Abstract
Despite advances in cardiovascular disease risk stratification, traditional risk prediction models often fail to identify high-risk individuals before adverse events occur, underscoring the need for more precise tools. Polygenic risk scores (PRS) quantify genetic susceptibility by aggregating genetic variants but face challenges in practice. This systematic review investigates how artificial intelligence (AI) and machine learning algorithms can optimize PRS (AI-optimized PRS) to improve cardiovascular disease prediction. Analyzing 13 studies, we found that AI-optimized PRS models enhance predictive accuracy by improving feature selection, handling high-dimensional data, and integrating diverse variables-including clinical risk factors, biomarkers, imaging, and combining multiple PRS. These models outperform nonoptimized PRS models, providing a more comprehensive understanding of individual risk profiles. Evidence suggests that AI-optimized PRS can better stratify patients and guide personalized prevention strategies. Future research is needed to explore sex differences, include diverse populations, integrate AI-optimized PRS into electronic health records, and assess cost-effectiveness.