We evaluated the performance of various polygenic risk score (PRS) models derived from European (EU), South Asian (SA), and Punjabi Asian Indians (AI) studies on 13,974 subjects from AI ancestry. While all models successfully predicted Coronary artery disease (CAD) risk, the AI, SA, and EU + AI were superior predictors and more transportable than the EU model; the predictive performance in training and test sets was 18% and 22% higher in AI and EU + AI models, respectively than in EU. Comparing individuals with extreme PRS quartiles, the AI and EU + AI captured individuals with high CAD risk showed 2.6 to 4.6 times higher efficiency than the EU. Interestingly, including the clinical risk score did not significantly change the performance of any genetic model. The enrichment of diversity variants in EU PRS improves risk prediction and transportability. Establishing population-specific normative and risk factors and inclusion into genetic models would refine the risk stratification and improve the clinical utility of CAD PRS.

Graphical abstract

中文翻译：

---

亚洲印度人冠状动脉疾病的多基因风险评分评估

我们评估了源自欧洲 (EU)、南亚 (SA) 和旁遮普亚洲印第安人 (AI) 对 13,974 名 AI 血统受试者的研究中衍生的各种多基因风险评分 (PRS) 模型的性能。虽然所有模型都成功预测了冠状动脉疾病 (CAD) 风险，但 AI、SA 和 EU + AI 是更优秀的预测因子，并且比 EU 模型更易于移植； AI 和 EU + AI 模型的训练和测试集预测性能分别比 EU 高 18% 和 22%。比较具有极端 PRS 四分位数的个体，AI 和 EU + AI 捕获具有高 CAD 风险的个体的效率比 EU 高 2.6 至 4.6 倍。有趣的是，包括临床风险评分并没有显着改变任何遗传模型的性能。 EU PRS 中多样性变体的丰富提高了风险预测和可转移性。建立特定人群的规范和风险因素并将其纳入遗传模型将完善风险分层并提高 CAD PRS 的临床效用。

图形概要