While the trans-ancestry PRS was significantly associated with T2D status across the ancestral groups we examined and showed potential for clinical translation, the raw PRS had major distributional shifts by ancestry (Additional File 1: Fig. S4; left panel), impeding its implementation in clinical care where a single cutoff of the PRS distribution across diverse populations is needed to identify high-risk individuals. By modeling the mean and variance of the trans-ancestry PRS as functions of population structure captured by genetic PCs, the post hoc ancestry adjustment method eliminated major distributional differences in the PRS across ancestries in the eMERGE dataset (Additional File 1: Fig. S4; right panel). The adjusted PRS showed good overall calibration across eMERGE samples, as demonstrated by largely concordant predicted and observed risk (i.e., the proportion of cases) in each PRS decile (Additional File 1: Fig. S5). To examine the impact of this post hoc adjustment on the tail discrimination of the PRS, we compared the OR of individuals in the top percentiles of the adjusted PRS across ancestries with the OR of individuals in the top percentiles
