Table 2. Comparison of the prediction performance of the Gaussian naïve Bayes (GNB)-based wrapper, logistic regression (LogR)-based wrapper, decision tree (DT)-based wrapper, k-nearest neighbor (KNN)-based wrapper, and two support vector machine (SVM)-based wrappers with the RBF and polynomial kernels (denoted as SVM-RBF and SVM-Poly respectively).

Wrapper method	Five-fold CV (average of 10 runs)				Jackknife test
	Sen	Spe	Acc	MCC	Sen	Spe	Acc	MCC
GNB	0.815±0.010	0.767±0.009	0.791±0.007	0.583±0.014	0.828	0.781	0.805	0.610
DT	0.716±0.019	0.704±0.025	0.710±0.011	0.421±0.021	0.684	0.700	0.692	0.384
LogR	0.801±0.008	0.699±0.005	0.750±0.006	0.502±0.012	0.805	0.704	0.754	0.511
KNN	0.716±0.015	0.770±0.010	0.743±0.008	0.487±0.016	0.721	0.771	0.746	0.492
SVM-Poly	0.867±0.008	0.668±0.011	0.768±0.009	0.547±0.019	0.855	0.687	0.771	0.550
SVM-RBF	0.830±0.013	0.746±0.006	0.788±0.008	0.578±0.016	0.848	0.754	0.801	0.605

Note: The CV tests were based on ten runs and the averages and the standard deviations are shown. The highest values are shown in bold.