Meta-analysis of the heritability of human traits based on fifty years of twin studies

Journal name:
Nature Genetics
Volume:
47,
Pages:
702–709
Year published:
DOI:
doi:10.1038/ng.3285
Received
Accepted
Published online

Abstract

Despite a century of research on complex traits in humans, the relative importance and specific nature of the influences of genes and environment on human traits remain controversial. We report a meta-analysis of twin correlations and reported variance components for 17,804 traits from 2,748 publications including 14,558,903 partly dependent twin pairs, virtually all published twin studies of complex traits. Estimates of heritability cluster strongly within functional domains, and across all traits the reported heritability is 49%. For a majority (69%) of traits, the observed twin correlations are consistent with a simple and parsimonious model where twin resemblance is solely due to additive genetic variation. The data are inconsistent with substantial influences from shared environment or non-additive genetic variation. This study provides the most comprehensive analysis of the causes of individual differences in human traits thus far and will guide future gene-mapping efforts. All the results can be visualized using the MaTCH webtool.

At a glance

Figures

  1. Distribution of the investigated traits in virtually all twin studies published between 1958 and 2012.
    Figure 1: Distribution of the investigated traits in virtually all twin studies published between 1958 and 2012.

    (a) The number of investigated traits in classical twin studies across all countries. (b) The average number of twin pairs included per study across countries. (c) The number of investigated traits according to functional trait domain and trait characteristic (inset). (d) Monozygotic and dizygotic twin correlations and reported estimates of h2 and c2 as a function of sample size. Contour lines indicate the density of the data in that region. The lines are colored by 'heat' from blue to red, indicating increasing data density.

  2. Twin correlations and heritabilities for all human traits studied.
    Figure 2: Twin correlations and heritabilities for all human traits studied.

    (a) Distribution of rMZ and rDZ estimates across the traits investigated in 2,748 twin studies published between 1958 and 2012. rMZ estimates are based on 9,568 traits and 2,563,628 partly dependent twin pairs; rDZ estimates are based on 5,220 traits and 2,606,252 partly dependent twin pairs (Table 1). (b) Relationship between rMZ and rDZ, using all 5,185 traits for which both were reported. (c) Random-effects meta-analytic estimates of twin correlations (top) and reported variance components (bottom) across all traits separately for four age cohorts. Error bars, standard errors. (d) Random-effects meta-analytic estimates of twin correlations (top) and reported variance components (bottom) across all traits, and within functional domains for which data on all correlations and variance components were available. Error bars, standard errors.

  3. Twin correlations for the top 20 most investigated specific traits by age and sex.
    Figure 3: Twin correlations for the top 20 most investigated specific traits by age and sex.

    Alc., alcohol; dis., disorders; depr., depressive; endocr., endocrine; imm., immunological; funct., functions; maint., maintenance; metab., metabolic; ment. beh., mental and behavioral; spec. personal., specific personality; temp. pers., temperament and personality; tob., tobacco; r, correlation; MZ, monozygotic twins; DZ, dizygotic twins; M, males; F, females; SS, same-sex pairs only; DOS, dizygotic opposite-sex pairs. Inclusion for the top 20 most investigated traits was conditional on the reporting of rMZ and rDZ. Empty cells denote insufficient information available to calculate weighted estimates; error bars, standard errors. We note that estimates and graphs for all specific traits are available from the online MaTCH webtool.

  4. Authorship co-occurrence matrix on 2,748 twin studies published between 1958 and 2012.
    Supplementary Fig. 1: Authorship co-occurrence matrix on 2,748 twin studies published between 1958 and 2012.

    Each colored cell represents two authors who appeared on the same paper; darker cells indicate authors that co-published more frequently. The filter of at least 25 papers per author was set for readability. The web application MaTCH has an interactive version of this matrix.

  5. Funnel plots across all traits for twin correlations and variance components.
    Supplementary Fig. 2: Funnel plots across all traits for twin correlations and variance components.

    Z, Z-converted correlation; MZ, monozygotic twins; DZ, dizygotic twins; DZSS, DZ same-sex twins; MZM, MZ male twins; MZF, MZ female twins; DZM, DZ male twins; DZF, DZ female twins; DOS, DZ opposite-sex twins; h2, heritability; c2, shared environment; h2 same sex; c2 same sex; h2 males; c2 males; h2 female; c2 females.

  6. Funnel plots for rMZ across the major trait domains.
    Supplementary Fig. 3: Funnel plots for rMZ across the major trait domains.

    The plots denote the relationship between the Z-transformed rMZ and its standard error. SE, standard error.

  7. Funnel plots for rDZ across the major trait domains.
    Supplementary Fig. 4: Funnel plots for rDZ across the major trait domains.

    The plots denote the relationship between the Z-transformed rDZ and its standard error. SE, standard error.

