Li-Fraumeni syndrome (LFS) confers high lifetime cancer risk due to germline TP53 pathogenic variants (PV). A comprehensive surveillance regimen termed the 'Toronto Protocol', has been adopted for early tumor detection, demonstrating improved survival among TP53 PV carriers. However, the protocol's "one-size-fits-all" approach fails to consider individual cancer risk. To personalize screening, we developed a support vector machine model to predict early onset of primary tumors (age < 6) using peripheral blood methylation data of TP53 PV carriers (n = 237). Validation (n = 64) and external testing (n = 79) showed AUROC = 0.928 [0.835-1.000], F1-score = 0.692 [0.435-0.867], and NPV = 0.984 [0.946-1.000]. The model achieved 91% accuracy, correctly classifying 90% of patients with cancer before the age of six and 87% of cancer-free individuals in the external test set. Our tool enables risk stratification for early-onset malignancies, to optimize clinical surveillance and improve patient outcomes.
© 2025. The Author(s).