The effects of human population structure on large genetic association studies

Large-scale association studies hold substantial promise for unraveling the genetic basis of common human diseases. A well-known problem with such studies is the presence of undetected population structure, which can lead to both false positive results and failures to detect genuine associations. Here we examine approximately 15,000 genome-wide single-nucleotide polymorphisms typed in three population groups to assess the consequences of population structure on the coming generation of association studies. The consequences of population structure on association outcomes increase markedly with sample size. For the size of study needed to detect typical genetic effects in common diseases, even the modest levels of population structure within population groups cannot safely be ignored. We also examine one method for correcting for population structure (Genomic Control). Although it often performs well, it may not correct for structure if too few loci are used and may overcorrect in other settings, leading to substantial loss of power. The results of our analysis can guide the design of large-scale association studies.     

Meta-analysis of genetic association studies supports a contribution of common variants to susceptibility to common disease

Association studies offer a potentially powerful approach to identify genetic variants that influence susceptibility to common disease, but are plagued by the impression that they are not consistently reproducible. In principle, the inconsistency may be due to false positive studies, false negative studies or true variability in association among different populations. The critical question is whether false positives overwhelmingly explain the inconsistency. We analyzed 301 published studies covering 25 different reported associations. There was a large excess of studies replicating the first positive reports, inconsistent with the hypothesis of no true positive associations (P < 10(-14)). This excess of replications could not be reasonably explained by publication bias and was concentrated among 11 of the 25 associations. For 8 of these 11 associations, pooled analysis of follow-up studies yielded statistically significant replication of the first report, with modest estimated genetic effects. Thus, a sizable fraction (but under half) of reported associations have strong evidence of replication; for these, false negative, underpowered studies probably contribute to inconsistent replication. We conclude that there are probably many common variants in the human genome with modest but real effects on common disease risk, and that studies using large samples will convincingly identify such variants.     

Demonstrating stratification in a European American population

Population stratification occurs in case-control association studies when allele frequencies differ between cases and controls because of ancestry. Stratification may lead to false positive associations, although this issue remains controversial. Empirical studies have found little evidence of stratification in European-derived populations, but potentially significant levels of stratification could not be ruled out. We studied a European American panel discordant for height, a heritable trait that varies widely across Europe. Genotyping 178 SNPs and applying standard analytical methods yielded no evidence of stratification. But a SNP in the gene LCT that varies widely in frequency across Europe was strongly associated with height (P < 10(-6)). This apparent association was largely or completely due to stratification; rematching individuals on the basis of European ancestry greatly reduced the apparent association, and no association was observed in Polish or Scandinavian individuals. The failure of standard methods to detect this stratification indicates that new methods may be required.     

Principal components analysis corrects for stratification in genome-wide association studies

Population stratification--allele frequency differences between cases and controls due to systematic ancestry differences-can cause spurious associations in disease studies. We describe a method that enables explicit detection and correction of population stratification on a genome-wide scale. Our method uses principal components analysis to explicitly model ancestry differences between cases and controls. The resulting correction is specific to a candidate marker's variation in frequency across ancestral populations, minimizing spurious associations while maximizing power to detect true associations. Our simple, efficient approach can easily be applied to disease studies with hundreds of thousands of markers.     

The common PPARgamma Pro12Ala polymorphism is associated with decreased risk of type 2 diabetes

Genetic association studies are viewed as problematic and plagued by irreproducibility. Many associations have been reported for type 2 diabetes, but none have been confirmed in multiple samples and with comprehensive controls. We evaluated 16 published genetic associations to type 2 diabetes and related sub-phenotypes using a family-based design to control for population stratification, and replication samples to increase power. We were able to confirm only one association, that of the common Pro12Ala polymorphism in peroxisome proliferator-activated receptor-gamma(PPARgamma) with type 2 diabetes. By analysing over 3,000 individuals, we found a modest (1.25-fold) but significant (P=0.002) increase in diabetes risk associated with the more common proline allele (85% frequency). Moreover, our results resolve a controversy about common variation in PPARgamma. An initial study found a threefold effect, but four of five subsequent publications failed to confirm the association. All six studies are consistent with the odds ratio we describe. The data implicate inherited variation in PPARgamma in the pathogenesis of type 2 diabetes. Because the risk allele occurs at such high frequency, its modest effect translates into a large population attributable risk-influencing as much as 25% of type 2 diabetes in the general population.     

