GWAPower: a statistical power calculation software for genome-wide association studies with quantitative traits
© Feng et al; licensee BioMed Central Ltd. 2011
Received: 2 March 2010
Accepted: 21 January 2011
Published: 21 January 2011
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.
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.
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
Statistical power and sample size calculation is an important step during experiment design in genome-wide association (GWA) studies. It estimates the probability that a true genetic effect can be detected under the experimental constraints and other reasonable assumptions. To date, more than 230 GWA studies have been conducted and reported. Among these, case control design (>100 studies) and cohort studies with unrelated subjects (>70 studies) are the two most popular designs. In the majority of cohort studies, one or more phenotypic variables are quantitative traits. A number of statistical power calculation softwares have been developed for case-control GWA studies [1–5]. This report presents a simple and user-friendly statistical software, GWAPower, for power calculation of GWA studies with quantitative traits.
General statistical power calculation procedures for quantitative measures often assume normality and require information of the first two distribution moments (means and variances) of the quantitative measure. These power calculation procedures cannot be applied directly in genetic studies. One challenge is that a key parameter, i.e. the size of the hypothesized genetic effect, is represented in a genetic term, "heritability". In its broad sense, heritability is defined as the amount of phenotypic variation in the population that can be ascribed to the genetic variation among individuals. If the total variability of the phenotype is V(t) and the variations that can be explained by genetic variants and error terms are V(g) and V(e), respectively, where V(t) = V(g) + V(e), then the heritability (H) is estimated as: H = V(g)/V(t). Although H is intrinsically related to the distribution of the quantitative measure, the exact mathematical transformation formula is not clear; given the value of H, the first two moments of the quantitative measure cannot be determined directly.
However, a power calculation procedure based on heritability is possible for GWA studies, since most applied genetic models in GWA analysis are simple and have only one degree of freedom. Using straightforward algebra, a sound statistical procedure and convenient graphical user interface software have been developed to estimate the statistical power for quantitative traits in GWA studies.
Results and Discussion
Methods for calculating power with heritability in one-degree-of-freedom models
where τi is the phenotype mean of genotype group i; σ2 is the error variation; and ri is the number of individuals in each genotype group i. Therefore, distribution parameters such as means and variances are still necessary for this approach.
We provide a solution that avoids assuming the non-centrality parameter in power calculations. In GWA data analysis, some popular genetic models have one degree of freedom including the dominant model, the recessive model and the additive model. When these genetic models are assumed, the one-way ANOVA model (dominant and recessive models) and the simple linear regression model (additive models) are used in the majority of GWA data analyses for quantitative traits. In these studies, the F-statistic is related to Student's t statistic, i.e. if a random variable X has a t distribution with d degrees of freedom, then X2 has F distribution (1, d). This is true for central and non-central distributions . This suggests that so long as the genetic model contains only one degree of freedom, then an F distribution can be used to perform the square root transformation to obtain the t statistics for the power calculation.
The ANOVA table for the dominate and the recessive model
Degree of Freedom
Sum of Squares
P = 1
It should be noted that (1) the absolute values of SST and SSM are unknown, as are the absolute values of numerator and denominator of the t-statistic. However, in this approach, only the relative values are necessary to calculate power; (2) the t-test is associated with the ANOVA table. In this study, an ANOVA (and the associated F-test) is used owing to its ability to accommodate the concept of heritability and to adjust for covariate effects. Therefore, if a dominant or a recessive genetic model is considered, the t-test compares two sample means within an ANOVA framework, although an ANOVA t-test is different from a "two sample t-test".
Covariates: users can input the percentage of the variation explained by the covariates, in addition to the genetic effects and the number of total covariates. The information is included in the ANOVA table and the statistical power is adjusted accordingly, i.e. the degree of freedom of the t-test is N-2-number of covariates;
Linkage Disequilibrium (LD): users can input the two-locus LD (r2) between the SNP marker and the hypothesized causal variant. When the LD between the genotyped and "causative" marker is r2, the sample size is increased by 1/r2, because in order to have the same power to detect a causative variant when using a marker that is r2 away, an increase in sample size of 1/r2 is required [7, 8].
The GWAPower package was developed using Microsoft Visual C++ 2008 and consists a main dialog graphical user interface (GUI). Users can download and install the compiled versions of the GUI that run as stand-alone applications under Windows operating systems. A demonstration of the GUI for power calculations with the corresponding parameters setting is presented in Figure 1. By clicking the button "SNPs Power Curves", the GUI displays power curves of various sample size with H = 1% ~ 10%. The graphs can be copied into a Windows clipboard by using Menu - Copy to Clipboard. The graphs can be pasted into other Windows applications.
Example for HIV studies
The statistical power estimated by GWAPower for the HIV studies
The GWAPower package provides a simple and useful statistical power calculation procedure for GWAS with quantitative traits. It is designed specifically to allow genetic researchers to use the genetic term, heritability, instead of the general statistical term, 'phenotype means of each genotype', in power calculations. Current users have reported that the tool is very user-friendly and easy to operate.
