Skip to main content
[Preprint]. 2023 Sep 2:2023.08.31.23294849. [Version 1] doi: 10.1101/2023.08.31.23294849

Figure 3:

Figure 3:

ROC curves, and the 95% bootstrapped confidence intervals of the AUCs, for predictive performance of the CS and MPC models on the CCG cohort under two scenarios: (a), (c) all validation subjects are included, and (b), (d) only known and inferred mutation carriers and wildtypes are included. For comparison, the KM method is used to predict at least one cancer versus no cancer. Sample sizes in scenario (a), (c): n(unaffected) = 1264, n(SPC) = 259, n(MPC) = 31, n(BR) = 94, n(SA) = 18, n(OT) = 220, n(D) = 497, n(A) = 879. Sample sizes in scenario (b), (d): n(unaffected) = 907, n(SPC) = 180, n(MPC) = 27, n(BR) = 69, n(SA) = 16, n(OT) = 157, n(D) = 379, n(A) = 617. BR: breast cancer, SA: sarcoma, OT: all other cancer types combined, D: death, A: alive.