Figure 5. Error rates calculated within the identified classifiability sub-categories.

Boxplots of the leave-two-out cross-validation balanced error-rates calculated within each of the two identified classifiability sub-categories for each cancer for models built using DLDA [16] in the gene expression data and logistic regression in the clinical data in the a. melanoma, b. breast cancer and c. colon cancer cohorts. For the gene expression data boxplots, two completely independent cross-validated classification schemes were performed within both of the two patient groups: a) with and without tumor extranodal spread; b) with and without hormonal therapy; c) with and without adjuvant chemotherapy.