We tested all 14 cohorts individually for association under an additive logistic regression model using PLINK [28]. For samples derived from parent child trios, we applied a case-pseudocontrol design in which the pseudocontrol was created with the non-transmitted alleles from the parents. Since these pseudocontrols are perfectly matched to each case, no covariates were used in association analysis from these trio cohorts. For the non-trio cohorts, each regression included derived principal components as covariates [29]. To minimise the putative reduction in power observed when non-confounding covariates are included in GWAS for rare disease [30], sex was not included as a covariate in these analyses. Individual PP plots for each cohort are reported in the Additional file 1: Figure S1).
