We performed imputation using the reference genotype in combination with the CNPs and the SNPs in the GWAS. We selected markers within a 1 cM region around the top CNP (Figure S4). The SNPs in the commonly deleted region were then imputed as if they were tri-allelic markers (Figure S11). The Beagle software [33] used played an important role because it can perform multi-allelic imputation. The association of these tri-allelic SNPs with CPD was evaluated using a linear regression model in conjunction with the covariates used in the GWAS. Here, the allele dosage of the difference in the number of A and B alleles (ranging from to ) was assessed with the assumption of an additive effect on CPD, as proposed previously [34].