Including known covariates can reduce power to detect genetic effects in case-control studies

Genome-wide association studies (GWAS) search for associations between genetic variants and disease status, typically via logistic regression. Often there are covariates, such as sex or well-established major genetic factors, that are known to affect disease susceptibility and are independent of tested genotypes at the population level. We show theoretically and with data from recent GWAS on multiple sclerosis, psoriasis and ankylosing spondylitis that inclusion of known covariates can substantially reduce power for the identification of associated variants when the disease prevalence is lower than a few percent. Whether the inclusion of such covariates reduces or increases power to detect genetic effects depends on various factors, including the prevalence of the disease studied. When the disease is common (prevalence of >20%), the inclusion of covariates typically increases power, whereas, for rarer diseases, it can often decrease power to detect new genetic associations.     

A unified mixed-model method for association mapping that accounts for multiple levels of relatedness

As population structure can result in spurious associations, it has constrained the use of association studies in human and plant genetics. Association mapping, however, holds great promise if true signals of functional association can be separated from the vast number of false signals generated by population structure. We have developed a unified mixed-model approach to account for multiple levels of relatedness simultaneously as detected by random genetic markers. We applied this new approach to two samples: a family-based sample of 14 human families, for quantitative gene expression dissection, and a sample of 277 diverse maize inbred lines with complex familial relationships and population structure, for quantitative trait dissection. Our method demonstrates improved control of both type I and type II error rates over other methods. As this new method crosses the boundary between family-based and structured association samples, it provides a powerful complement to currently available methods for association mapping.     

Replication validity of genetic association studies

The rapid growth of human genetics creates countless opportunities for studies of disease association. Given the number of potentially identifiable genetic markers and the multitude of clinical outcomes to which these may be linked, the testing and validation of statistical hypotheses in genetic epidemiology is a task of unprecedented scale. Meta-analysis provides a quantitative approach for combining the results of various studies on the same topic, and for estimating and explaining their diversity. Here, we have evaluated by meta-analysis 370 studies addressing 36 genetic associations for various outcomes of disease. We show that significant between-study heterogeneity (diversity) is frequent, and that the results of the first study correlate only modestly with subsequent research on the same association. The first study often suggests a stronger genetic effect than is found by subsequent studies. Both bias and genuine population diversity might explain why early association studies tend to overestimate the disease protection or predisposition conferred by a genetic polymorphism. We conclude that a systematic meta-analytic approach may assist in estimating population-wide effects of genetic risk factors in human disease.     

Differential confounding of rare and common variants in spatially structured populations

Well-powered genome-wide association studies, now made possible through advances in technology and large-scale collaborative projects, promise to characterize the contribution of rare variants to complex traits and disease. However, while population structure is a known confounder of association studies, it remains unknown whether methods developed to control stratification are equally effective for rare variants. Here, we demonstrate that rare variants can show a stratification that is systematically different from, and typically stronger than, common variants, and this is not necessarily corrected by existing methods. We show that the same process leads to inflation for load-based tests and can obscure signals at truly associated variants. Furthermore, we show that populations can display spatial structure in rare variants, even when Wright's fixation index F(ST) is low, but that allele frequency-dependent metrics of allele sharing can reveal localized stratification. These results underscore the importance of collecting and integrating spatial information in the genetic analysis of complex traits.     

An Icelandic example of the impact of population structure on association studies

The impact of population structure on association studies undertaken to identify genetic variants underlying common human diseases is an issue of growing interest. Spurious associations of alleles with disease phenotypes may be obtained or true associations overlooked when allele frequencies differ notably among subpopulations that are not represented equally among cases and controls. Population structure influences even carefully designed studies and can affect the validity of association results. Most study designs address this problem by sampling cases and controls from groups that share the same nationality or self-reported ethnic background, with the implicit assumption that no substructure exists within such groups. We examined population structure in the Icelandic gene pool using extensive genealogical and genetic data. Our results indicate that sampling strategies need to take account of substructure even in a relatively homogenous genetic isolate. This will probably be even more important in larger populations.     

