Fig. 5.

Grouped dotplots of the average cross-validation resampling results (representative example using a specific random seed), for the 3 groups of variables: only the 94 hub genes, only the 8 PET variables and the combination of both (ROC, Sensitivity, Specificity-10 fold cross-validation, repeated 10 times for 10 different random seeds). Using the total merged microarray dataset as the training set, the composite signature yields overall better performance measures than either PET, or genes separately.