Ni Hms 357611

Most Reported Genetic Associations with General IntelligenceAre Probably False Positives

Christopher F. Chabris1,*, Benjamin M. Hebert2, Daniel J. Benjamin3, Jonathan P.Beauchamp2, David Cesarini4,5, Matthijs J.H.M. van der Loos6, Magnus Johannesson7,Patrik K.E. Magnusson8, Paul Lichtenstein8, Craig S. Atwood9,10, Jeremy Freese11, TaissaS. Hauser12, Robert M. Hauser12,13, Nicholas A. Christakis14,15, and David Laibson21Department of Psychology, Union College2Department of Economics, Harvard University3Department of Economics, Cornell University4Department of Economics, New York University5IFN-Research Institute for Industrial Economics, Stockholm6Erasmus School of Economics, Rotterdam7Stockholm School of Economics8Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm9Department of Medicine, University of Wisconsin-Madison Medical School10Veterans Administration Hospital, Madison, Wisconsin11Department of Sociology, Northwestern University12Center for Demography of Health and Aging, University of Wisconsin-Madison13Department of Sociology, University of Wisconsin-Madison14Department of Sociology, Harvard University15Department of Medicine, Harvard Medical School

AbstractGeneral intelligence (g) and virtually all other behavioral traits are heritable. Associations betweeng and specific single-nucleotide polymorphisms (SNPs) in several candidate genes involved inbrain function have been reported. We sought to replicate published associations between 12specific genetic variants and g using three independent, longitudinal datasets of 5571, 1759, and2441 well-characterized individuals. Of 32 independent tests across all three datasets, only onewas nominally significant at the p < .05 level. By contrast, power analyses showed that we shouldhave expected 1015 significant associations, given reasonable assumptions for genotype effectsizes. As positive controls, we confirmed accepted genetic associations for Alzheimer disease andbody mass index, and we used SNP-based relatedness calculations to replicate estimates that abouthalf of the variance in g is accounted for by common genetic variation among individuals. Weconclude that different approaches than candidate genes are needed in the molecular genetics ofpsychology and social science.

*Address correspondence to: Christopher F. Chabris, Department of Psychology, Union College, 807 Union Street, Schenectady, NY12308, [email protected].

NIH Public AccessAuthor ManuscriptPsychol Sci. Author manuscript; available in PMC 2012 November 15.

Published in final edited form as:Psychol Sci. 2012 November 1; 23(11): 13141323. doi:10.1177/0956797611435528.

$waterm

ark-text$w

atermark-text

$waterm

ark-text

Genetics has great potential to contribute to psychology and the social sciences for at leasttwo reasons. First, as human behavior involves the operation of the brain, understanding thegenes whose expression affects the development and physiology of the brain can further ourunderstanding of the causal chains connecting evolution, brain, and behavior. Second,because genetic differences can potentially account for some of the differences amongindividuals in cognitive function, behavior, and outcomes, any effort to paint a picture of thestructure of human differences that does not incorporate genetics will be incomplete andpossibly misleading.

Within psychology, the genetics of behavior has been explored since the earliest twin studies(for an overview, see Plomin et al., 2008). Behavior genetic studies have shown that nearlyall human behavioral traits are heritable (Turkheimer, 2000). If a trait is heritable in thegeneral population, thenwith sufficiently large samplesit should be possible in principleto identify molecular genetic variants that are associated with the trait. General cognitiveability, or g (Spearman, 1904; Neisser et al., 1996; Plomin et al., 2008) is among the mostheritable behavioral traits. Estimates of broad heritability as high as 0.80 have been reportedfor adult IQ measured in modern Western populations (Bouchard, 1998). Although the exactfigures have been the topic of much debate, the claim that IQ is at least moderately heritableis widely accepted. IQ may in fact be similar in heritability to the physical trait of height(Weedon & Frayling, 2008). Both height and IQ are genetically complex because thesetraits are influenced by many genes, acting in concert with environmental factors, rather thanbeing determined by single genetic variants. Finding genes associated with g could yieldmany potential benefits, among them new insights into the biology of cognition and itsdisorders. Such discoveries might suggest new therapeutic targets or pathways for potentialtreatments to improve cognition. Uncovering the molecular genetics of other traits andabilities, such as personality, time and risk preferences, and social skills could have similarlybeneficial consequences (Benjamin et al., 2007).

By now there is a large literature of candidate gene studies showing associations betweenmany single-nucleotide polymorphisms (SNPs) and g.1 Payton (2009) produced acomprehensive review of these studies. Here we report the results of a series of attempts toreplicate as many published SNP-g associations as possible, using data from threeindependent, large, well-characterized, longitudinal samples. We begin, in Study 1, with theWisconsin Longitudinal Study (WLS; www.ssc.wisc.edu/wlsresearch), which includesgenotypes for 13 of the SNPs reported by Payton (2009) to have published associations withg. These 13 SNPs are located in or near 10 different genes. In followup studies, we test 10 ofthe original 13 SNPs that were available in two other samples. In Study 2, we use theFramingham Heart Study (FHS; www.framinghamheartstudy.org), and in Study 3, we usedata from the Swedish Twin Registry (STR; ki.se/ki/jsp/polopoly.jsp?d=9610&l=en) toexamine associations with g. Although we analyzed them separately, the combined samplesize of these datasets is almost 10,000 individuals, which gives us considerable statisticalpower.

