meta-analysis were used. A total of 96 polygenic risk scores were evaluated in each phenotype exploring the impact of ancestral population (two scenarios), p value threshold (16 scenarios), and variant weighting (three scenarios). The proportion of variation explained by each PRS (partial-R2) approach was assessed for UKB European ancestry and African ancestry individuals separately. The partial-R2 was calculated from the difference in R2 values following linear regression of HbA1c levels on age, sex, BMI, and first 10 PCs with and without the PRS also included. Similarly, for asthma and prostate cancer, we determined the Nagelkerke’s pseudo partial-R2 following logistic regression of case status on age, sex (asthma only), BMI (prostate cancer only), and first 10 PCs with and without the PRS. Additionally, in African ancestry individuals we created a combined PRS (α1PRSEUR+α2PRSAFR), where PRSEUR and PRSAFR was the most optimal PRS using variants from the designated population and the weight and p value that resulted in the highest accuracy; albeit in-sample, optimization was done within a single PRS to ensure limited overfitting of the combined model.10 We used 5-fold cross-validation to assess model performance in which 80% of the cohort was used to estimate the mixing coefficients (α1 and α2)