Although GWAPower is developed for GWA studies, it can be used for general genetic studies if researchers wish to use heritability as the parameter for genetic effect size. However, this approach should be used with caution. Some complicating factors are not considered such as population stratification, various interactions among genetic variants, and environmental factors. Furthermore, Bonferroni correction for multiple tests is known to be over-conservative. Therefore, for complicated designs with multiple confounding factors, researchers are encouraged to consult a statistical geneticist.
One reviewer has reported that another software "genetic power calculator" (GPC)  can be used to estimate statistical power in GWA studies with quantitative traits, with some simple adjustments. A major difference between GPC and GWAPower is that GPC is based on a comprehensive mixed model approach , making it amenable for use in various experimental designs. GWAPower takes advantage of the fact that most GWA study designs are reasonably simple, and much simpler models such as ANOVA or regression models are often adequate. In such cases, the approximations of estimating distribution parameters are not necessary.
This study was supported by funding from the Measurement to Understand Re-Classification of Disease of Cabarrus and Kannapolis (MURDOCK) Study and by a grant from the NIH CTSA (Clinical and Translational Science Award) 1UL1RR024128-01 to the Duke University.
- Purcell S, Cherny SS, Sham PC: Genetic Power Calculator: design of linkage and association genetic mapping studies of complex traits. Bioinformatics. 2003, 19 (1): 149-150. 10.1093/bioinformatics/19.1.149.View ArticlePubMedGoogle Scholar
- Gordon D, Haynes C, Blumenfeld J, Finch SJ: PAWE-3D: visualizing power for association with error in case-control genetic studies of complex traits. Bioinformatics. 2005, 21 (20): 3935-3937. 10.1093/bioinformatics/bti643.View ArticlePubMedGoogle Scholar
- Skol AD, Scott LJ, Abecasis GR, Boehnke M: Joint analysis is more efficient than replication-based analysis for two-stage genome-wide association studies. Nat Genet. 2006, 38 (2): 209-213. 10.1038/ng1706.View ArticlePubMedGoogle Scholar
- Menashe I, Rosenberg PS, Chen BE: PGA: power calculator for case-control genetic association analyses. BMC Genet. 2008, 13: 9-36.Google Scholar
- Edwards BJ, Haynes C, Levenstien MA, Finch SJ, Gordon D: Power and sample size calculations in the presence of phenotype errors for case/control genetic association studies. BMC Genet. 2005, 8: 6-18.Google Scholar
- Shao J: Mathematical statistics: exercises and solutions. 2005, Springer, 2Google Scholar
- Risch N, Teng J: The relative power of family-based and case-control designs for linkage disequilibrium studies of complex human diseases I. DNA pooling. Genome Res. 1998, 8 (12): 1273-1288.PubMedGoogle Scholar
- Pritchard JK, Przeworski M: Linkage disequilibrium in humans: models and data. Am J Hum Genet. 2001, 69 (1): 1-14. 10.1086/321275.PubMed CentralView ArticlePubMedGoogle Scholar
- Weedon MN, Lango H, Lindgren CM, Wallace C, Evans DM, Mangino M: Genome-wide association analysis identifies 20 loci that influence adult height. Nat Genet. 2008, 40 (5): 575-583. 10.1038/ng.121.PubMed CentralView ArticlePubMedGoogle Scholar
- Fellay J, Shianna KV, Ge D, Colombo S, Ledergerber B, Weale M, Zhang K, Gumbs C, Castagna A, Cossarizza A, Cozzi-Lepri A, Luca AD, Easterbrook P, Francioli P, Mallal S, Martinez-Picado J, Miro JM, Obel N, Smith JP, Wyniger J, Descombes P, Antonarakis SE, Letvin NL, McMichael AJ, Haynes BF, Telenti A, Goldstein DB: A whole-genome association study of major determinants for host control of HIV-1. Science. 2007, 317 (5840): 944-947. 10.1126/science.1143767.PubMed CentralView ArticlePubMedGoogle Scholar
- Fellay J, Ge D, Shianna KV, Colombo S, Ledergerber B, Cirulli ET, Urban TJ, Zhang K, Gumbs CE, Smith JP, Castagna A, Cozzi-Lepri A, De Luca A, Easterbrook P, Günthard HF, Mallal S, Mussini C, Dalmau J, Martinez-Picado J, Miro JM, Obel N, Wolinsky SM, Martinson JJ, Detels R, Margolick JB, Jacobson LP, Descombes P, Antonarakis SE, Beckmann JS, O'Brien SJ, Letvin NL, McMichael AJ, Haynes BF, Carrington M, Feng S, Telenti A, Goldstein DB: Common genetic variation and the control of HIV-1 in humans. PLoS Genet. 2009, 5 (12): e1000791-10.1371/journal.pgen.1000791.PubMed CentralView ArticlePubMedGoogle Scholar
- Sham PC, Cherny SS, Purcell S, Hewitt JK: Power of linkage versus association analysis of quantitative traits, by use of variance-components models, for sibship data. Am J Hum Genet. 2000, 66: 1616-1630. 10.1086/302891.PubMed CentralView ArticlePubMedGoogle Scholar