If the published SNP-g associations we examined were true positives in the generalpopulation, then we would expect many of them to replicate at the 5% significance level inour much larger datasets. However, if the literature on SNP-g associations consists mostly offalse positives, then we would expect very few replications in our data. Such a result wouldnot likely be due to differences in the methods used to estimate g in the various datasets

1Because our goal is to replicate the results of published candidate gene studies of g, we do not consider the results of genome-wideassociation studies (GWAS), none of which have yet identified replicable SNPs that meet conventional thresholds for significantassociations with g (e.g., Butcher et al., 2008; Davies et al., 2011; Seshadri et al., 2007).

Chabris et al. Page 2

Psychol Sci. Author manuscript; available in PMC 2012 November 15.

$waterm

ark-text$w

atermark-text

$waterm

ark-text

under comparison, since g is consistently measured by a wide variety of well-designed tests(Ree & Earles, 1991).

Study 1Method

The Wisconsin Longitudinal Study (WLS) is based on a one-third sample of all Spring 1957Wisconsin high school graduates (initial N = 10,317). A randomly selected sibling of asubsample of these graduates was enrolled in 1977 and a randomly selected sibling of eachremaining graduate was enrolled in 1993 (N = 5,219). g was measured by the Henmon-Nelson Test of Mental Ability (Lamke & Nelson, 1957) for both graduate and siblingsample members when they were in the 11th grade, and obtained from administrativerecords. Percentile scores were rescaled to the conventional IQ metric of a mean of 100 andstandard deviation of 15.

We studied all 13 SNPs that were both previously associated with g according to Paytonsreview (2009) and included among the 90 SNPs genotyped in the WLS. They were:rs429358 and rs7412 in APOE (these SNPs define the e2/e3/e4 haplotype associated withAlzheimer disease), rs6265 in BDNF, rs2061174 in CHRM2, rs8191992 in CHRM2/CHRNA4, rs4680 in COMT, rs17571 in CTSD, rs821616 in DISC1, rs1800497 in DRD2/ANKK1, rs1018381 in DTNBP1, rs760761 in DTNBP1, rs363050 in SNAP25, andrs2760118 in SSADH (aka ALDH5A1).

Of the 6,908 WLS respondents with adequate covariate and genotype data, 5,571 had datafor g and for all 13 SNPs previously associated with g. All 13 SNP genotypes were inHardy-Weinberg equilibrium, and their frequencies matched those reported in the literaturefor European samples.

As positive controls for global problems in genotyping or data quality, we considered twogenotype-phenotype associations that have been established and accepted: APOE andAlzheimers disease (AD), and FTO and body mass index (BMI). We tested the two SNPs inthe APOE gene that define the common, well-established risk haplotype for AD (e2/e3/e4)for association with parental AD status. As expected, subjects with at least one e4 allelewere more likely to report having a parent with AD than were subjects with no e4 alleles (p< .0001). Likewise, the previously reported and replicated association between the numberof C alleles of SNP rs1421085 in FTO and body mass index (Tung & Yeo, 2011) wasobserved here (self-reported BMIs of 27.5, 27.9, and 28.3 for 0, 1, and 2 C alleles,respectively; p < .001).

For each SNP we adopted a standard linear allele dosage model; we regressed Henmon-Nelson IQ on the number minor (less frequent) alleles. However, for the two APOE SNPs,we instead analyzed a dummy variable indicating the presence of at least one e4 allele, sincethis allele is defined by a haplotype of these two SNPs and is the genotype previouslystudied in conjunction with g (and AD). All of our analyses controlled for graduate/siblingstatus, age, gender, and the interactions of these factors, as well as the first three principalcomponents of the genetic data from the full set of 90 genotyped SNPs (to account forpossible population stratification). [For additional Methods details, see Supporting OnlineMaterial.]

ResultsTable 1 displays the results of this analysis. None of the 12 genotypes (11 SNPs and theAPOE e4 variable) were significantly associated with g (p .10 in all cases). We conductedan omnibus F-test for all 11 SNPs and the APOE dummy combined in a single regression,



$waterm

ark-text$w

atermark-text

$waterm

ark-text

and could not reject the null hypothesis that all of the SNPs jointly have zero effect on g (F =0.88, p = .56). We calculated the statistical power associated with this omnibus test andfound that if, in aggregate, our 12 genotypic predictors jointly explain at least 0.52% of thevariance of g, the F-test should reject the null hypothesis more than 99% of the time. Thethresholds associated with 80% and 95% rejection are 0.26% and 0.39% of the variance,respectively.

A recent meta-analysis (Barnett et al., 2008) suggests that the well-researched Val158Metpolymorphism in COMT (rs4680) may explain around 0.10% of the variance of g. Thisestimate is likely to still be biased upward, because it assumes no publication bias orwinners curse is affecting the literature on this association. If we make the reasonableassumption that our SNPs, which are mostly distributed across several chromosomes, areindependent, these results imply that the average effect size of the 12 genotypic predictors(which include rs4680) must be even smaller than 0.05% of the variance (because 0.52%/12= 0.043%), although we cannot rule out the possibility that most are zero and a few exceed0.10%. These effect sizes are smalle.g., 0.05% of the variance is about 0.45 IQ points fora SNP whose minor allele frequency is close to 50%, as in the case of rs4680and muchlower than the effect sizes reported for the SNPs in the initial publications of their gassociations. From these calculations, we conclude that our analysis has a high level ofstatistical power for effect sizes of meaningful magnitude.

Study 2Method

In study 2, we attempted to repeat the same analysis as closely as possible with data fromthe Initial and Offspring cohorts of the Framingham Heart Study (FHS), which hastracked residents of Framingham, Massachusetts, and their descendants since the 1940s.Dawber et al. (1951) and Feinleib et al. (1975) provide more details on these two cohorts ofthe FHS. Our dataset included 1759 individuals, of whom 45.4% were male. Participantsranged from 40100 years in age when they completed a battery of cognitive tests as part ofa neuropsychological component of the FHS. These tests included Trails A and B, WRAT-Reading, Boston Naming, WAIS Similarities, Hooper Visual Organization, WMS VisualReproductions, and WMS Logical Memory (for more information see Seshadri et al., 2007).

To estimate general cognitive ability, we first conducted a principal component analysis onthe cognitive test data (controlling for sex, birth year, and cohort); the first componentaccounted for 45.6% of the variance in test performance, consistent with the normal patternin studies of general intelligence (Chabris, 2007). For each individual in the full sample, gwas then defined as the subjects score on the first principal component. Finally, the scoreswere normalized to have mean 100 and variance 15.

Ten of the 13 WLS SNPs were available in a set of genotypes previously imputed. (The twoSNPs in APOE, rs7412 and rs429358, and one in SNAP25, rs363050, were not available.)[For additional Methods details, see Supporting Online Material.]

ResultsTests of association with each SNP were conducted using the standard linear allele dosagemodel as with the WLS data, with the standard errors clustered by extended family. Table 2displays the results. Nine of the ten SNPs were not significantly associated with g, p .10 inall cases. We also did an omnibus F-test for all 10 SNPs in a single regression, and could notreject the null hypothesis that all of the SNPs have zero joint effect on g (F = 0.85, p = .58).



$waterm

ark-text$w

atermark-text

$waterm

ark-text

One SNP, rs2760118 in SSADH (also known as ALDH5A1), exhibited a nominallysignificant association with g (t = 2.01, p = .04), but this association did not survive aBonferroni correction. The mean g values (transformed to the IQ scale) by genotype for thisSNP were 98.3, 99.7, and 100.6 for genotypes TT, TC, and CC respectively. This SSADHpolymorphism was first reported to be associated with g by Plomin et al. (2004), withdirectionality the same as in our FHS data, and some rare SSADH mutations are robustlyassociated with mental retardation and seizures via a well-known biological pathwayinvolving the metabolism of the inhibitory neurotransmitter GABA (Pearl et al., 2009).

Benjamin et al. (2011) reported that rs2760118 was associated with educational attainmentin an Icelandic sample; the association was replicated in a second Icelandic sample andappeared to be partially mediated by an association between SSADH and cognitive functionin both samples. However, the same study reported that the association between rs2760118and education did not replicate in three other datasets (WLS, FHS, and a control group fromthe NIMH Swedish Schizophrenia Study). It is possible that this SSADH SNP has a true, butsmall, effect on g that is only observed in some studies and/or under some environmentalconditions.

Study 3Method

To verify that the results of Study 1 and Study 2 were not artifacts of any factors specific tothe WLS and FHS datasets, we repeated the analysis in a sample of recently genotypedSwedish twins born between 1936 and 1958. The subjects were all participants in the SALTsurvey (see Lichtenstein et al., 2002, for a description of the sample); 10,946 of the SALTrespondents have been genotyped.

Until recently, Swedish men were required by law to participate in military conscription ator around the age of 18, and a test of cognitive ability was part of the screening process.Since performance on the test influenced a recruits ultimate position in the military,incentives to perform well on the test were strong. The recruits studied here took either fouror five cognitive tests, depending on their cohort; the tests used included measures ofproblem solving, concept discrimination, technical comprehension, multiplication, andmechanical or spatial ability. Carlstedt (2000) describes the batteries in more detail andreports evidence that they provide good measures of g. Since there are minor variationsacross years in the specific questions asked, we conducted a separate principal componentanalysis of the subtests for each birth year. For each individual in the full sample, g was thendefined as the subjects score on the first principal component. As with the WLS and FHS,we normalized the scores to have mean 100 and standard deviation 15.

Ten of the original 12 WLS genotypes were available in the imputed data, exactly the sameSNPs as in the Framingham data. Tests of association with each SNP were conducted usinglinear regression analysis. The sample is exclusively male, g was estimated separately foreach cohort defined by birth year, and there is no meaningful variation in the age at whichthe men take the test (as conscription nearly always occurs around the age of 18), so age andsex were not included as covariates, but the first ten principal components of genetic datawere included. The final sample includes 2,441 individuals for whom genetic and IQ testdata is available: 811 twins without a co-twin in the sample, 418 complete MZ pairs, and397 complete DZ pairs. [For additional Methods details, see Supporting Online Material.]

ResultsTests of association with each SNP were conducted using the same approach as with theWLS and FHS data; Table 3 displays the results. The association that came closest to



$waterm

ark-text$w

atermark-text

$waterm

ark-text

significance is with SNP rs2760118 in SSADH (t = 1.58, p = .11), the same SNP that wasnominally significant in the FHS sample. However, the direction of the association here isthe opposite of what was observed in the FHS. In STR the mean IQ scores were 99.2, 100.4,and 100.9 for genotypes CC, TC and TT respectively. The omnibus F-test for all 10 SNPs ina single regression fails to reject the null hypothesis that the SNPs jointly have zero effect ong (F = 0.89, p = .55).

DiscussionWe attempted to replicate published associations of 12 specific genotypes with measures ofgeneral cognitive ability in three large, well-characterized longitudinal datasets. In theWisconsin Longitudinal Study, none of the 12 genotypes were significantly associated withg. In the Framingham Heart Study, 9 of the 10 SNPs we were able to test were also notassociated with g. The only nominally significant association involved SNP rs27660118. Inthe Swedish Twin Registry sample, none of the 10 available SNPs were significantlyassociated with g. The association between rs27660118 and IQ approached significance(before correction for multiple hypothesis testing), but the effect was opposite to thatobserved in the FHS sample.

