Table 4.

Sensitivity and specificity of ