Abstract

Polygenic risk scores (PRSs) for coronary artery disease (CAD) hold promise for improving early risk 
stratification and preventive cardiology, yet their clinical translation across diverse populations remains 
challenging. This paper examines the transferability, calibration, and clinical decision thresholds of CAD PRSs in 
Latinx populations from a bench-to-population perspective. It synthesizes current evidence on how differences in 
genetic ancestry, linkage disequilibrium structure, allele frequencies, and environmental context influence 
predictive performance when PRSs derived largely from European genome-wide association studies are applied to 
Latinx cohorts. The analysis highlights that while some multi-ancestry models retain a partial predictive signal, 
performance often declines, with systematic miscalibration leading to over- or underestimation of absolute risk. 
Recalibration methods, including adjustment of intercepts and slopes and incorporation of population-specific 
prevalence data, can improve clinical reliability. Determining appropriate decision thresholds is also critical, as 
differences in disease prevalence and baseline risk necessitate ancestry-informed cut-points to balance sensitivity 
and specificity in screening and prevention strategies. The paper further discusses methodological, ethical, and 
implementation considerations, including dataset representation, governance of genomic data, and integration 
with traditional cardiovascular risk assessment tools. Strengthening Latinx representation in genomic research 
and developing ancestry-aware modelling strategies are essential to ensure equitable clinical utility. Ultimately, 
optimizing PRS transferability, calibration, and threshold selection could enable more accurate and inclusive 
genomic risk prediction for CAD in Latinx populations.