To identify independent associations within significant signals in EA, we performed conditional haplotype analysis using WHAP [45]. This analysis showed that controlling for rs2476601 accounted for most of the association between PTPN22 and SLE in EA (Table S1). However, since rs1217414 remained significant after being conditioned on rs2476601 (P = 0.022), and rs2476601 remained significant (P = 7.0×10−7) after conditioned on rs1217414, we performed haplotype analysis using these SNPs. The overall haplotype was significant (Phap = 2.40×10−9) at a similar magnitude as individual SNP rs2476601 (P = 4.13×10−9). In fact, these SNPs are physically 35.6 kb apart and have a very low LD (r2 = 0.04), implying independence. The risk haplotype at both SNPs (AG) was the most significant (P = 9.58×10−9, OR = 1.38), and the non-risk haplotype (GA) was significant as well (P = 7.39×10−4, OR = 0.88). The haplotype which was non-risk for rs2476601/risk for rs1217414 was not significant (GG) (P = 0.63). When we compared the AG (both risk alleles) vs. GA (both non-risk alleles), the OR increased to 1.47 (p = 9.5×10−9). However, we did not observe the (AA) haplotype in our data.