There have been previous failures to replicate published candidate gene studies of g (e.g.,Houlihan et al., 2009). Our research is distinguished by a large combined sample of almost10,000 individuals across three independent samples and an attempt to replicate allpublished associations for which we had available data in all three datasets. The contrastbetween the outcome expected from the literature and the outcome we actually observed inour investigation is striking. Assuming that the SNPs are independently distributed, underthe null hypothesis that every genotype we examined was unrelated to g, the expectednumber of significant associations at the 5% level is 1.6 (out of our 32 total tests). Weobserved exactly one nominally significant association, slightly less than would be expectedby chance alone.

This result is not likely due to lack of statistical power. Figure 1 shows the number ofsignificant associations expected under a range of alternative hypotheses for the size of eachgenotypes effect on g, with the effect size ranging from R2 = 0% to 1% of the variance. Forexample, had all of the associations that we tested been true positives in the population withan effect size of R2 = 0.1%the effect size that Barnett et al.s (2008) meta-analysis foundfor COMTthen the expected number of significant (p < .05) associations would have beenapproximately 14.7 in the 32 tests we did: the sum of 8.7 out of 12 in the WLS data, 2.6 outof 10 in the FHS data, and 3.4 out of 10 in the STR data.2 Even after accountingconservatively for the genetic relatedness of some participants (siblings in the WLS, familymembers in the FHS, and twins in the STR), we would still expect 10.6 total associations, orten times more than we found. And an effect of one tenth of one percent of the phenotypicvariance is tiny; as Figure 1 shows, assuming anything larger increases the power of ourstudies, and thus the divergence between the number of associations expected and thenumber we observed.

To assess the potential size of any effects on g of the genotypes we examined, we meta-analyzed the results from our three studies. Figure 2 shows that the pooled estimates aresufficiently precise to rule out anything but very small effects. Even the widest 95%

2For our full samples, power at R2 = 0.1% (the dotted line in Figure 1) is .72 for WLS, .26 for FHS, and .34 for STR. Assumingindependence across SNPsa reasonable assumption since almost all of the SNPs are far apart or on separate chromosomestheexpected number of significant associations in a sample is the power times the number of SNPs tested. (For the smaller samples ofunrelated individuals, the power values are .56, .13, and .25 respectively.)



$waterm

ark-text$w

atermark-text

$waterm

ark-text

confidence interval excludes effect sizes larger than 1.3 IQ points, which is less than onetenth of a standard deviation. Most of the effects are estimated with considerably greaterprecision.

The failure thus far to find genes associated with g does not mean that g has no geneticcomponent. Davies et al. (2011) used data from five different genome-wide associationstudies (GWAS) and failed to identify any individual markers robustly associated withcrystalized or fluid intelligence. They then applied a recently developed method (Yang et al.,2010; Visscher et al., 2010) for testing the cumulative effects of all the genotyped SNPs. Inessence, this method calculates the overall genetic similarity between each pair ofindividuals in a sample and then correlates this genetic similarity with phenotypic similarityacross all pairs. Following Yang et al. (2010), we dropped one twin per pair, and thenestimated all pairwise genetic relationships in the resulting sample. We then droppedindividuals whose relatedness exceeded .025, just as in Davies et al. (2011). Davies et al.reported that the ~550,000 SNPs in their data could jointly explain 40% of the variation incrystalized g (N = 3,254) and 51% of the variation in fluid g (N = 3,181). We applied thesame procedure to the STR sample from Study 3 and estimated that the ~630,000 SNPs inour data jointly account for 47% of the variance in g (p < .02), confirming the Davies et al.(2011) findings in an independent sample. These and our other results, together with thefailure of whole-genome association studies of g to date, are consistent with generalintelligence being a highly polygenic trait on which common genetic variants individuallyhave only small effects.

ConclusionA consensus is emerging that most published results from candidate gene studies thatoriginally used small samples fail to replicate (Siontis et al., 2010; Ioannidis et al., 2011; cf.Ioannidis, 2005). There are several possible reasons, none of them mutually exclusive, forthis state of affairs. Failure to replicate can be attributed to lack of statistical power in thereplication sample, but this is unlikely to apply here, because our replication samples aremuch larger than the samples used in the original studies or in most candidate gene studies.Genetic associations may also fail to replicate when the identified variants are not the onesthat cause the trait variation, but are correlated with the true causal variants, with differentpatterns of linkage disequilibrium in different samples. Patterns of failed replication mayalso arise due to differing effects of genes on traits across environments.

By far the most plausible explanation in our case, however, is that the original studies weseek to replicate did not have sufficient sample sizesand not because of any error indesign or execution. Expectations that individual SNPs might have large effects on g, whichcould be detected with small samples, seemed reasonable before genome-wide associationstudies were possible, and when genotyping was orders of magnitude more expensive than itis now. But if the true effect sizes of common variants are small, as now seems clear, thenthe early studies whose results we have failed to replicate were inadvertently underpowered.Bayesian calculations imply that results reported from underpowered studies, even ifstatistically significant, are likely to be false positives (e.g., Ioannidis, 2005; Benjamin,2010).

The results reported here illustrate for g the problem of missing heritability (Manolio etal., 2009), which is the failureso farto find specific molecular variants that account forthe substantial genetic influences identified by twin and family studies of medical andpsychiatric phenotypes. For comparison, height is approximately 90% heritable in Westernpopulations, but so far no common variants contributing more than 0.5cm per allele havebeen discovered, and the set of 180 height-associated SNPs identified by the most



$waterm

ark-text$w

atermark-text

$waterm

ark-text

comprehensive meta-analysis only explains about 10% of the population phenotypicvariance (Lango Allen et al., 2010). We suspect that our results for g are not an isolatedexception, but instead illustrative of a larger pattern in the genetics of cognition and socialscience (Beauchamp et al., 2011; Benjamin, 2010). There are several possible explanationsfor the missing heritability. One view is that common variants explain much of the heritablevariation but that the individual effects are so small that enormous samples are required toreliably detect them (Visscher, 2008; Visscher et al., 2008). An alternative view is that muchof the heritable variation comes from rare, perhaps structural, genetic variants with modestto large effect sizes (Dickson et al., 2010; Yeo et al., 2011).

