The last evaluation step was to recover known GWAS hits for COPD24,27. As shown in Fig. 2c, we conducted three GWAS for COPD, as follows: (1) COPD cases from COPDGene EA with WGS data (N = 2736) vs. controls from COGEND EA with array data (N = 1961); (2) COPD cases from ECLIPSE EA with array data (N = 1764) vs. controls from COPDGene EA with WGS data (N = 2475); and (3) COPD cases from COPDGene AA with WGS data (N = 813) vs. controls from COGEND AA with array data (N = 712). Because COPD is highly comorbid with smoking history, only smokers (current and former) were used as controls to compare with COPD cases across these GWAS analyses. All association models include ten principal components as covariates to account for population substructure. Results for each GWAS analysis are presented in Supplementary Fig. 6. Meta-analysis of the 3 analyses successfully recovered 5 out of 7 loci reported as COPD-associated (Fig. 4 and Table 2) at genome-wide significance (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$P\, <
