Table 4 Prediction accuracy of integrative models in terms of AUC from 5-fold 20 CV
Model1Mean AUC2SD3Proportion of times model in row (column) had AUC > model in column (row) over 20 CV
AgeTmzSNPMethGeCNVAge +
TmzAge +
Tmz + SNPAge +
Tmz + MethAge + Tmz + GeAge +
Tmz + CNVAge + Tmz + Ge + Meth + SNP + CNV
Age + Tmz0.7050.0630.71>0.95>0.95>0.95>0.95>0.950.420.080.250.380.33
Age + Tmz + SNP0.7060.0650.71>0.95>0.95>0.95>0.95>0.950.580.080.290.380.33
Age + Tmz + Meth0.7170.0730.79>0.95>0.95>0.95>0.95>0.950.920.920.500.710.46
Age + Tmz + Ge0.7180.0580.92>0.95>0.95>0.95>0.95>0.950.750.710.500.710.71
Age + Tmz + CNV0.7060.0640.79>0.95>0.95>0.95>0.95>0.950.630.630.290.290.29
Age + Tmz + Ge + Meth + SNP + CNV0.7150.0670.82>0.95>0.95>0.95>0.95>0.950.430.640.370.430.71
  • 1 Models used for analysis, described on Table 2; 2Mean of prediction accuracy measured by AUC by using 20 5-fold cross-validations for each model; 3Standard deviation of AUC.