We developed a multi-ancestry meta-regression method to meta-analyse ancestrally diverse genome-wide association study (GWAS) summary statistics from 60 cohorts with 3,383,199 individuals (Supplementary Table 1; see Supplementary Fig. 1 for an overview of the project), representing major clines of recent human ancestry (Fig. 1a). The meta-analytic method used here uses meta-regression to account for per study axes of genetic ancestry variation combined with a random effect to capture further unexplained heterogeneity in the effect of a given genetic variant. Although ancestry here is continuous, we also performed secondary analyses of continental groups reflecting four ancestry clines, including individuals of African (AFR; maximum n = 119,589) and American (AMR; n = 286,026) recently admixed ancestries primarily from the United States; individuals of East Asian ancestries (EAS; n = 296,438) primarily from the United States, People’s Republic of China and Japan; and individuals of European ancestries (EUR; n = 2,669,029) from the United States, Europe and Australia (see Extended Data Fig. 1 and Supplementary Note). Smoking phenotypes were selected to represent different stages of tobacco use and addiction, including initiation, the onset
