Think of a polygenic risk score as a “credit score for heart disease” built from thousands of tiny changes in your DNA. This paper reviews how AI can act like a smarter credit bureau—sifting through massive genomic and clinical datasets to build more accurate and personalized scores that predict who is at high risk of heart problems, long before symptoms start.
Traditional cardiovascular risk tools (like Framingham scores) don’t fully capture genetic risk, especially when thousands of small genetic effects interact in complex ways and differ across populations. The reviewed AI methods aim to improve how polygenic risk scores (PRS) are constructed and applied to better predict cardiovascular events, personalize prevention and treatment, and reduce missed high‑risk patients and over‑treatment of low‑risk ones.
Access to large, well‑phenotyped genomic–cardiology cohorts (biobanks, EHR‑linked registries), robust validation across diverse populations, and tight integration of PRS outputs into clinical workflows and decision-support systems form the main defensible advantages; algorithmic techniques alone are becoming increasingly commoditized.
Classical-ML (Scikit/XGBoost)
Structured SQL
High (Custom Models/Infra)
Need for very large, well‑curated, multi-ethnic genomic and clinical datasets plus stringent validation and regulatory evidence to move models from retrospective studies into routine clinical practice.
Early Adopters
Unlike generic PRS work, this systematic review is focused specifically on the intersection of genomics and cardiology clinical practice, emphasizing how AI/ML can optimize polygenic risk scores for real-world cardiovascular decision-making (screening, prevention, and treatment) and highlighting practical issues such as performance across ancestries, integration with traditional clinical risk factors, and translational barriers.