least one non-European population (with minimum sample size of 8,000) that are listed in the NHGRI-EBI GWAS Catalog (MacArthur et al., 2017), where our approach could potentially be valuable (S21 Table). Intuitively, our approach leverages both large training sample sizes and training data with target-matched LD patterns. We note that the effects of differential tagging (or different causal effect sizes) in different populations can potentially be quantified using cross-population genetic correlation (Brown, Ye, Price, & Zaitlen, in press; de Candia et al., 2013; Mancuso et al., 2016), and that leveraging data from a different population to improve predictions is a natural analogue to leveraging data from a correlated trait (Maier et al., 2015).
