Table 4.

Means and standard derivations of AUC, accuracy (ACC), sensitivity (SEN) and specificity (SPE) for 5-fold cross validation on van de Vijver et al. [23] (top) and Wang et al [24] (bottom) for netSVM, other network-based methods and gene-based methods.

	network-based method				gene-based method			previous study
	net SVM	F∞-norm SVM	Larsnet	Chuang's method*	SVM	Lasso	t-test + LDA	70 genes [23]*	76 genes [24]*
AUC	0.68 ±	0.60 ±	0.69 ±	-	0.63 ±	0.61 ±	0.64 ±	-	-
	0.02	0.02	0.02		0.02	0.02	0.02
ACC	0.65 ±	0.62 ±	0.67 ±	0.70	0.63 ±	0.53 ±	0.64 ±	0.62	-
	0.02	0.02	0.02		0.02	0.02	0.02
SEN	0.51 ±	0.50 ±	0.71 ±	0.90	0.49 ±	0.73 ±	0.44 ±	0.93	-
	0.04	0.02	0.03		0.04	0.02	0.02
SPE	0.74 ±	0.67 ±	0.67 ±	0.63	0.71 ±	0.47 ±	0.71 ±	0.53	-
	0.03	0.02	0.02		0.03	0.02	0.02
AUC	0.73 ±	0.64 ±	0.70 ±	-	0.72 ±	0.68 ±	0.60 ±	-	-
	0.02	0.01	0.02		0.02	0.02	0.02
ACC	0.71 ±	0.63 ±	0.68 ±	0.72	0.70 ±	0.68 ±	0.59 ±	-	0.62
	0.02	0.01	0.02		0.02	0.02	0.02
SEN	0.42 ±	0.61 ±	0.65 ±	0.90	0.42 ±	0.63 ±	0.49 ±	-	0.93
	0.04	0.02	0.03		0.04	0.03	0.02
SPE	0.81 ±	0.64 ±	0.71 ±	0.62	0.80 ±	0.72 ±	0.65 ±	-	0.53
	0.02	0.02	0.02		0.02	0.02	0.02

*: results are extracted from the study [14].