Abstract The large number of disease susceptibility loci identified from genome-wide association studies (GWAS) is enabling polygenic risk scores (PRS) to deliver on their promise to improve health outcomes and to transform the practice of personalized medicine. The reduced quality of PRS for common diseases and related quantitative traits for populations of recent African, Asian, and Native American ancestries relative to those for populations of recent European ancestries, however, threatens to create a new class of disparities in the delivery of healthcare based on PRS. The number of individuals of non-European ancestry with genome interrogation are growing much more rapidly now than 5 years ago; nevertheless, the number of individuals of recent European ancestries with genome interrogation grows still more rapidly and it is likely to be many years before sample sizes for genome interrogation in even major continental groups are close to proportional to relative population sizes. Thus, it is critical to optimize PRS performance for diverse populations in as many ways as we can. Given the substantial and growing fraction of the US population with genomes admixed from different continental ancestries, we believe that high quality PRS for much of the US population is unlikely to be achieved without properly accounting for local ancestries. Similarly, focused strategies to identify high-impact but population-specific variants could improve the quality of PRS in populations with such alleles. DNA variants from regions with a signature of natural selection often demonstrate such properties, and have been shown to be enriched among top associations for a number of hematological and immune/inflammatory traits that are important biomarkers for key chronic diseases. We propose to focus our PRS studies on hematological and immune/inflammatory traits and their associated chronic diseases and to extend methods for the development of PRS to accommodate estimates of local ancestry, high impact population-specific variants and multiple endophenotypes. Thus, our Specific Aims are: 1) Assemble and harmonize data sets needed to accomplish the goals of the project, including hematological traits (red blood cell, white blood cell, platelet), and immune/inflammatory traits (CRP, fibrinogen, D-dimer) from: Jackson Heart Study, Women’s Health Initiative, BioVU, and GeneSTAR. 2) Extend PRS methods to: a) explicitly model local ancestry; b) accommodate large-effect but population-specific risk alleles (such as those from regions with a signature of natural selection); and c) enable joint modeling of multiple endophenotypes; and 3) Develop and apply novel PRS and overall disease prediction models to: a) estimate risk of common diseases and related biomarkers affected by hematological, thrombotic and immune/inflammatory biology; and b) enable calculation of PRS-adjusted clinical laboratory values to reduce structural health disparities.

PROJECT NARRATIVE We will leverage existing biomarker data to derive new polygenic risk scores (PRS) to improve prediction of disease across ancestrally diverse populations, particularly focusing on inflammation and hematological measures under selective pressure from differential infectious disease exposures. We will develop high quality PRS for common laboratory measures that differ in means and/or variances by ancestry, which would enable improved diagnostic accuracy and reduce unnecessary repeat testing and downstream healthcare utilization.