Genetic association studies of complex traits: design and analysis issues

Most common diseases and many important quantitative traits are complex genetic traits, with multiple genetic and environmental variables contributing to the observed phenotype. Because of the multi-factorial nature of complex traits, each individual genetic variant generally has only a modest effect, and the interaction of genetic variants with each other or with environmental factors can potentially be quite important in determining the observed phenotype. It remains largely unknown what sort of genetic variants explain inherited variation in complex traits, but recent evidence suggests that common genetic variants will explain at least some of the inherited variation in susceptibility to common disease. Genetic association studies, in which the allele or genotype frequencies at markers are determined in affected individuals and compared with those of controls (either population- or family-based), may be an effective approach to detecting the effects of common variants with modest effects. With the explosion in single nucleotide polymorphism (SNP) discovery and genotyping technologies, large-scale association studies have become feasible, and small-scale association studies have become plentiful. We review the different types of association studies and discuss issues that are important to consider when performing and interpreting association studies of complex genetic traits. Heritable and accurately measured phenotypes, carefully matched large samples, well-chosen genetic markers, and adequate standards in genotyping, analysis, and interpretation are all integral parts of a high-quality association study.     

Combined linkage and association sib-pair analysis for quantitative traits.

An extension to current maximum-likelihood variance-components procedures for mapping quantitative-trait loci in sib pairs that allows a simultaneous test of allelic association is proposed. The method involves modeling of the allelic means for a test of association, with simultaneous modeling of the sib-pair covariance structure for a test of linkage. By partitioning of the mean effect of a locus into between- and within-sibship components, the method controls for spurious associations due to population stratification and admixture. The power and efficacy of the method are illustrated through simulation of various models of both real and spurious association.     

A genome-wide association study uncovers novel genomic regions and candidate genes of yield-related traits in upland cotton

Key message

A total of 62 SNPs associated with yield-related traits were identified by a GWAS. Based on significant SNPs, two candidate genes pleiotropically increase lint yield.

Abstract

Improved fibre yield is considered a constant goal of upland cotton (Gossypium hirsutum) breeding worldwide, but the understanding of the genetic basis controlling yield-related traits remains limited. To better decipher the molecular mechanism underlying these traits, we conducted a genome-wide association study to determine candidate loci associated with six yield-related traits in a population of 719 upland cotton germplasm accessions; to accomplish this, we used 10,511 single-nucleotide polymorphisms (SNPs) genotyped by an Illumina CottonSNP63K array. Six traits, including the boll number, boll weight, lint percentage, fruit branch number, seed index and lint index, were assessed in multiple environments; large variation in all phenotypes was detected across accessions. We identified 62 SNP loci that were significantly associated with different traits on chromosomes A07, D03, D05, D09, D10 and D12. A total of 689 candidate genes were screened, and 27 of them contained at least one significant SNP. Furthermore, two genes (Gh_D03G1064 and Gh_D12G2354) that pleiotropically increase lint yield were identified. These identified SNPs and candidate genes provide important insights into the genetic control underlying high yields in G. hirsutum, ultimately facilitating breeding programmes of high-yielding cotton.

     

Bivariate association analysis for quantitative traits using generalized estimation equation

Quantitative traits often underlie risk for complex diseases. Many studies collect multiple correlated quantitative phenotypes and perform univariate analyses on each of them respectively. However, this strategy may not be powerful and has limitations to detect plei- otropic genes that may underlie correlated quantitative traits. In addition, testing multiple traits individually will exacerbate perplexing problem of multiple testing. In this study, generalized estimating equation 2 （GEE2） is applied to association mapping of two correlated quantitative traits. We suppose that a quantitative trait locus is located in a chromosome region that exerts pleiotropic effects on multiple quantitative traits. In that region, multiple SNPs are genotyped. Genotypes of these SNPs and the two quantitative traits affected by a causal SNP were simulated under various parameter values： residual correlation coefficient between two traits, causal SNP heritability, minor allele frequency of the causal SNP, extent of linkage disequilibrium with the causal SNP, and the test sample size. By power ana- lytical analyses, it is showed that the bivariate method is generally more powerful than the univariate method. This method is robust and yields false-positive rates close to the pre-set nominal significance level. Our real data analyses attested to the usefulness of the method.     

