Leveraging genetic ancestry continuum information to interpolate PRS for admixed populations

Abstract

Calculating optimal polygenic risk scores (PRS) across diverse ancestries, particularly in admixed populations, is necessary to enable equitable genetic research and clinical translation. However, the relatively low representation of admixed populations in both discovery and fine-tuning individual-level datasets limits PRS development for admixed populations. Under the assumption that the most informative PRS weight for a homogeneous sample, which can be approximated by a data point in the ancestry continuum space, varies linearly in that space, we introduce a Genetic Distance-assisted PRS Combination Pipeline for Diverse Genetic Ancestries (DiscoDivas) to interpolate a harmonized PRS for diverse, especially admixed, ancestries, leveraging multiple PRS weights fine-tuned within single-ancestry samples and the genetic ancestry continuum information. DiscoDivas treats ancestry as a continuous variable and does not require shifting between different models when calculating PRS for different ancestries. We generated PRS with DiscoDivas and the current conventional method, i.e. fine-tuning multiple GWAS PRS using the matched or similar ancestry sample, for simulated datasets and large-scale biobank datasets (UK Biobank [UKBB] N=415,402, Mass General Brigham Biobank N=53,306, All of Us N=245,394) and compared our method with the conventional method with quantitative traits and complex disease traits. DiscoDivas generated a harmonized PRS of the accuracy comparable to or higher than the conventional approach, with the greatest advantage exhibited in admixed samples: DiscoDivas PRS for admixed samples was more statistically accurate than the PRS fine-tuned in matched or similar ancestry sample in 12 out of 16 simulated scenarios and was statistically equivalent in the remaining four scenarios; when tested with quantitative trait data in UKBB, DiscoDivas increased the PRS accuracy of admixed sample by 5% on average; yet no statistical difference was observed when tested for binary traits in UKBB where ancestry-matched data was available. For the single ancestry samples, the accuracy of DiscoDivas PRS and PRS fine-tuned in match samples was similar. In summary, our method DiscoDivas yields a harmonized PRS of robust accuracy for individuals across the genetic ancestry spectrum, including where ancestry-matched training data may be incomplete.