Table 4.

The results of repeated 10-fold cross-validation experiments for predicting treatment response using multilayer feedforward neural networks (MFNNs) and logistic regression with 16 biomarkers and 6 clinical biomarkers only.

Algorithm	AUC	Sensitivity	Specificity	Number of biomarkers
MFNN with 1 hidden layer	0.8211 ± 0.0571	0.7496 ± 0.0579	0.6775 ± 0.0731	16
MFNN with 2 hidden layers	0.8228 ± 0.0571	0.7546 ± 0.0619	0.6922 ± 0.0765	16
MFNN with 3 hidden layers	0.8220 ± 0.0570	0.7535 ± 0.0611	0.6951 ± 0.0731	16
Logistic Regression	0.8168 ± 0.0553	0.7493 ± 0.0626	0.7066 ± 0.0785	16
MFNN with 1 hidden layer	0.5597 ± 0.0808	0.6081 ± 0.0113	0.3919 ± 0.0113	6
MFNN with 2 hidden layers	0.5606 ± 0.0836	0.6081 ± 0.0113	0.3919 ± 0.0113	6
MFNN with 3 hidden layers	0.5571 ± 0.0788	0.6081 ± 0.0113	0.3919 ± 0.0113	6
Logistic Regression	0.5374 ± 0.0762	0.5881 ± 0.0432	0.4112 ± 0.0418	6

AUC, the area under the receiver operating characteristic curve.

Data are presented as mean ± standard deviation.