Simulating association studies: a data-based resampling method for candidate regions or whole genome scans

MOTIVATION: Reductions in genotyping costs have heightened interest in performing whole genome association scans and in the fine mapping of candidate regions. Improvements in study design and analytic techniques will require the simulation of datasets with realistic patterns of linkage disequilibrium and allele frequencies for typed SNPs. METHODS: We describe a general approach to simulate genotyped datasets for standard case-control or affected child trio data, by resampling from existing phased datasets. The approach allows for considerable flexibility in disease models, potentially involving a large number of interacting loci. The method is most applicable for diseases caused by common variants that have not been under strong selection, a class specifically targeted by the International HapMap project. RESULTS: Using the three population Phase I/II HapMap data as a testbed for our approach, we have implemented the approach in HAP-SAMPLE, a web-based simulation tool.     

GWAPower: a statistical power calculation software for genome-wide association studies with quantitative traits

Background

In designing genome-wide association (GWA) studies it is important to calculate statistical power. General statistical power calculation procedures for quantitative measures often require information concerning summary statistics of distributions such as mean and variance. However, with genetic studies, the effect size of quantitative traits is traditionally expressed as heritability, a quantity defined as the amount of phenotypic variation in the population that can be ascribed to the genetic variants among individuals. Heritability is hard to transform into summary statistics. Therefore, general power calculation procedures cannot be used directly in GWA studies. The development of appropriate statistical methods and a user-friendly software package to address this problem would be welcomed.

Results

This paper presents GWAPower, a statistical software package of power calculation designed for GWA studies with quantitative traits, where genetic effect is defined as heritability. Based on several popular one-degree-of-freedom genetic models, this method avoids the need to specify the non-centrality parameter of the F-distribution under the alternative hypothesis. Therefore, it can use heritability information directly without approximation. In GWAPower, the power calculation can be easily adjusted for adding covariates and linkage disequilibrium information. An example is provided to illustrate GWAPower, followed by discussions.

Conclusions

GWAPower is a user-friendly free software package for calculating statistical power based on heritability in GWA studies with quantitative traits. The software is freely available at: http://dl.dropbox.com/u/10502931/GWAPower.zip     

Multilocus association testing of quantitative traits based on partial least-squares analysis

Because of combining the genetic information of multiple loci, multilocus association studies (MLAS) are expected to be more powerful than single locus association studies (SLAS) in disease genes mapping. However, some researchers found that MLAS had similar or reduced power relative to SLAS, which was partly attributed to the increased degrees of freedom (dfs) in MLAS. Based on partial least-squares (PLS) analysis, we develop a MLAS approach, while avoiding large dfs in MLAS. In this approach, genotypes are first decomposed into the PLS components that not only capture majority of the genetic information of multiple loci, but also are relevant for target traits. The extracted PLS components are then regressed on target traits to detect association under multilinear regression. Simulation study based on real data from the HapMap project were used to assess the performance of our PLS-based MLAS as well as other popular multilinear regression-based MLAS approaches under various scenarios, considering genetic effects and linkage disequilibrium structure of candidate genetic regions. Using PLS-based MLAS approach, we conducted a genome-wide MLAS of lean body mass, and compared it with our previous genome-wide SLAS of lean body mass. Simulations and real data analyses results support the improved power of our PLS-based MLAS in disease genes mapping relative to other three MLAS approaches investigated in this study. We aim to provide an effective and powerful MLAS approach, which may help to overcome the limitations of SLAS in disease genes mapping.     

The association of flowering time quantitative trait loci with duplicated regions and candidate loci in Brassica oleracea.

A population of 150 doubled haploid lines of rapid cycling Brassica oleracea, derived from an F1 from a var. alboglabra x var. italica cross, was scored for flowering time in two trials. Using information on 82 mapped molecular markers, spread evenly across the nine linkage groups, QTL were identified at six locations; one each on linkage groups O2 and O3 and two each on linkage groups O5 and O9. In total, these QTL explained 58 and 93% of the genetical variation in the two trials. Three of these QTL, on linkage groups O2, O3, and O9, were situated in regions showing considerable homology both with each other and with chromosome regions of B. nigra that have been shown to affect flowering time. These same regions are all homologous to a single tract of Arabidopsis chromosome 5, which contains a number of the flowering-related genes, one or more of which may be candidates for the QTL found in Brassica.     

