Abstract TH957: Performance of Cardiometabolic Polygenic Scores in a high-altitude Peruvian Population

Abstract

Background: Polygenic scores (PGS) aggregate the effects of multiple genetic variants to estimate disease risk. Limited ancestral diversity in discovery and validation datasets may reduce performance in underrepresented populations. We evaluated the transferability of associations for existing PGS for cardiometabolic phenotypes in a Peruvian population. Methods: We conducted a cross-sectional analysis of 685 participants from Puno, the high-altitude CRONICAS site with substantial Native South American (Quechua and Aymara) ancestry. We evaluated previously validated PGS from large, multi-ancestry, genome-wide association studies for 10 traits: systolic blood pressure (SBP), diastolic blood pressure (DBP), LDL-C, HDL-C, triglycerides, waist circumference, BMI, fasting glucose, fasting insulin, and type 2 diabetes mellitus (T2DM). We also tested PGS associations with clinically defined categorical variables for lipids, waist circumference, and blood pressure. All PGS were mean-centered and standardized. Models were adjusted for age, sex, and ancestry-informative principal components. Bonferroni correction accounted for testing across phenotypes. We estimated ancestry proportions using ADMIXTURE software. Results: Participants were predominantly of Native South American ancestry (93.2%), with minor European (5.8%), African (0.7%), and East Asian (0.2%) contributions. PGS were significantly associated (P < 0.0050) with several cardiometabolic traits after adjustment: SBP increased by 3.80 mmHg (95% CI 2.72–4.88), DBP by 2.78 mmHg (2.11–3.46), LDL-C by 9.38 mg/dL (6.68–12.07), HDL-C by 3.34 mg/dL (2.53–4.15), triglycerides by 13.07 mg/dL (6.72–19.42), and waist circumference by 1.94 cm (1.09–2.78) per 1-SD increase in their respective PGS. PGS for BMI, fasting glucose, fasting insulin, and T2D were not significantly associated with their respective phenotypes. PGS for SBP, LDL-C, HDL-C, and triglycerides were also associated with their respective clinically defined categorical variables, according to expected gradients. Conclusion: In a Peruvian cohort with predominantly Native South American ancestry, PGS derived from multi-ancestry datasets were strongly associated with multiple cardiometabolic phenotypes. These findings underscore the transferability of PGS based on cross-population genetic studies and the need for broader South American representation in genetic studies to further improve PGS performance in these distinct populations.