 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 conductown data

    In some cases, the procedure for computing statistical significance of differences between genetic correlations yielded correlations outside of [−1, 1], leading to convergence errors; when this occurred, the model was rerun constraining the relevant estimates to fall within [−0.9999, 0.9999].

    “In some cases, this procedure yielded correlations outside of [−1, 1], leading to convergence errors. When this occurred, we reran the model, constraining the relevant estimates to fall within [−0.9999, 0.9999].”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    A relationship between cross-mate phenotype correlations and DGE genetic correlations can be explained by pleiotropy and univariate AM, which would not be expected to generate spurious genetic correlations between phenotypes without any pleiotropy.

    “Since DGE based genetic correlations should be almost entirely free from the influence of xAM, a relationship between cross-mate phenotype correlations and DGE genetic correlations can be explained by pleitropy and univariate AM, which would not be expected to generate spurious genetic correlations between phenotypes without any pleiotropy.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: DGE-based genetic correlations should be almost entirely free from the influence of xAM.

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    To reduce the impact of variants with low precision estimates, variants with effective sample size less than 0.8 of the median effective sample size for DGEs were filtered out.

    “To reduce the impact of variants with low precision estimates, we filtered out variants with effective sample size less than 0.8 of the median effective sample size for DGEs.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationcitation

    Under xAM, a relationship between cross-mate phenotype correlation and population effect genetic correlations will occur even in the absence of pleiotropy, and when there is some pleiotropy, it will be overestimated.

    “However, under xAM, a relationship between cross-mate phenotype correlation and population effect genetic correlations will occur even in the absence of pleiotropy, and when there is some pleiotropy, it will be overestimated7.”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [7] 2022, Cross-trait assortative mating is widespread and inflates genetic correlation estimates, Science

      Not checked in the free run.

64 of 64

  • The authors state this as an assumption.

    Under random mating, the population effect βl equals the sum of DGE δl and average NTC αl (βl=δl+αl), and deviations from this relationship are negligible for samples with low levels of structure.

    “Under random-mating, βl=δl+αl, where αl=(αpl+αml)/2 is the average NTC3 and deviations from this relationship are negligible for samples with low levels of structure3, such as those typically used in GWAS.”

  • The authors state this as an assumption.

    Under assortative mating at equilibrium, βl=δl/(1−rδ), where rδ is the correlation between parents’ DGE components, and thus αl=βl−δl=rδ/(1−rδ)δl.

    “To see this, consider that under AM at equilibrium, βl=δl/(1−rδ), where rδ is the correlation between parents’ DGE components23. Thus, αl=βl−δl=rδ/(1−rδ)δl.”

  • The authors state this as an assumption.

    The effects have expectation zero across the SNPs.

    “we assume the effects have expectation zero across the SNPs”

  • The authors state this as an assumption.

    The random bias term sl has mean zero and variance σs².

    “We assume that El[sl]=0 and Varl(sl)=σs².”

  • The authors state this as an assumption.

    Population effects can be modeled as scaled direct genetic effects plus a random bias term with mean zero.

    “In effect, we are modeling population effects as scaled DGEs plus a random bias term with mean zero.”

  • The authors state this as an assumption.

    Under random mating, the population effect βl equals δl + αl.

    “Under random-mating, the population effect, βl, is βl=δl+αl.”

  • The authors state this as an assumption.

    For samples with minimal structure, deviations from the relationship βl = δl + αl are negligible.

    “For samples with minimal structure — such as used in our meta-analysis — deviations from this relationship will be negligible3,20.”

  • The paper relies on this without support.

    DGE-based genetic correlations should be almost entirely free from the influence of xAM.

    “Since DGE based genetic correlations should be almost entirely free from the influence of xAM, a relationship between cross-mate phenotype correlations and DGE genetic correlations can be explained by pleitropy and univariate AM, which would not be expected to generate spurious genetic correlations between phenotypes without any pleiotropy.”
8 of 8

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.

  • Deflated population effects counted as both ten and nine

    Minor

    The Results identify ten phenotypes with deflated population effects, but a later paragraph refers to deflation observed for nine phenotypes.

  • Number of phenotypes with correlations below one differs

    Minor

    The Results state that the correlation was below one for 27 of 30 phenotypes, with the remaining three not estimable. The Discussion gives 26 phenotypes for the same finding.

2 of 2

Cites

Flagged first, then by citations

8 of the 62 references couldn’t be matched to a known paper.