At the time most of the results we have attempted to replicate were obtained, candidate genestudies of complex traits were commonplace in medical genetics research. Such studies arenow rarely published in leading journals. Our results add IQ to the list of phenotypes thatmust be approached with great caution when evaluating published molecular geneticassociations. In our view, excitement over the value of behavioral and molecular geneticstudies in the social sciences should be temperedas it has been in the medical sciencesby an appreciation that for complex phenotypes, individual common genetic variants of thesort assayed by SNP microarrays are likely to have very small effects. Associations ofcandidate genes with psychological and other social science traits should be viewed astentative until they have been replicated in multiple large samples. Doing otherwise mayhamper scientific progress by proliferating potentially false positive results, which may theninfluence the research agendas of other scientists who do not appreciate that the associationsthey take as a starting point for their efforts may not be real. And the dissemination of falseresults to the public risks creating an incorrect perception about the state of knowledge in thefield, especially the existence of genes described as being for traits on the basis ofunintentionally inflated estimates of effect size and statistical significance.

We think that a profitable way forward for molecular genetic investigations in social scienceis to follow the lead of medical genetics researchers, who have formed internationalconsortia that include as many large studies with genomic and (harmonized) phenotypic dataas possible. A plausible sample size of 100,000 individuals has statistical power of 80% todiscover genetic variants accounting for as little as 0.04% of the variance in a trait at agenome-wide significance level of p < 5 108. With sufficient power, it will also befeasible to study gene-gene interactions (e.g., Roetker et al., 2011), which may account formore of the variance in complex phenotypes than individual SNPs considered in isolation.

Finally, we emphasize that the negative results reported here should not detract fromresearch into the behavioral and molecular genetics of g and other social science traits, butrather point the way to study designs that are more likely to yield robust knowledge.

Supplementary MaterialRefer to Web version on PubMed Central for supplementary material.

AcknowledgmentsThis research was supported by the NIA (grants P01AG005842 and T32-AG000186-23). The Swedish TwinRegistry is supported by the Swedish Department of Higher Education, the European Commission (grant QLG2-CT-2002-01254), the Swedish Research Council, the Swedish Foundation for Strategic Research, the Jan Wallanderand Tom Hedelius Foundation, and the Swedish Council for Working Life and Social Research. We thank Paul deBakker and the Broad Institute for imputing the Framingham Heart Study genotypic data and for making the resultsavailable to other FHS researchers. We thank Emil Rehnberg of the Karolinska Institute for conducting theimputation and computing the principal components in the Study 3 dataset. We thank Yeon Sik Cho for researchassistance.



$waterm

ark-text$w

atermark-text

$waterm

ark-text

ReferencesBarnett JH, Scoriels L, Munaf MR. Meta-analysis of the cognitive effects of the catechol-O-

methyltransferase gene Val158/108Met polymorphism. Biological Psychiatry. 2008; 64:137144.[PubMed: 18339359]

Beauchamp JP, Cesarini D, Johannesson M, van der Loos M, Koellinger P, Groenen PJF, Fowler JH,Rosenquist N, Thurik AR, Christakis NA. Molecular genetics and economics. Journal of EconomicPerspectives. 2011; 25(4):5782. [PubMed: 22427719]

Benjamin, DJ.; Chabris, CF.; Glaeser, EL.; Gudnason, V.; Harris, T.; Laibson, DI.; Launer, L.; Purcell,S. Genoeconomics. In: Weinstein, M.; Vaupel, JW.; Watcher, KW., editors. Biosocial surveys.Washington, DC: The National Academies Press; 2007. p. 304-335.

Benjamin, DJ. White paper on genoeconomics. In: Lupia, A., editor. Genes, Cognition, and SocialBehavior: Next Steps for Foundations and Researchers. University of Michigan; 2010. p.66-77.manuscript. [www.isr.umich.edu/cps/workshop/NSF_Report_Final.pdf]

Benjamin, DJ.; Cesarini, DA.; Chabris, CF.; Glaeser, EL.; Laibson, DI., et al. The Promise and Pitfallsof Genoeconomics. 2011. Manuscript submitted for publication

Bouchard TJ Jr. Genetic and environmental influences on adult intelligence and special mentalabilities. Human Biology. 1998; 70:257179. [PubMed: 9549239]

Butcher LM, Davis OS, Craig IW, Plomin R. Genome-wide quantitative trait locus association scan ofgeneral cognitive ability using pooled DNA and 500K single nucleotide polymorphism microarrays.Genes, Brain, and Behavior. 2008; 7(4):435446.

Carlstedt, B. PhD thesis. Gothenburg University; 2000. Cognitive abilities: Aspects of structure,process and measurement. [http://gupea.ub.gu.se/bitstream/2077/9600/3/gupea_2077_9600_3.pdf]

Chabris, CF. Cognitive and neurobiological mechanisms of the Law of General Intelligence. In:Roberts, MJ., editor. Integrating the mind: Domain specific versus domain general processes inhigher cognition. Hove, UK: Psychology Press; 2007. p. 449-491.

