Prediction performance. A) Raw prediction scores (breast cancer >0> non-breast cancer). Subjects with breast cancer are indicated by red bars and healthy subjects are indicated by green bars. Among the 67 breast cancer patients, 54 were correctly predicted, while 47 of the 60 healthy samples were assigned to the correct class. False negatives (n = 13) and false positives (n = 13) can be seen in the centre of the figure. Bars marked with * are samples from pregnant women. B) ROC curve based on the double cross-validation results Prediction of the 127 samples based on the 738 probe list. The prediction accuracy is 79.5% and the area under curve (AUC) is 0.88 reflecting a good separation of the two groups.