Fig. 1

Classification of four different healthy tissue types within TCGA dataset: breast, lung, head and neck, and kidney. Five-fold cross-validation was performed 10 times for 171 KEGG and BioCarta pathways using each of five methods. In each panel, the pathways are ordered according to the classification performance of GRAPE (green). The viewing range is lower bounded by 0.6, as only PC had any pathways (n=15) that failed to achieve this threshold. b Proportion of pathways for which each method had largest accuracy. If multiple methods achieved the max (within.001 counted), the credit for the pathway was fractionated. c Proportion of pathways that achieved classification accuracy within 0.05 of the best performance (left) and proportion that achieve at least 0.90 classification accuracy (right)