Four gene sets that satisfied p < 0.001 in both the Affymetrix and Illumina datasets and their p-values, FDR, accuracy and AUC.

P-values and FDR were calculated using the “gage” R package, and the accuracy was obtained from leave-one-out cross validation. AUC generated by K-OPLS for scoring classifiers of four gene sets.

P值（P-values）与错误发现率（False Discovery Rate，FDR）通过"gage"R包计算得到，模型准确率通过留一交叉验证（leave-one-out cross validation）获取。针对4个基因集的分类器评分任务，其受试者工作特征曲线下面积（Area Under Curve，AUC）由核正交偏最小二乘（K-OPLS）生成。