Y (that is, 36% versus 10%, based on the principles of Hardy–Weinberg equilibrium where f(AA+Aa) = p 2+2pq). Furthermore, owing to differences in population history (including effective population size and distinctive rates of immigration and inbreeding), inheritance patterns of genomic variants are more variable and complex in some ethnic groups than others (for example, persons of African ancestry tend to have shorter haplotype blocks compared with persons of European ancestry) 18– 20. More complex inheritance patterns will require a greater number of gene variants to be included in the PRS, and because the SNPs included in PRS are often not themselves causal, causal variants may be “tagged” by different SNPs in different populations, giving rise to different detected effect sizes and different groups of maximally informative SNPs. Effect sizes are also influenced by gene-by-environment interactions, which may result in population-specific variability. Inclusion of quantitative genetic features of ethnicity as covariates in the association tests will adjust—to some degree—these ethnicity-specific effects in the primary GWAS, but application of those data to independent samples of different ethnicity to the discovery sample is especially prone to errors and fraught with potential misinterpretation 16, 17. For this reason (among many others), it is imperative
