Tobacco and alcohol use are heritable behaviours that can be radically affected by environmental factors, including cultural context37 and public health policies38,39. Despite this, we found that a large majority of associated genetic variants showed homogeneous effect size estimates across diverse ancestries, suggesting that the genetic variants associated with substance use affect such individuals similarly. The limited extent of variant effect size heterogeneity, coupled with similar heritability estimates and cross-trait genetic correlations, indicates that the genetic architecture underlying substance use is not markedly different across ancestries. There are some potentially interesting exceptions of ancestrally heterogeneous effects in genes such as ADH1B and CACNA1B. By contrast, polygenic scores generally performed well in EUR ancestries but with mixed-to-limited results in other ancestries, suggesting that portability of such scores across ancestries remains challenging, even when discovery sample sizes across all ancestries are more than 100,000. Explanations for this apparent discrepancy have been proposed40, but more stringent and sensitive tests will be required to draw strong conclusions about such patterns of heredity.
