Title:Efficient and powerful familywise error control in genome-wide association studies using generalized linear models

Abstract:In genetic association studies, detecting phenotype--genotype association is a primary goal. We consider testing a large number of common genetic variants, and stress the importance of controlling the familywise error rate. Multiple testing correction methods may achieve this by estimating a local significance level for the individual tests. We assume that data -- phenotype, genetic markers and environmental covariates -- are available from independent individuals in a case--control, cohort study or cross-sectional study. Further, we assume that the relationship between the phenotype and the genetic and environmental covariates can be modelled by a generalized linear model (GLM). We use a multivariate score statistic, which under the complete null hypothesis of no phenotype--genotype association asymptotically has a multivariate normal distribution with a covariance matrix that can be estimated from key features of the fitted GLM model, under the complete null hypothesis, and the genetic markers. For common genetic variants, it is well known that the genotype correlation between variants can be markedly non-zero. We use an approximation to the distribution of the complete multivariate score test statistics to estimate the local significance level. We show, using simulated and real data (one study on schizophrenia and bipolar disorder and one study on maximal oxygen uptake), that our method is a powerful alternative to the popular Bonferroni method. For GLMs without environmental covariates, we show that our method is an efficient alternative to permutation methods for multiple testing. As a byproduct, the local significance level can be used to define an effective number of independent tests.