Regarding the study design we focused on CNV testing in the family-based discovery sample as a first step while using the case–control discovery GWAS sample as secondary discovery filter. The focus on our family-based sample has several advantages: (i) the known robustness of the family-based design against population stratification; (ii) better control for biases (e.g. plate effects) due to the within-family comparisons; (iii) due to the incorporation of only one affected offspring per pedigree, the CNV-FBAT leads to quite conservative results, which lowers the risk of false positives. The conservativeness of the CNV-FBAT is displayed in the QQ plots for the two discovery samples (Supplementary Material, Fig. S2). In contrast to the CNV-FBAT, the results for the case–control analyses are too liberal although we performed normalizations of the intensity signals to, for example, address technical problems (e.g. plate effects).