  8. Funnel plots for h2 across the major trait domains.
    Supplementary Fig. 5: Funnel plots for h2 across the major trait domains.

    The plots denote the relationship between the Z-transformed h2 and its standard error. SE, standard error.

  9. Funnel plots for c2 across the major trait domains.
    Supplementary Fig. 6: Funnel plots for c2 across the major trait domains.

    The plots denote the relationship between the Z-transformed c2 and its standard error. SE, standard error.

  10. Distribution of twin correlations and variance components in full and best models across all traits from 2,748 studies.
    Supplementary Fig. 7: Distribution of twin correlations and variance components in full and best models across all traits from 2,748 studies.

    rMZ, monozygotic twin correlation; rDZ, dizygotic twin correlation; rDZSS, DZ same-sex twin correlation; rMZM, MZ male twin correlation; rMZF, MZ female twin correlation; rDZM, DZ male twin correlation; rDZF, DZ female twin correlation; rDOS, DZ opposite-sex twin correlation; h2, heritability; c2, shared environment; h2 same sex;c2 same sex; h2 males; c2 males; h2 females; c2 females. “BEST” denotes estimates from the most parsimonious models per study. All other estimates are from “FULL” models.

  11. Distribution of differences between MZ and DZ correlations.
    Supplementary Fig. 8: Distribution of differences between MZ and DZ correlations.

    rMZ, monozygotic twin correlation; rDZ, dizygotic twin correlation; rDZSS, DZ same-sex twin correlation; rMZM, MZ male twin correlation; rMZF, MZ female twin correlation; rDZM, DZ male twin correlation; rDZF, DZ female twin correlation; rDOS, DZ opposite-sex twin correlation.

  12. The correlation between variance component estimates (h2 or c2) from maximum-likelihood (BEST or FULL models) (x axis) compared to the least-squares estimates (y axis).
    Supplementary Fig. 9: The correlation between variance component estimates (h2 or c2) from maximum-likelihood (BEST or FULL models) (x axis) compared to the least-squares estimates (y axis).
  13. The difference between variance components estimates from maximum-likelihood (BEST model) and least-squares (h2, left panel; c2, right panel) for given sample size.
    Supplementary Fig. 10: The difference between variance components estimates from maximum-likelihood (BEST model) and least-squares (h2, left panel; c2, right panel) for given sample size.
  14. Scatterplots of all MZ versus DZ correlations.
    Supplementary Fig. 11: Scatterplots of all MZ versus DZ correlations.

    Contour lines indicate the density of the data in that region. The lines are ‘heat’ colored from blue to red, indicating increasing data density.

  15. QQ plots of the [chi]2 test statistics for testing the null hypothesis that 2(rMZ - rDZ) = 0 and 2rDZ - rMZ = 0 and relationship with sample size.
    Supplementary Fig. 12: QQ plots of the χ2 test statistics for testing the null hypothesis that 2(rMZrDZ) = 0 and 2rDZrMZ = 0 and relationship with sample size.

    (a) The deviation from the null hypotheses is quantified with the inflation λ in the QQ plots. (b) Effects as a function of sample size.

