Early antidepressant treatment response prediction in major depression using clinical and TPH2 DNA methylation features based on machine learning approaches

Table 4 Averaged prediction metrics for each classifier based on TPH2 CpGs

Feature selection	Classifier	ROC (95%CI)	F-Measure	G-Mean	Accuracy	Sensitivity	Specificity	PPV	NPV
Without RFE	Logistic	0.524 (0.502–0.546)	0.661	0.432	0.544	0.720	0.259	0.612	0.363
	Rpart	0.530 (0.517–0.542)	0.677	0.472	0.569	0.731	0.305	0.630	0.411
	SVM-RBF	0.516 (0.503–0.528)	0.762	0.045	0.615	0.994	0.002	0.618	0.171
	LogitBoost	0.571 (0.559–0.583)	0.691	0.536	0.599	0.717	0.401	0.660	0.466
	RF	0.590 (0.578–0.603)	0.750	0.407	0.628	0.901	0.184	0.642	0.534
With RFE	Logistic	0.519 (0.498–0.540)	0.710	0.378	0.580	0.831	0.172	0.619	0.386
	Rpart	0.568 (0.556–0.581)	0.683	0.522	0.588	0.716	0.381	0.652	0.453
	SVM-RBF	0.526 (0.514–0.539)	0.760	0.160	0.617	0.981	0.026	0.620	0.458
	LogitBoost	0.608 (0.596–0.620)	0.725	0.574	0.638	0.756	0.436	0.685	0.524
	RF	0.666 (0.654–0.678)	0.763	0.558	0.670	0.859	0.363	0.686	0.614

ISSN: 1471-244X