Table 3.

Discrimination of vancomycin-susceptible E. faecium (VSE) from vancomycin-resistant E. faecium (VRE) isolates and, within the latter group, differentiation between the strains hosting the vanA and vanB resistance genes. The actual classification is shown in columns and the predictive classification in rows. (A) Results from the support vector machine (SVM) algorithm using the full-spectrum-TIC method. Accuracy: 80.9%; F1 score: 80.46%; sensitivity: 81.4% specificity: 80.43%; positive predictive value (PPV) or precision: 79.55%; negative predictive value (NPV): 82.22%. (B) Scores from the partial least squares–discrimination analysis (PLS-DA) algorithm using the total ion current normalization after peak finding (TICp) method. Accuracy: 86.05%; predictive value for VanA VRE: 80.65%; predictive value for VanB VRE: 89.09%.

Actual Classification	Predicted Classification
(A) SVM Full Spectrum	VRE	VSE
VRE	70	16
VSE	18	74
(B) PLS Threshold TICp	VanA	VanB
VanA	25	6
VanB	6	49