When genotype data are available, such as in individual GWA studies, confounding effects on , e.g. gene size, can be corrected for using phenotype permutation analysis that does not require a priori knowledge of the confounders (described in Materials and Methods). However, to exploit the power of large GWA study meta-analyses, where permutation analysis cannot be performed due to unavailability of genotype data, we needed an alternative correction method. We chose a linear regression-based approach that adjusts for the effects of multiple confounders on the gene score. This required identifying a substantial amount of the confounding effects on .
