Blood Lipid Polygenic Risk Score Development and Application for Atherosclerosis Ultrasound Parameters

Abstract

Background: The present study investigates the feasibility of using three previously published genome-wide association studies (GWAS) results on blood lipids to develop polygenic risk scores (PRS) for population samples from the European part of the Russian Federation. Methods: Two population samples were used in the study - one from the Ivanovo region (n = 1673) and one from the Vologda region (n = 817). We investigated three distinct approaches to PRS development: using the straightforward PRS approach with original effect sizes and fine-tuning with PRSice-2 and LDpred2. Results: In total, we constructed 56 PRS scales related to four lipid phenotypes: low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, total cholesterol, and triglyceride levels. Compared with previous results for the Russian population, we achieved an additional R² increase of 2-4%, depending on the approach and lipid phenotype studied. Overall, the R² PRS estimates approached those described for other populations. We also evaluated the clinical utility of blood lipid PRS for predicting carotid and femoral artery atherosclerosis. Specifically, we found that PRS for total cholesterol, low-density lipoprotein cholesterol, and triglycerides were positively correlated with ultrasound parameters of carotid and femoral artery atherosclerosis (ρ = 0.09-0.13, p < 0.001), whereas PRS for high-density lipoprotein cholesterol were inversely correlated with the number of plaques in the femoral arteries (ρ = -0.08, p = 8.71 × 10-3). Conclusions: PRS fine-tuning using PRSice-2 add LDpred2 improves the performance of blood lipid PRS. Our study demonstrates the potential for further use of blood lipid PRS for prediction of atherosclerosis risk.

Keywords: atherosclerosis; cardiovascular disease; high-density lipoprotein cholesterol; low-density lipoprotein cholesterol; polygenic risk scores; total cholesterol; triglycerides.

Figure 1

Comparison of R² distributions between training and validation data for (A) ESSE-Ivanovo HDL-C phenotype and (B) ESSE-Vologda logTG phenotype. No significant difference of means was shown for the ESSE-Ivanovo HDL-C phenotype, and a significant difference was shown for the ESSE-Vologda logTG phenotype.

Figure 2

Comparison of R² estimates across all developed PRS scales. For PRSice-2, R² estimates from both training and validation datasets are given. Estimates are provided along with 95% CI. HDL-C—high-density lipoprotein cholesterol, LDL-C—low-density lipoprotein cholesterol, TC—total cholesterol, logTG—logarithmic triglycerides.