We also performed haplotype analyses for CPD and the smoking-related diseases (COPD, lung cancer and ASO) using the GWAS and the replication samples. We assumed six haplotypes at the CYP2A6 locus, each of which consisted of a combination of an allele at rs11878604[T/C] and a haplotypic copy number at rs8102683[0/1/2]. Here, the MAP copy number dosage [8] was used as the copy number genotype. Population haplotype frequencies and their relative effect on CPD were jointly evaluated using the haplotype-specific linear regression model, as described previously [27]. In this model, we used the same covariates that were used in the GWAS and the replication study. All model parameters, including the effect sizes of the six haplotypes, were estimated using a standard EM (expectation-maximization) algorithm [28].