A new multipoint method for genome-wide association studies by imputation of genotypes

Genome-wide association studies are set to become the method of choice for uncovering the genetic basis of human diseases. A central challenge in this area is the development of powerful multipoint methods that can detect causal variants that have not been directly genotyped. We propose a coherent analysis framework that treats the problem as one involving missing or uncertain genotypes. Central to our approach is a model-based imputation method for inferring genotypes at observed or unobserved SNPs, leading to improved power over existing methods for multipoint association mapping. Using real genome-wide association study data, we show that our approach (i) is accurate and well calibrated, (ii) provides detailed views of associated regions that facilitate follow-up studies and (iii) can be used to validate and correct data at genotyped markers. A notable future use of our method will be to boost power by combining data from genome-wide scans that use different SNP sets.     

Genomic screening and replication using the same data set in family-based association testing

The Human Genome Project and its spin-offs are making it increasingly feasible to determine the genetic basis of complex traits using genome-wide association studies. The statistical challenge of analyzing such studies stems from the severe multiple-comparison problem resulting from the analysis of thousands of SNPs. Our methodology for genome-wide family-based association studies, using single SNPs or haplotypes, can identify associations that achieve genome-wide significance. In relation to developing guidelines for our screening tools, we determined lower bounds for the estimated power to detect the gene underlying the disease-susceptibility locus, which hold regardless of the linkage disequilibrium structure present in the data. We also assessed the power of our approach in the presence of multiple disease-susceptibility loci. Our screening tools accommodate genomic control and use the concept of haplotype-tagging SNPs. Our methods use the entire sample and do not require separate screening and validation samples to establish genome-wide significance, as population-based designs do.     

Transferability of tag SNPs in genetic association studies in multiple populations

A general question for linkage disequilibrium-based association studies is how power to detect an association is compromised when tag SNPs are chosen from data in one population sample and then deployed in another sample. Specifically, it is important to know how well tags picked from the HapMap DNA samples capture the variation in other samples. To address this, we collected dense data uniformly across the four HapMap population samples and eleven other population samples. We picked tag SNPs using genotype data we collected in the HapMap samples and then evaluated the effective coverage of these tags in comparison to the entire set of common variants observed in the other samples. We simulated case-control association studies in the non-HapMap samples under a disease model of modest risk, and we observed little loss in power. These results demonstrate that the HapMap DNA samples can be used to select tags for genome-wide association studies in many samples around the world.     

'Racial' differences in genetic effects for complex diseases

'Racial' differences are frequently debated in clinical, epidemiological and molecular research and beyond. In particular, there is considerable controversy regarding the existence and importance of 'racial' differences in genetic effects for complex diseases influenced by a large number of genes. An important question is whether ancestry influences the impact of each gene variant on the disease risk. Here, we addressed this question by examining the genetic effects for 43 validated gene-disease associations across 697 study populations of various descents. The frequencies of the genetic marker of interest in the control populations often (58%) showed large heterogeneity (statistical variability) between 'races'. Conversely, we saw large heterogeneity in the genetic effects (odds ratios) between 'races' in only 14% of cases. Genetic markers for proposed gene-disease associations vary in frequency across populations, but their biological impact on the risk for common diseases may usually be consistent across traditional 'racial' boundaries.     

Joint analysis is more efficient than replication-based analysis for two-stage genome-wide association studies

Genome-wide association is a promising approach to identify common genetic variants that predispose to human disease. Because of the high cost of genotyping hundreds of thousands of markers on thousands of subjects, genome-wide association studies often follow a staged design in which a proportion (pi(samples)) of the available samples are genotyped on a large number of markers in stage 1, and a proportion (pi(samples)) of these markers are later followed up by genotyping them on the remaining samples in stage 2. The standard strategy for analyzing such two-stage data is to view stage 2 as a replication study and focus on findings that reach statistical significance when stage 2 data are considered alone. We demonstrate that the alternative strategy of jointly analyzing the data from both stages almost always results in increased power to detect genetic association, despite the need to use more stringent significance levels, even when effect sizes differ between the two stages. We recommend joint analysis for all two-stage genome-wide association studies, especially when a relatively large proportion of the samples are genotyped in stage 1 (pi(samples) >or= 0.30), and a relatively large proportion of markers are selected for follow-up in stage 2 (pi(markers) >or= 0.01).     

