Background: Li-Fraumeni syndrome (LFS) is an inherited cancer syndrome mainly caused by a deleterious mutation in TP53. An estimated 48% of LFS patients present due to a deleterious de novo mutation (DNM) in TP53. The knowledge of DNM status, DNM or familial mutation (FM), of an LFS patient requires genetic testing of both parents which is often inaccessible, making de novo LFS patients an understudied population. Famdenovo.TP53 is a Mendelian Risk prediction model used to predict DNM status of TP53 mutation carriers based on the cancer-family history and several input genetic parameters, including disease-gene penetrance. The good predictive performance of Famdenovo.TP53 was demonstrated using data collected from four historical US cohorts. We hypothesize that by incorporating penetrance estimates that are specific for different types of cancers diagnosed in family members, we can develop a model with further improved calibration, accuracy and prediction. Methods: We present Famdenovo.CS, which uses cancer-specific penetrance estimates that were derived previously using a Bayesian semi-parametric competing risk model, to calculate the DNM probability. We use our model to analyze 101 families recently collected from the Clinical Cancer Genetic program at MD Anderson Cancer Center (CCG-TP53) that includes 20 families with known DNM status and 81 families with unknown DNM status. We used the concordance index (AUC), observed:expected ratios (OE) and Brier score (BS) to measure our model’s discrimination, calibration and accuracy, respectively. We estimate the proportion of probands that present a DNM and compare DNM to FM carriers in several areas including: cancer types diagnosed, age at diagnosis, number of primary cancers diagnosed, sex, amino acid change caused by mutation in TP53. Results: Famdenovo.CS showed equally good discrimination and calibration performance to Famdenovo.TP53, while improving the overall accuracy, demonstrated by a decrease in the Brier score of -0.09 (95% CI: [-0.02, -0.19]). Of the 101 probands in the CCG-TP53 cohort, we predict 39 to be DNMs and 62 to be FMs. The cancer types and ages of diagnosis observed in FMs and DNMs are similarly distributed. Conclusions: Famdenovo.CS shows improved model accuracy in the CCG cohort. DNMs in TP53 are a prevalent cause of LFS and we did not find differences in the clinical characteristics of DNM and FM carriers. Our model allows for a systematic identification and characterization of TP53 DNM carriers.