The relative sample compositions of GWAS result in highly predictable disparities in prediction accuracy; population genetics theory predicts that genetic risk prediction accuracy will decay with increasing genetic divergence between the original GWAS sample and target of prediction, a function of population history13,14. This pattern can be attributed to several statistical observations which we detail below: 1) GWAS favor the discovery of genetic variants that are common in the study population; 2) linkage disequilibrium (LD) differentiates marginal effect size estimates for polygenic traits across populations, even when causal variants are the same; and 3) environment and demography differ across populations. Notably, the first two phenomena degrade prediction performance across populations substantially even when there exist no biological, environmental, or diagnostic differences, whereas the environment and demography may interact to drive differential forces of natural selection that in turn drive differences in causal genetic architecture. (We define the causal genetic architecture as the true effects of variants that impact a phenotype that would be identified in a population of infinite sample size. Unlike effect size estimates, true effects are typically modeled as invariant with respect to LD and allele frequency differences across populations.)
