To address the genome-wide multiple testing problem of our report, we also performed additional analyses. First, to address the impact of the 25 CNVRs detected in the family-based GWAS discovery sample, we stored the CNV-FBAT P-values of the respective 1314 CNV markers (those with two-sided asymptotic P-values ≤ 0.05; Table 2). Next, we drew 1000 random sets of 1314 CNV markers out of the 244 analysed CNVRs (8051 CNV markers) to derive the distribution of P-values under the null hypothesis. Next, both the quantiles of –log10 P-values of observed association signals and the corresponding 97.5, 50 and 2.5% percentiles of the 1000 random sets of equal size were plotted against the expected quantiles (from a uniform distribution) in a QQ plot (Supplementary Material, Fig. S1). While both the observed and sampled results indicated the conservativeness of the family-based discovery GWAS approach, the smaller P-values of the observed signals justify our subsequent proceeding.
