Figure 3

Evaluation of the outcome predictor of "ALL Relapse" derived from the mechanism-anchored expression profile obtained by PGnet. The 87 leukemia patients with "CCR" or "relapse" information were randomly divided into three folds, two of which were used to identify the predictor of outcome ("CCR" vs. "relapse", Figure 2c). The predictor consisted of GEMs and ESGs associated to "relapse" to train a linear SVM model, and the remaining one was used as a blinded test set. Three-fold cross-validations were repeated 100 times. The resulting Receiver Operating Characteristic (ROC) curve, the area under the curve (AUC) and corresponding p-values were calculated by Bioconductor package verification. Horizontal and vertical "error bars" represent the 95% confidence intervals of the predictor. Regions where the error bars are above the diagonal line represent a better prediction than chance. Overall, the AUC was significantly different than that of a random predictor of "ALL relapse" (P = 1.6%).