1、GWAS分析中的加法模型Statistical analysis. GWAS analysis was performed using the additive model by logistic regression analysis. Population structure was evalsuated by PCA in the software package EIGENSTRAT 3.0 (ref. 20). We used PLINK 1.07 for general

statistical analysis. SNPs for the first stage of replication were selected based on a separate meta-analysis for every SNP from the Nanjing and the Beijing GWAS.

Our default meta-analysis used a fixed-effect model with inverse variance weighting and a calculation of Cochran’s Q statistic and the I2 statistic for heterogeneity22. When there was no indication of heterogeneity for a SNP (P for Q >0.05), the fixed-effect model was kept in place. When heterogeneity was present (P for Q ≤ 0.05), we adopted a random-effects model (DerSimonian-Laird) for that SNP. The Manhattan plot of −log10 P was generated using R 2.11.1. Weused MACH 1.0 software to impute ungenotyped SNPs using the LD information from the HapMap 3 database (CHB and JPT as reference set, released Feb.2009). The chromosome region was plotted using an online tool, LocusZoom 1.1. PCA identified one significant (P < 0.05) eigenvector, which we included in the logistic regression analysis with the other covariates of age, gender, and smoking and drinking status for both the GWAS and the combined analysis.

The chi-squared (χ2)-based Cochran’s Q statistic was also calculated to test for heterogeneity between groups in stratified analysis. Gene-gene and geneenvironment multiplicative interactions were tested by a general logistic regression model using the equation

Y b = + 0 1b A × + b B 2 3 × + b A × × ( B e ) + where Y is the logit of case-control status, A and B are factors (SNP or environmental), b0 is constant, b1 and b2 are the main effects of factor A and B,respectively, and b3 is the interaction term. P values are two sided, and the ORs reported in the manuscript are from an additive model by logistic regression analyses unless otherwise specified. The analyses were also performed using SAS version 9.1.3 (SAS Institute) or Stata version 9.2 (StataCorp LP).

(责任编辑：广东会GDH基因)