Research works printed but not linked: 2

  1. 10.7554/elife.39324 · DOI · Paper page
  2. 10.1038/ng.3643 · DOI · Paper page

Cited by

Sorted by citations

Related

Authors
Tan, T., Jayashankar, H., Guan, J., Nehzati, S. M., Mir, M., Bennett, M., Agerbo, E., Ahlskog, R., de Andrade Anapaz, V. P., Benonisdottir, S., Boomsma, D. I., Campbell, A., Chabris, C. F., Chen, Z., China Kadoorie Biobank Collaborative Group, de Geus, E., Ehli, E. A., Elnahas, A. G., Estonian Biobank Research Team, Finngen, Ganna, A., Giannelis, A., Hakaste, L., Hayward, C., Hottenga, J.-J., Houmark, M. A., Kaprio, J., Latvala, A., Lee, J. J., Lehtovirta, M., Li, L., Study, L. C., Lin, K., Linnér, R. K., Lombardi, S., Martin, N. G., McGue, M., Medland, S. E., Metspalu, A., Mitchell, B. L., Ni, G., Nolte, I. M., Oetjens, M. T., Oskarsson, S., Palviainen, T., Prasad, R. B., Reigo, A., Reis, K., Sidorenko, J., Silventoinen, K., Snieder, H., Tuomi, T., Vilhjálmsson, B. J., Walters, R. G., Willoughby, E. A., Flint, J., Yengo, L., Visscher, P. M., Kong, A., Tucker-Drob, E. M., Border, R., Cesarini, D., Turley, P., Okbay, A., Benjamin, D. J., & Young, A. S.
Affiliations
  • National Bureau of Economic Research (US)
  • University of California, Los Angeles (US)
  • Aarhus University (DK)
  • Uppsala University (SE)
  • Institute for Molecular Medicine Finland FIMM, HILife, University of Helsinki, Helsinki, FIN-00140 Finland
  • Big Data Institute, Li Ka Shing Centre for Health Information Discovery, University of Oxford, Oxford OX3 7LF, UK
  • Vrije Universiteit Amsterdam (NL)
  • Western General Hospital (GB)
  • Geisinger Health System (US)
  • University of Oxford (GB)
  • Department of Biological Psychology, Vrije Universiteit Amsterdam, De Boelelaan 1105 1081 HV Amsterdam, The Netherlands
  • Avera McKennan Hospital & University Health Center (US)
  • University of Tartu (EE)
  • The University of Texas at Austin (US)
  • University of Helsinki (FI)
  • University of Minnesota (US)
  • Peking University (CN)
  • Leiden University (NL)
  • QIMR Berghofer Medical Research Institute (AU)
  • The University of Queensland (AU)
  • University Medical Center Groningen (NL)
  • Autism & Developmental Medicine Institute (US)
  • Broad Institute (US)
  • UCLA Department of Psychiatry and Biobehavioral Sciences, Los Angeles, CA, USA
  • Institute for Molecular Bioscience, University of Queensland, Australia
  • Department of Psychology, The University of Texas at Austin, 108 E. Dean Keeton Stop A8000, Austin, TX 78712, USA
  • University of Southern California (US)
  • UCLA Anderson School of Management, Los Angeles, CA, USA
ORCID
Venue
medRxiv
Publisher
openRxiv
Published
2024-10-04
DOI
10.1101/2024.10.01.24314703
PMC
PMC12633112
OpenAlex
W4403119557
Semantic Scholar
8d27ef42535e305cd79539723f7a5bb36940433e
Cited by
47
References
62
Citations by year
2026: 25 · 2025: 20 · 2024: 2
Type
posted content
Language
en
Open access
Licence
Locations
  • medRxiv (acceptedVersion, cc-by)
  • PubMed (publishedVersion)
  • Europe PMC (PubMed Central) (submittedVersion, cc-by)
  • University of Groningen research database (University of Groningen / Centre for Information Technology) (submittedVersion)
Funders
  • University of Southern California
  • Open Philanthropy Project
  • National Institutes of Health (R01AG042568, R00-AG062787)
  • University of California, Los Angeles
  • National Institute on Aging (R24AG065184, R00AG062787, R01AG083379)
  • National Institute of Mental Health (R01MH120219)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genetic Associations and Epidemiology
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genetic and phenotypic traits in livestock
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genetic Mapping and Diversity in Plants and Animals
Keywords
direct genetic effects, gene-environment correlation, population stratification, assortative mating, indirect genetic effects, pleiotropy, SNP heritability, polygenic prediction, phenotype prediction, genetic correlation, natural selection

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