Joint tests of linkage and association for quantitative traits

In this paper we describe various study designs and analytic techniques for testing the joint hypothesis that a genetic marker is both linked to and associated with a quantitative phenotype. Issues of power and sampling are addressed. The distinction between methods that explicitly examine association and those that infer association by examining the distribution of allelic transmissions from a heterozygous parent is examined. Extensions to multivariate, multiallelic, and multilocus situations are addressed. Recent approaches that combine variance-components-based linkage analyses with joint tests of linkage in the presence of association for disentanglement of the linkage and association and the application of such methods to fine mapping are discussed. Finally, new classes of joint tests of linkage and association that do not require samples of related individuals are described.     

Designing candidate gene and genome-wide case-control association studies

This protocol describes how to appropriately design a genetic association case-control study, either focusing on a candidate gene (CG) or region or implementing a genome-wide approach. The steps described involve: (i) defining the case phenotype in adequate detail; (ii) checking the heritability of the disease in question; (iii) considering whether a population-based study is the appropriate design for the research question; (iv) the appropriate selection of controls; (v) sample size calculations and (vi) giving due consideration to whether it is a de novo or replication study. General guidelines are given, as well as specific examples of a CG and a genome-wide association study into type 2 diabetes. Software and websites used in this protocol include the International HapMap Consortium website, Genetic Power Calculator, CaT, and SNPSpD. Running each of the programs takes only a few seconds; the rate-limiting steps involve thinking through the designs and parameters in the disease models.     

Nonparametric evaluation of quantitative traits in population-based association studies when the genetic model is unknown

Statistical association between a single nucleotide polymorphism (SNP) genotype and a quantitative trait in genome-wide association studies is usually assessed using a linear regression model, or, in the case of non-normally distributed trait values, using the Kruskal-Wallis test. While linear regression models assume an additive mode of inheritance via equi-distant genotype scores, Kruskal-Wallis test merely tests global differences in trait values associated with the three genotype groups. Both approaches thus exhibit suboptimal power when the underlying inheritance mode is dominant or recessive. Furthermore, these tests do not perform well in the common situations when only a few trait values are available in a rare genotype category (disbalance), or when the values associated with the three genotype categories exhibit unequal variance (variance heterogeneity). We propose a maximum test based on Marcus-type multiple contrast test for relative effect sizes. This test allows model-specific testing of either dominant, additive or recessive mode of inheritance, and it is robust against variance heterogeneity. We show how to obtain mode-specific simultaneous confidence intervals for the relative effect sizes to aid in interpreting the biological relevance of the results. Further, we discuss the use of a related all-pairwise comparisons contrast test with range preserving confidence intervals as an alternative to Kruskal-Wallis heterogeneity test. We applied the proposed maximum test to the Bogalusa Heart Study dataset, and gained a remarkable increase in the power to detect association, particularly for rare genotypes. Our simulation study also demonstrated that the proposed non-parametric tests control family-wise error rate in the presence of non-normality and variance heterogeneity contrary to the standard parametric approaches. We provide a publicly available R library nparcomp that can be used to estimate simultaneous confidence intervals or compatible multiplicity-adjusted p-values associated with the proposed maximum test.     

Density of common complex ocular traits in the aging eye: analysis of secondary traits in genome-wide association studies

Genetic association studies are identifying genetic risks for common complex ocular traits such as age-related macular degeneration (AMD). The subjects used for discovery of these loci have been largely from clinic-based, case-control studies. Typically, only the primary phenotype (e.g., AMD) being studied is systematically documented and other complex traits (e.g., affecting the eye) are largely ignored. The purpose of this study was to characterize these other or secondary complex ocular traits present in the cases and controls of clinic-based studies being used for genetic study of AMD. The records of 100 consecutive new patients (of any diagnosis) age 60 or older for which all traits affecting the eye had been recorded systematically were reviewed. The average patient had 3.5 distinct diagnoses. A subset of 10 complex traits was selected for further study because they were common and could be reliably diagnosed. The density of these 10 complex ocular traits increased by 0.017 log-traits/year (P = 0.03), ranging from a predicted 2.74 at age 60 to 4.45 at age 90. Trait-trait association was observed only between AMD and primary vitreomacular traction (P = 0.0009). Only 1% of subjects age 60 or older had no common complex traits affecting the eye. Extrapolations suggested that a study of 2000 similar subjects would have sufficient power to detect genetic association with an odds ratio of 2.0 or less for 4 of these 10 traits. In conclusion, the high prevalence of complex traits affecting the aging eye and the inherent biases in referral patterns leads to the potential for confounding by undocumented secondary traits within case-control studies. In addition to the primary trait, other common ocular phenotypes should be systematically documented in genetic association studies so that adjustments for potential trait-trait associations and other bias can be made and genetic risk variants identified in secondary analyses.     