References

  1. Moore, J.H. Analysis of gene-gene interactions. Curr. Protoc. Hum. Genet. Chapter 1, Unit 1.14 (2004).
  2. Hill, W.G., Goddard, M.E. & Visscher, P.M. Data and theory point to mainly additive genetic variance for complex traits. PLoS Genet. 4, e1000008 (2008).
  3. Traynor, B.J. & Singleton, A.B. Nature versus nurture: death of a dogma, and the road ahead. Neuron 68, 196200 (2010).
  4. Zuk, O., Hechter, E., Sunyaev, S.R. & Lander, E.S. The mystery of missing heritability: genetic interactions create phantom heritability. Proc. Natl. Acad. Sci. USA 109, 11931198 (2012).
  5. Phillips, P.C. Epistasis—the essential role of gene interactions in the structure and evolution of genetic systems. Nat. Rev. Genet. 9, 855867 (2008).
  6. Visscher, P.M., Brown, M.A., McCarthy, M.I. & Yang, J. Five years of GWAS discovery. Am. J. Hum. Genet. 90, 724 (2012).
  7. Manolio, T.A. et al. Finding the missing heritability of complex diseases. Nature 461, 747753 (2009).
  8. Stranger, B.E., Stahl, E.A. & Raj, T. Progress and promise of genome-wide association studies for human complex trait genetics. Genetics 187, 367383 (2011).
  9. Maher, B. Personal genomes: the case of the missing heritability. Nature 456, 1821 (2008).
  10. Eichler, E.E. et al. Missing heritability and strategies for finding the underlying causes of complex disease. Nat. Rev. Genet. 11, 446450 (2010).
  11. Nelson, R.M., Pettersson, M.E. & Carlborg, Ö. A century after Fisher: time for a new paradigm in quantitative genetics. Trends Genet. 29, 669676 (2013).
  12. Barker, J.S. Inter-locus interactions: a review of experimental evidence. Theor. Popul. Biol. 16, 323346 (1979).
  13. Cockerham, C.C. An extension of the concept of partitioning hereditary variance for analysis of covariances among relatives when epistasis is present. Genetics 39, 859882 (1954).
  14. Cockerham, C.C. in Statistical Genetics and Plant Breeding 5394 (Nat. Acad. Sci. Nat. Res. Council Publ., 1963).
  15. Kempthorne, O. On the covariances between relatives under selfing with general epistacy. Proc. R. Soc. Lond. B Biol. Sci. 145, 100108 (1956).
  16. Crow, J.F. & Kimura, M. An Introduction To Population Genetics Theory (Harper and Row, 1970).
  17. Carlborg, O. & Haley, C.S. Epistasis: too often neglected in complex trait studies? Nat. Rev. Genet. 5, 618625 (2004).
  18. Falconer, D.S. & Mackay, T.F.C. Quantitative Genetics (Longman Group, 1996).
  19. Lynch, M. & Walsch, B. Genetics and Analysis of Quantitative Traits (Sinauer Associates, 1998).
  20. Begg, C.B. & Mazumdar, M. Operating characteristics of a rank correlation test for publication bias. Biometrics 50, 10881101 (1994).
  21. Egger, M., Davey Smith, G., Schneider, M. & Minder, C. Bias in meta-analysis detected by a simple, graphical test. Br. Med. J. 315, 629634 (1997).
  22. Rosenthal, R. The file drawer problem and tolerance for null results. Psychol. Bull. 86, 91106 (1979).
  23. Schulze, R. Meta-Analysis: A Comparison Of Approaches (Hogrefe & Huber, 2004).
  24. Drasgow, F. in Encyclopedia of Statistical Sciences (eds. Kotz, S., Read, C.B., Balakrishnan, N. & Vidakovic, B.) Vol. 7, 6874 (John Wiley & Sons, 2006).
  25. Olsson, U. Maximum likelihood estimation of the polychoric correlation coefficient. Psychometrika 44, 443460 (1979).
  26. Falconer, D.S. The inheritance of liability to certain diseases, estimated from the incidence among relatives. Ann. Hum. Genet. 29, 5176 (1965).
  27. Smith, C. Concordance in twins: methods and interpretation. Am. J. Hum. Genet. 26, 454466 (1974).
  28. Jiang, H. & Doerge, R.W. Estimating the proportion of true null hypotheses for multiple comparisons. Cancer Inform. 6, 2532 (2008).
  29. Storey, J.D. & Tibshirani, R. Statistical significance for genomewide studies. Proc. Natl. Acad. Sci. USA 100, 94409445 (2003).
  30. Bulik-Sullivan, B.K. & Sullivan, P.F. The authorship network of genome-wide association studies. Nat. Genet. 44, 113 (2012).
  31. Blondel, V., Guillaume, J., Lambiotte, R. & Lefebvre, E. Fast unfolding of community hierarchies in large networks. J. Stat. Mech. 10, P10008 (2008).

Download references

Author information

  1. These authors contributed equally to this work.

    • Tinca J C Polderman &
    • Beben Benyamin
  2. These authors jointly supervised this work.

    • Peter M Visscher &
    • Danielle Posthuma

Affiliations

  1. Department of Complex Trait Genetics, VU University, Center for Neurogenomics and Cognitive Research, Amsterdam, the Netherlands.

    • Tinca J C Polderman,
    • Christiaan A de Leeuw &
    • Danielle Posthuma
  2. Queensland Brain Institute, University of Queensland, Brisbane, Queensland, Australia.

    • Beben Benyamin &
    • Peter M Visscher
  3. Institute for Computing and Information Sciences, Radboud University Nijmegen, Nijmegen, the Netherlands.

    • Christiaan A de Leeuw
  4. Center for Psychiatric Genomics, Department of Genetics, University of North Carolina, Chapel Hill, North Carolina, USA.

    • Patrick F Sullivan
  5. Department of Psychiatry, University of North Carolina, Chapel Hill, North Carolina, USA.

    • Patrick F Sullivan
  6. Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden.

    • Patrick F Sullivan
  7. Faculty of Sciences, VU University, Amsterdam, the Netherlands.

    • Arjen van Bochoven
  8. University of Queensland Diamantina Institute, Translational Research Institute, Brisbane, Queensland, Australia.

    • Peter M Visscher
  9. Department of Clinical Genetics, VU University Medical Center, Neuroscience Campus Amsterdam, Amsterdam, the Netherlands.

