Fig. 3

Performance of risk prediction models for hereditary breast cancer. A DrABC performed better than previous models in predicting germline pathogenic variants (GPVs) in any cancer predisposition genes (CPGs) (AUCs of 0.74 for DrABC, 0.65 for BRCAPRO, 0.57 for BOADICEA, 0.56 for Myriad, and 0.61 for PENNII). B In predicting GPVs in BRCA1/2, the AUC of DrABC was 0.79 (95% CI, 0.74–0.85) for the validation cohort, which was superior to those for previous models (0.70 for BRCAPRO, 0.59 for BOADICEA, 0.59 for Myriad, and 0.63 for PENN II). C, D The probabilities generated by DrABC were distributed differently between non-carriers and patients with GPVs in any CPG (C) or BRCA1/2 (D). **p < 0.01, ****p < 0.0001, when comparing with the DrABC