Sibling-based tests of linkage and association for quantitative traits.

The transmission/disequilibrium test (TDT) developed by Spielman et al. can be a powerful family-based test of linkage and, in some cases, a test of association as well as linkage. It has recently been extended in several ways; these include allowance for implementation with quantitative traits, allowance for multiple alleles, and, in the case of dichotomous traits, allowance for testing in the absence of parental data. In this article, these three extensions are combined, and two procedures are developed that offer valid joint tests of linkage and (in the case of certain sibling configurations) association with quantitative traits, with use of data from siblings only, and that can accommodate biallelic or multiallelic loci. The first procedure uses a mixed-effects (i.e., random and fixed effects) analysis of variance in which sibship is the random factor, marker genotype is the fixed factor, and the continuous phenotype is the dependent variable. Covariates can easily be accommodated, and the procedure can be implemented in commonly available statistical software. The second procedure is a permutation-based procedure. Selected power studies are conducted to illustrate the relative power of each test under a variety of circumstances.     

famCNV: copy number variant association for quantitative traits in families

A program package to enable genome-wide association of copy number variants (CNVs) with quantitative phenotypes in families of arbitrary size and complexity. Intensity signals that act as proxies for the number of copies are modeled in a variance component framework and association with traits is assessed through formal likelihood testing. AVAILABILITY AND IMPLEMENTATION: The Java package is made available at www.imperial.ac.uk/medicine/people/m.falchi/. CONTACT: m.falchi@imperial.ac.uk.     

Linkage and association to candidate regions in Swedish atopic dermatitis families

We have studied, in 406 families with at least two siblings affected with atopic dermatitis (in total 1514 individuals) from the Swedish population, linkage and association to five chromosomal regions (2q35, 5q31-33, 6p21, 11q13 and 14q11) previously implicated in atopic diseases. The region on 14q11 gave evidence for linkage to atopic dermatitis (NPL-score: 2.36, P<0.009). In the 11q13 region, there was a clear association to an intragenic marker in the beta-subunit of the high-affinity IgE receptor for raised allergen-specific serum IgE levels (P<0.009). When a quantitative variable for the severity of atopic dermatitis was studied, evidence was found in favour of linkage to the 5q31-33 region, with the highest Z-score (2.06) close to the marker D5S458 (P<0.005).     

A powerful and flexible multilocus association test for quantitative traits

Association mapping of complex traits typically employs tagSNP genotype data to identify a trait locus within a region of interest. However, considerable debate exists regarding the most powerful strategy for utilizing such tagSNP data for inference. A popular approach tests each tagSNP within the region individually, but such tests could lose power as a result of incomplete linkage disequilibrium between the genotyped tagSNP and the trait locus. Alternatively, one can jointly test all tagSNPs simultaneously within the region (by using genotypes or haplotypes), but such multivariate tests have large degrees of freedom that can also compromise power. Here, we consider a semiparametric model for quantitative-trait mapping that uses genetic information from multiple tagSNPs simultaneously in analysis but produces a test statistic with reduced degrees of freedom compared to existing multivariate approaches. We fit this model by using a dimension-reducing technique called least-squares kernel machines, which we show is identical to analysis using a specific linear mixed model (which we can fit by using standard software packages like SAS and R). Using simulated SNP data based on real data from the International HapMap Project, we demonstrate that our approach often has superior performance for association mapping of quantitative traits compared to the popular approach of single-tagSNP testing. Our approach is also flexible, because it allows easy modeling of covariates and, if interest exists, high-dimensional interactions among tagSNPs and environmental predictors.