    • Danielle Posthuma

Contributions

D.P., B.B., P.F.S. and P.M.V. performed the analyses. D.P. conceived the study. D.P., T.J.C.P. and P.M.V. designed the study. T.J.C.P. and D.P. collected and entered the data. D.P. and P.F.S. categorized traits according to standard classifications. A.v.B. and C.A.d.L. checked data entries, and checked and wrote statistical scripts. A.v.B. designed and programmed the webtool. D.P., T.J.C.P. and P.M.V. wrote the manuscript. All authors discussed the results and commented on the manuscript.

Competing financial interests

The authors declare no competing financial interests.

Corresponding author

Correspondence to:

Author details

Supplementary information

Supplementary Figures

  1. Supplementary Figure 1: Authorship co-occurrence matrix on 2,748 twin studies published between 1958 and 2012. (368 KB)

    Each colored cell represents two authors who appeared on the same paper; darker cells indicate authors that co-published more frequently. The filter of at least 25 papers per author was set for readability. The web application MaTCH has an interactive version of this matrix.

  2. Supplementary Figure 2: Funnel plots across all traits for twin correlations and variance components. (380 KB)

    Z, Z-converted correlation; MZ, monozygotic twins; DZ, dizygotic twins; DZSS, DZ same-sex twins; MZM, MZ male twins; MZF, MZ female twins; DZM, DZ male twins; DZF, DZ female twins; DOS, DZ opposite-sex twins; h2, heritability; c2, shared environment; h2 same sex; c2 same sex; h2 males; c2 males; h2 female; c2 females.

  3. Supplementary Figure 3: Funnel plots for rMZ across the major trait domains. (375 KB)

    The plots denote the relationship between the Z-transformed rMZ and its standard error. SE, standard error.

  4. Supplementary Figure 4: Funnel plots for rDZ across the major trait domains. (366 KB)

    The plots denote the relationship between the Z-transformed rDZ and its standard error. SE, standard error.

  5. Supplementary Figure 5: Funnel plots for h2 across the major trait domains. (346 KB)

    The plots denote the relationship between the Z-transformed h2 and its standard error. SE, standard error.

  6. Supplementary Figure 6: Funnel plots for c2 across the major trait domains. (343 KB)

    The plots denote the relationship between the Z-transformed c2 and its standard error. SE, standard error.

  7. Supplementary Figure 7: Distribution of twin correlations and variance components in full and best models across all traits from 2,748 studies. (379 KB)

    rMZ, monozygotic twin correlation; rDZ, dizygotic twin correlation; rDZSS, DZ same-sex twin correlation; rMZM, MZ male twin correlation; rMZF, MZ female twin correlation; rDZM, DZ male twin correlation; rDZF, DZ female twin correlation; rDOS, DZ opposite-sex twin correlation; h2, heritability; c2, shared environment; h2 same sex;c2 same sex; h2 males; c2 males; h2 females; c2 females. “BEST” denotes estimates from the most parsimonious models per study. All other estimates are from “FULL” models.

  8. Supplementary Figure 8: Distribution of differences between MZ and DZ correlations. (209 KB)

    rMZ, monozygotic twin correlation; rDZ, dizygotic twin correlation; rDZSS, DZ same-sex twin correlation; rMZM, MZ male twin correlation; rMZF, MZ female twin correlation; rDZM, DZ male twin correlation; rDZF, DZ female twin correlation; rDOS, DZ opposite-sex twin correlation.

  9. Supplementary Figure 9: The correlation between variance component estimates (h2 or c2) from maximum-likelihood (BEST or FULL models) (x axis) compared to the least-squares estimates (y axis). (476 KB)
  10. Supplementary Figure 10: The difference between variance components estimates from maximum-likelihood (BEST model) and least-squares (h2, left panel; c2, right panel) for given sample size. (169 KB)
  11. Supplementary Figure 11: Scatterplots of all MZ versus DZ correlations. (517 KB)

    Contour lines indicate the density of the data in that region. The lines are ‘heat’ colored from blue to red, indicating increasing data density.

  12. Supplementary Figure 12: QQ plots of the χ2 test statistics for testing the null hypothesis that 2(rMZrDZ) = 0 and 2rDZrMZ = 0 and relationship with sample size. (222 KB)

    (a) The deviation from the null hypotheses is quantified with the inflation λ in the QQ plots. (b) Effects as a function of sample size.

PDF files

  1. Supplementary Text and Figures (4,901 KB)

    Supplementary Figures 1–12, Supplementary Note and Supplementary Tables 1–19, 22–24 and 26–31.

Excel files

  1. Supplementary Tables 20, 21, 25, 32 and 33. (624 KB)

    Supplementary Tables 20, 21, 25, 32 and 33.

Additional data