Abstract P049: Blood Pressure Polygenic Risk Score Stratifies Individuals With High Risk Of Hypertension In Different Ancestries

Abstract

The heritability of arterial hypertension ranges from 25 to 68%, but the underlying genetic determinants remain elusive. Thousands of variants with small effects on systolic blood pressure (SBP) from European Ancestry Genome-Wide Association Studies have been used to derive genetic risk scores, whose predictive power in samples from diverse ancestries has yet to be determined. We used the LDPred2 algorithm and 423 thousand genetic variants from UK Biobank that were common to an admixed sample population from Brazil to derive SBP polygenic risk scores (PRS). We selected the best model that displayed a score with hyper-parameters ρ = 0.056, h 2 = 0.0739 and without sparsity, which presented the highest correlation coefficient with observed SBP (0.206) and SBP change per standard deviation increase (4.04mmHg) in a different set of data from the UK Biobank prior to test its predictive power on SBP and hypertension in a third set of the UK Biobank and two admixed Brazilian population samples. All three populations displayed similar normal distribution of PRS, with the UK Biobank PRS distribution slightly shifted to the right, presenting higher PRS average (7.8 vs. 5.6 and 6.4 in Baependi and Pelotas, respectively), but with similar standard deviation (3.3 vs. 3.2 for both Baependi and Pelotas). We used the PRS deciles to stratify the populations and observed a significant difference on SBP average between the top and bottom deciles in all three populations (14.43 mmHg, 9.34 mmHg and 7.51 mmHg for the UK Biobank, Baependi and Pelotas, respectively). We then compared the top 10% PRS versus the remaining 90% of the PRS distribution, and observed that high PRS is associated with 1.8, 2 to 3 and 3 to 4-fold increase in prehypertension, hypertension stage I and hypertension stage II risk, respectively, demonstrating that the increased risk is independently of age and ancestry. Considering the highly admixed nature of the Brazilian population, we provide evidence for the generalization of PRS between populations with different genetic structure.