Table 4. The performance of Support vector machine (SVM) and Random Forest (RF) models developed using different sets of features selected via SVM technique on training and independent or external validation dataset.

Features	Dataset	Technique	Performance Measures
			Sensitivity	Specificity	Accuracy (%)	MCC	ROC
setA-2 (5 genes)	Training	SVM	68.9	73.94	70.88	0.42	0.75
	Validation		65.08	70.73	67.31	0.35	0.75
	Training	RF	81.5	56.97	71.84	0.4	0.73
	Validation		77.78	46.34	65.38	0.25	0.68
setB-2 (5 genes)	Training	SVM	68.9	70.91	69.69	0.39	0.76
	Validation		60.32	70.73	64.42	0.3	0.72
	Training	RF	71.65	64.85	68.97	0.36	0.71
	Validation		69.84	56.1	64.42	0.26	0.70
Combo-2 (10 genes)	Training	SVM	72.44	72.89	72.62	0.45	0.78
	Validation		71.43	68.29	70.19	0.39	0.77
	Training	RF	76.19	65.85	72.12	0.42	0.76
	Validation		70.47	70.3	70.41	0.4	0.76