Davies G, Tenesa A, Payton A, Yang J, Harris SE, Liewald D, et al. Genome-wide association studiesestablish that human intelligence is highly heritable and polygenic. Molecular Psychiatry. 2011

Dawber TR, Meadors GF, Moore FE. Epidemiological approaches to heart disease: The FraminghamStudy. American Journal of Public Health. 1951; 41:279286. [PubMed: 14819398]

Dickson S, Wang K, Krantz I, Hakonarson H, Goldstein D. Rare variants create synthetic genome-wide associations. PLoS Biology. 2010; 8(1)

Feinleib M, Kannel WB, Garrison RJ, McNamara PM, Castelli WP. The Framingham OffspringStudy: Design and preliminary data. Preventive Medicine. 1975; 4:518552. [PubMed: 1208363]

Hirschhorn JN, Lohmueller K, Byrne E, Hirschhorn K. A comprehensive review of genetic associationstudies. Genetics in Medicine. 2002; 4:4561. [PubMed: 11882781]

Houlihan LM, Harris SE, Luciano M, Gow AJ, Starr JM, Visscher PM, Deary IJ. Replication study ofcandidate genes for cognitive abilities: The Lothian Birth Cohort 1936. Genes, Brain andBehavior. 2009; 8:238247.

Ioannidis JPA, Ntzani EE, Trikalinos TA, Contopoulos-Ioannidis DG. Replication validity of geneticassociation studies. Nature Genetics. 2001; 29:306309. [PubMed: 11600885]

Ioannidis JP. Why most published research findings are false. PLoS Medicine. 2005; 2(8):e124.[PubMed: 16060722]

Ioannidis JP, Tarone R, McLaughlin JK. The false-positive to false-negative ratio in epidemiologicstudies. Epidemiology. 2011; 22(4):450456. [PubMed: 21490505]

Lamke, TA.; Nelson, MJ. Henmon-Nelson Tests of Mental Ability. Boston: Houghton Mifflin; 1957.(rev. ed.)

Lango Allen H, et al. Hundreds of variants clustered in genomic loci and biological pathways affecthuman height. Nature. 2010; 467:832838. [PubMed: 20881960]

Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika. 1986;73:1322.

Lichtenstein P, de Faire U, Floderus B, Svartengren M, Svedberg P, Pedersen NL. The Swedish TwinRegistry: A unique resource for clinical, epidemiological and genetic studies. Journal of InternalMedicine. 2002; 252:184205. [PubMed: 12270000]



$waterm

ark-text$w

atermark-text

$waterm

ark-text

http://gupea.ub.gu.se/bitstream/2077/9600/3/gupea_2077_9600_3.pdf

Manolio TA, Collins FS, Cox NJ, Goldstein DB, Hindor LA, et al. Finding the missing heritability ofcomplex diseases. Nature. 2009; 461:747753. [PubMed: 19812666]

Neisser U, et al. Intelligence: Knowns and unknowns. American Psychologist. 1996; 51(2):77101.

Payton A. The impact of genetic research on our understanding of normal cognitive ageing: 1995 to2009. Neuropsychology Review. 2009; 19:451477. [PubMed: 19768548]

Pearl PL, Gibson KM, Cortez MA, Wu Y, Snead OC 3rd, Knerr I, Forester K, Pettiford JM, Jakobs C,Theodore W. Succinic semialdehyde dehydrogenase deficiency: Lessons from mice and men.Journal of Inherited Metabolic Disease. 2009; 32(3):343352. [PubMed: 19172412]

Plomin R, Turic DM, Hill L, Turic DE, Stephens M, Williams J, et al. A functional polymorphism inthe succinate-semialdehyde dehydrogenase (aldehyde dehydrogenase 5 family, member A1) geneis associated with cognitive ability. Molecular Psychiatry. 2004; 9:582586. [PubMed: 14981524]

Plomin R, Kennedy JKJ, Craig IW. The quest for quantitative trait loci associated with intelligence.Intelligence. 2006; 34(6):513526.

Plomin, R.; McClearn, GE.; McGuffin, P.; DeFries, J. Behavioral Genetics. 5. New York: Worth;2008.

Purcell S, Cherny SS, Sham PC. Genetic Power Calculator: Design of linkage and association geneticmapping studies of complex traits. Bioinformatics. 2003; 19(1):149150. [PubMed: 12499305]

Ree MJ, Earles JA. The stability of g across different methods of estimation. Intelligence. 1991;15:271278.

Roetker, NS.; Yonker, JA.; Lee, C.; Chang, V.; Basson, J.; Roan, CL., et al. Exploring epistasis inclinically diagnosed depression in the Wisconsin Longitudinal Study: A pilot study utilizingrecursive partitioning analysis. 2011. Manuscript submitted for publication

Seshadri S, DeStefano AL, Au R, Massaro JM, Beiser AS, Kelly-Hayes M, et al. Genetic correlates ofbrain aging on MRI and cognitive test measures: a genome-wide association and linkage analysisin the Framingham Study. BMC Medical Genetics. 2007; 8:S15. [PubMed: 17903297]

Siontis KC, Patsopoulos NA, Ioannidis JP. Replication of past candidate loci for common diseases andphenotypes in 100 genome-wide association studies. European Journal of Human Genetics. 2010;18(7):832837. [PubMed: 20234392]

Spearman C. General intelligence, objectively determined and measured. American Journal ofPsychology. 1904; 15:201293.

Tung YC, Yeo GS. From GWAS to biology: Lessons from FTO. Annals of the New York Academy ofSciences. 2011; 1220:162171. [PubMed: 21388413]

Turkheimer E. Three laws of behavior genetics and what they mean. Current Directions inPsychological Science. 2000; 9:160164.

