To maximize power, it is critical to make use of meta-analyses of multiple independent GWA studies whose increasing sample size (from thousands of people in single studies to tens of thousands in meta-analyses) boosts the statistical power for detecting clustering of modest associations. Yet, as opposed to traditional GWA studies, information about individual genotypes is not available for most meta-analyses, making it impossible to evaluate statistical significance through standard phenotype permutation analysis. While several GSEA variations have been recently applied to meta-analyses, the extent to which they account for confounding effects on gene association scores has not been tested.
