Figure 4

Method for predicting the clinical outcome using MCCV. Prediction performance of signatures from 35 miRNAs (Table 2 and 3). Left: Mean accuracy, specificity, and sensitivity (% in vertical axis) as a function of the training set size were determined for the 100 random splits of patients (non-SVR (NR+R) vs. SVR). Right: prediction performance of the training set size was determined by performing 48 random splits of patients (NR vs. R). A. Prediction performance (mean accuracy, specificity, and sensitivity) and number of miRNAs which was used for the prediction are also shown in each training set (TS) and validation set (VS).