Weedon MN, Frayling TM. Reaching new heights: Insights into the genetics of human stature. Trendsin Genetics. 2008; 24(12):595603. [PubMed: 18950892]

Visscher PM, Hill WG, Wray NR. Heritability in the genomics era: Concepts and misconceptions.Nature Reviews Genetics. 2008; 9(4):255266.

Visscher PM. Sizing up human height variation. Nature Genetics. 2008; 40(5):489490. [PubMed:18443579]

Visscher PM, Yang J, Goddard ME. A commentary on Common SNPs explain a large proportion ofthe heritability for human height by Yang et al. (2010). Twin Research and Human Genetics.2010; 13:517524. [PubMed: 21142928]

Yang J, Benyamin B, McEvoy BP, Gordon S, Henders AK, Nyholt DR, et al. Common SNPs explain alarge proportion of the heritability for human height. Nature Genetics. 2010; 42:565569.[PubMed: 20562875]

Yeo RA, Gangestad SW, Liu J, Calhoun VD, Hutchison KE. Rare copy number deletions predictindividual variation in intelligence. PLoS One. 2011; 6(1):e16339. [PubMed: 21298096]



$waterm

ark-text$w

atermark-text

$waterm

ark-text

Figure 1.Statistical power of Studies 13 to detect significant associations between SNPs and g,plotted as a function of the percentage of variance in g explained by the SNP (or genotype inthe case of APOE e4). Note that the x-axis runs from 0% to 1% out of a total of 100%variance in g, so that 0.1 corresponds to 1/1000 of the total trait variance. Power wasestimated for the three studies using the full sample size (Upper bound on power for WLS,STR, and FHS) and using the number of unrelated individuals only (Lower bound onpower for WLS, STR, and FHS), yielding six power curves. Calculations were performedusing the tool created by Purcell, Cherny, and Sham (2003) [pngu.mgh.harvard.edu/~purcell/gpc/qtlassoc.html]. Assuming an effect size of 0.1% of variance for each genotypetested (shown by the dashed line), we should have observed between 10.6 and 14.7significant associations (for the unrelated and full samples, respectively), but we onlyobserved 1.



$waterm

ark-text$w

atermark-text

$waterm

ark-text

Figure 2.Regression coefficients for each genotype (i.e., difference in number of IQ points associatedwith each copy of the minor allele), pooled across Studies 13. To minimize the variance ofthe estimator, pooling was done by weighting the three estimated regression coefficients foreach SNP by the inverse of their estimated variances, with the weights then normalized sothat they sum to one. Error bars show 95% confidence intervals. For APOE, the bar showsthe number of IQ points associated with possessing at least one e4 allele.



$waterm

ark-text$w

atermark-text

$waterm

ark-text

$waterm

ark-text$w

atermark-text

$waterm

ark-text


Tabl

e 1

Res

ults

of

Stud

y 1.

Eac

h lin

e gi

ves

the

resu

lts f

or e

ach

SNP

of a

sep

arat

e lin

ear

regr

essi

on o

f g

(Hen

mon

-Nel

son

IQ)

on d

osag

e of

the

min

or a

llele

(0,

1,

or 2

cop

ies)

, con

trol

ling

for

age,

sex

, gra

duat

e/si

blin

g st

atus

, and

the

inte

ract

ions

of

thes

e fa

ctor

s, a

s w

ell a

s th

e fi

rst t

hree

pri

ncip

al c

ompo

nent

s of

the

90-

SNP

geno

type

cor

rela

tion

mat

rix

avai

labl

e in

the

Wis

cons

in L

ongi

tudi

nal S

tudy

dat

aset

. Sam

ple

size

var

ies

slig

htly

am

ong

SNPs

due

to m

issi

ng d

ata.

The

last

two

row

s sh

ow g

enot

ypes

that

wer

e av

aila

ble

in th

e W

LS

data

set,

but n

ot in

the

FHS

data

set (

Stud

y 2)

. The

R2

colu

mn

give

s th

e pe

rcen

tage

of

vari

ance

exp

lain

ed b

y a

univ

aria

te r

egre

ssio

n of

g o

n m

inor

alle

le d

osag

e fo

r ea

ch S

NP.

SNP

CH

RG

ene

NR

2 (%

)B

eta

Stan

dard

Err

ort

pM

AF

Min

or A

llele

Maj

or A

llele

rs10

1838

16p

DT

NB

P165

070.

040.

809

0.51

41.

57.1

2.0

80C

T

rs17

571

11p

CT

SD64

640.

010.

310

0.48

10.

64.5

2.0

79A

G

rs18

0049

711

qD

RD

2/A

NK

K1

6469

0.00

0.00

70.

356

0.02

.98

.191

AG

rs20

6117

47q

CH

RM

263

920.

000.

091

0.29

40.

31.7

6.3

28G

A

rs27

6011

86p

SSA

DH

(A

LD

H5A

1)64

790.

01

0.11

40.

340

0.

34.7

4.3

40T

C

rs46

8022

qC

OM

T64

200.

02

0.35

00.

270

1.

30.2

0.4

71G

A

rs62

6511

pB

DN

F64

890.

020.

367

0.33

11.

11.2

7.1

90T

C

rs76

0761

6pD

TN

BP1

6438

0.00

0.12

80.

330

0.39

.70

.206

AG

rs81

9199

27q

CH

RN

A4/

CH

RM

264

920.

000.

122

0.27

30.

45.6

6.4

74T

A

rs82

1616

1qD

ISC

164

780.

04

0.48

30.

293

1.

65.1

0.2

83T

A

rs42

9358

, rs7

412

19q

APO

E e

4 pr

esen

t/abs

ent

6390

0.00

0.04

10.

