to impute with SNPs, therefore, we imputed the GWAS genotype data with the classical HLA alleles (Supplementary Information), and studied whether specific HLA alleles would associate with ASPD. We detected four HLA alleles at HLA-DRB1 and HLA-DQA1 genes with similar significance as individual SNPs but with higher OR in association with ASPD (Supplementary Table 2). The strongest associations were seen with a common HLA-DRB1 allele DRB1*01:01 (OR=2.19 (1.53–3.14), P=1.9 × 10−5, f=0.17) and with DQA1*01:01 (OR=2.09 (1.46–2.99), P=5.6 × 10−5, f=0.17) that are known to be in tight LD (current data r2=0.981). In addition, protective associations with DRB1*04:04 and DQB1*03:02 alleles were detected with similar protective effects as the strongest individual SNPs of the GWAS analysis (DRB1*04:04 OR=0.39 (0.18–0.57), P=9.7 × 10−5, f=0.04 and DQB1*03:02 OR=0.52 (0.37–0.74), P=2.5 × 10−4, f=0.11). To investigate whether the HLA allele effects were directly explained by the individual SNPs, we conditioned the analysis for the strongest variant at the HLA region (rs9268528), which failed to remove the association with DRB1*01:01 (conditioned P=2.3 × 10−3). Similarly, conditioning for the strongest four HLA alleles failed to remove the association with rs9268528, suggesting independent roles for the HLA alleles and rs9268528 in ASPD.