Bayesian inference of epistatic interactions in case-control studies

Epistatic interactions among multiple genetic variants in the human genome may be important in determining individual susceptibility to common diseases. Although some existing computational methods for identifying genetic interactions have been effective for small-scale studies, we here propose a method, denoted 'bayesian epistasis association mapping' (BEAM), for genome-wide case-control studies. BEAM treats the disease-associated markers and their interactions via a bayesian partitioning model and computes, via Markov chain Monte Carlo, the posterior probability that each marker set is associated with the disease. Testing this on an age-related macular degeneration genome-wide association data set, we demonstrate that the method is significantly more powerful than existing approaches and that genome-wide case-control epistasis mapping with many thousands of markers is both computationally and statistically feasible.     

Systematic meta-analyses and field synopsis of genetic association studies in schizophrenia: the SzGene database

In an effort to pinpoint potential genetic risk factors for schizophrenia, research groups worldwide have published over 1,000 genetic association studies with largely inconsistent results. To facilitate the interpretation of these findings, we have created a regularly updated online database of all published genetic association studies for schizophrenia ('SzGene'). For all polymorphisms having genotype data available in at least four independent case-control samples, we systematically carried out random-effects meta-analyses using allelic contrasts. Across 118 meta-analyses, a total of 24 genetic variants in 16 different genes (APOE, COMT, DAO, DRD1, DRD2, DRD4, DTNBP1, GABRB2, GRIN2B, HP, IL1B, MTHFR, PLXNA2, SLC6A4, TP53 and TPH1) showed nominally significant effects with average summary odds ratios of approximately 1.23. Seven of these variants had not been previously meta-analyzed. According to recently proposed criteria for the assessment of cumulative evidence in genetic association studies, four of the significant results can be characterized as showing 'strong' epidemiological credibility. Our project represents the first comprehensive online resource for systematically synthesized and graded evidence of genetic association studies in schizophrenia. As such, it could serve as a model for field synopses of genetic associations in other common and genetically complex disorders.     

Evaluating potential for whole-genome studies in Kosrae, an isolated population in Micronesia

Whole-genome association studies are predicted to be especially powerful in isolated populations owing to increased linkage disequilibrium (LD) and decreased allelic diversity, but this possibility has not been empirically tested. We compared genome-wide data on 113,240 SNPs typed on 30 trios from the Pacific island of Kosrae to the same markers typed in the 270 samples from the International HapMap Project. The extent of LD is longer and haplotype diversity is lower in Kosrae than in the HapMap populations. More than 98% of Kosraen haplotypes are present in HapMap populations, indicating that HapMap will be useful for genetic studies on Kosrae. The long-range LD around common alleles and limited diversity result in improved efficiency in genetic studies in this population and augments the power to detect association of 'hidden SNPs'.     

Genome-wide efficient mixed-model analysis for association studies

Linear mixed models have attracted considerable attention recently as a powerful and effective tool for accounting for population stratification and relatedness in genetic association tests. However, existing methods for exact computation of standard test statistics are computationally impractical for even moderate-sized genome-wide association studies. To address this issue, several approximate methods have been proposed. Here, we present an efficient exact method, which we refer to as genome-wide efficient mixed-model association (GEMMA), that makes approximations unnecessary in many contexts. This method is approximately n times faster than the widely used exact method known as efficient mixed-model association (EMMA), where n is the sample size, making exact genome-wide association analysis computationally practical for large numbers of individuals.     

Meta-analysis of genome-wide association studies of asthma in ethnically diverse North American populations

Asthma is a common disease with a complex risk architecture including both genetic and environmental factors. We performed a meta-analysis of North American genome-wide association studies of asthma in 5,416 individuals with asthma (cases) including individuals of European American, African American or African Caribbean, and Latino ancestry, with replication in an additional 12,649 individuals from the same ethnic groups. We identified five susceptibility loci. Four were at previously reported loci on 17q21, near IL1RL1, TSLP and IL33, but we report for the first time, to our knowledge, that these loci are associated with asthma risk in three ethnic groups. In addition, we identified a new asthma susceptibility locus at PYHIN1, with the association being specific to individuals of African descent (P = 3.9 × 10(-9)). These results suggest that some asthma susceptibility loci are robust to differences in ancestry when sufficiently large samples sizes are investigated, and that ancestry-specific associations also contribute to the complex genetic architecture of asthma.