The expected polygenic risk score (ePRS) framework: an equitable metric for quantifying polygenetic risk via modeling of ancestral makeup

Abstract

Polygenic risk scores (PRSs) depend on genetic ancestry due to differences in allele frequencies between ancestral populations. This leads to implementation challenges in diverse populations. We propose a framework to calibrate PRS based on ancestral makeup. We define a metric called expected PRS (ePRS), the expected value of a PRS based on the global or local admixture patterns of an individual. We further define the residual PRS (rPRS), measuring the deviation of the PRS from the ePRS. Simulation studies confirm that it suffices to adjust for ePRS to obtain nearly unbiased estimates of the PRS-outcome association without further adjusting for PCs. Using the TOPMed dataset, the estimated effect size of the rPRS adjusting for the ePRS is similar to the estimated effect of the PRS adjusting for genetic PCs. The ePRS framework can protect from population stratification in association analysis and provide an equitable strategy to quantify genetic risk across diverse populations.