426

0.10

.92

.137

e4e2

/e3

rs36

3050

20p

SNA

P25

6464

0.04

0.32

30.

275

1.18

.24

.427

GA

Not

e: C

HR

= C

hrom

osom

e; M

AF

= M

inor

Alle

le F

requ

ency

.


$waterm

ark-text$w

atermark-text

$waterm

ark-text


Tabl

e 2

Res

ults

of

Stud

y 2.

Eac

h lin

e gi

ves

the

resu

lts f

or e

ach

SNP

of a

sep

arat

e lin

ear

regr

essi

on o

f g

(sco

re o

n th

e fi

rst p

rinc

ipal

com

pone

nt e

xtra

cted

fro

m a

batte

ry o

f ni

ne c

ogni

tive

test

s) o

n do

sage

of

the

min

or a

llele

(0,

1, o

r 2

copi

es),

con

trol

ling

for

a cu

bic

of a

ge, a

cub

ic o

f ag

e in

tera

cted

with

sex

, the

fir

st10

pri

ncip

al c

ompo

nent

s of

the

SNP

geno

type

cor

rela

tion

mat

rix,

and

stu

dy c

ohor

t, w

ith c

lust

erin

g by

ext

ende

d fa

mili

es, i

n th

e Fr

amin

gham

Hea

rt S

tudy

data

set (

N =

175

9). T

he R

2 co

lum

n gi

ves

the

perc

enta

ge o

f va

rian

ce e

xpla

ined

by

a un

ivar

iate

reg

ress

ion

of g

on

min

or a

llele

dos

age

for

each

SN

P.

SNP

CH

RG

ene

R2

(%)

Bet

aSt

anda

rd E

rror

tp

MA

FM

inor

Alle

leM

ajor

Alle

le

rs10

1838

16p

DT

NB

P10.

020.

607

0.92

80.

655

.51

.088

CT

rs17

571

11p

CT

SD0.

06

0.93

51.

105

0.

846

.40

.086

AG

rs18

0049

711

qD

RD

2/A

NK

K1

0.14

0.

914

0.63

2

1.44

8.1

5.2

02A

G

rs20

6117

47q

CH

RM

20.

00

0.00

90.

600

0.

014

.10

.318

GA

rs27

6011

86p

SSA

DH

(A

LD

H5A

1)0.

23

1.15

80.

576

2.

011

.04

.309

TC

rs46

8022

qC

OM

T0.

02

0.26

00.

539

0.

481

.63

.486

GA

rs62

6511

pB

DN

F0.

010.

298

0.69

50.

429

.67

.189

TC

rs76

0761

6pD

TN

BP1

0.01

0.21

80.

687

0.31

7.7

5.1

91A

G

rs81

9199

27q

CH

RN

A4/

CH

RM

20.

00

0.03

90.

551

0.

071

.94

.440

TA

rs82

1616

1qD

ISC

10.

02

0.38

70.

608

0.

636

.53

.287

TA

Not

e: C

HR

= C

hrom

osom

e; M

AF

= M

inor

Alle

le F

requ

ency

.


$waterm

ark-text$w

atermark-text

$waterm

ark-text


Tabl

e 3

Res

ults

of

Stud

y 3.

Eac

h lin

e gi

ves

the

resu

lts f

or e

ach

SNP

of a

sep

arat

e lin

ear

regr

essi

on o

f g

(sco

re o

n th

e fi

rst p

rinc

ipal

com

pone

nt e

xtra

cted

fro

m a

batte

ry o

f ni

ne c

ogni

tive

test

s) o

n do

sage

of

the

min

or a

llele

(0,

1, o

r 2

copi

es),

con

trol

ling

for

the

firs

t 10

prin

cipa

l com

pone

nts

of th

e SN

P ge

noty

peco

rrel

atio

n m

atri

x, a

nd s

tudy

coh

ort,

with

clu

ster

ing

by f

amily

. The

sam

ple

is c

ompr

ised

exc

lusi

vely

of

mal

e Sw

edis

h tw

ins

born

bet

wee

n 19

36 a

nd 1

958,

who

all

took

the

test

s ne

ar th

e ag

e of

18.

SNP

CH

RG

ene

NR

2 (%

)B

eta

Std

Err

ort

pM

AF

Min

or A

llele

Maj

or A

llele

rs10

1838

16p

DT

NB

P124

41.1

03

1.35

01.

120

1.

21.2

28.0

69C

T

rs17

571

11p

CT

SD24

41.0

440.

744

0.94

30.

79.4

30.0

73A

G

rs18

0049

711

qD

RD

2/A

NK

K1

2441

.007

0.

345

0.69

8

0.49

.621

.180

AG

rs20

6117

47q

CH

RM

224

41.0

05

0.11

20.

540

0.

21.8

35.3

19G

A

rs27

6011

86p

SSA

DH

(A

LD

H5A

1)24

41.1

630.

803

0.50

81.

58.1

14.3

75T

C

rs46

8022

qC

OM

T24

41.0

20

0.23

30.

498

0.

47.6

40.4

47G

A

rs62

6511

pB

DN

F24

41.0

380.

592

0.65

30.

91.3

65.1

95T

C

rs76

0761

6pD

TN

BP1

2441

.109

0.

907

0.63

1

1.44

.151

.221

AG

rs81

9199

27q

CH

RN

A4/

CH

RM

224

41.0

740.

524

0.49

51.

06.2

90.4

56T

A

rs82

1616

1qD

ISC

124

41.0

15

0.42

00.

520

0.

81.4

19.3

18T

A

Not

e: C

HR

= C

hrom

osom

e; M

AF

= M

inor

Alle

le F

requ

ency

.


Ni Hms 357611

Documents