Getting Power Back from Population Structure and Kinship in Genome-Wide Association Studies

Reporting false discoveries not only damages academic reputation, but also wastes resources by misleading follow-up studies. Population structure and kinship among individuals are two common factors that cause false positives in genome-wide association studies (GWAS). The Mixed Linear Model (MLM) is an effective method to eliminate false positives. MLM fits both testing genetic marker and population structure as fixed effects and incorporates kinship to define the variance and covariance structures of the random individual genetic effect. Unfortunately, the confounding of testing genetic marker with population structure and kinship makes the statistical test on the genetic marker less powerful. Even when the genetic marker is a functional polymorphism, its effect is deluded by the covariate population structure and the kinship. The confounding effect is more problematic on traits that are associated with population structure such as flowering time in plants. New MLM methods, such as SUPER GWAS method, have been developed to reduce the confounding. SUPER GWAS method uses the associate markers to define the kinship and exclude the associate markers that are in linkage disequilibrium with the testing markers. This presentation describes the principles and applications of these new MLM methods to simultaneously reduce false positives and retain high statistical power.