Fig. 1

Study design and flowchart for coronary artery disease (CAD). A Selection of PRS in the case–control study. B The cohort study. To select the parameters for each method with the best discrimination based on the area under the curve (AUC), clumping and thresholding, LDpred, lassosum, PRS-CS, sBayesR, LDpred-funct, and DBSLMM were used to calculate polygenic risk scores (PRSs) on the case–control set consisting of prevalent cases. For these calculations, summary data for three genome-wide association studies (GWAS) on CAD (CARDIoGRAMplusC4D, Finngen Biobank, Japan Biobank) that excluded the UK Biobank and data on linkage disequilibrium were used. The calculated PRSs were applied to a nonoverlapping set of participants from the UK Biobank with no preexisting CAD, aged 40 to 69 at baseline, and who were followed up for incident CAD events. In this population, the pooled cohort equations (PCE) model was calculated and different models (PRS, PCE, PRS-enhanced PCE) were compared in terms of their predictive accuracy based on discrimination, calibration, and reclassification metrics