Evaluation of the 500 gene signatures. Each of the 100-gene signatures for 80 randomly selected tumors in the training set was used to predict relapsed patients in the corresponding test set. Its performance was measured by the AUC of the ROC analysis. (A) Performance of the gene signatures for ER-positive patients in test sets. (B) Performance of the gene signatures for ER-negative patients in test sets. (Left) Frequency of AUC in 500 prognostic signatures panels as derived following the flow chart presented in Figure 1. (Right) Frequency of AUC in 500 random gene lists. To generate a gene list as a control, the survival data for the ER-positive patients or ER-negative patients was permutated randomly and reassigned to the chip data.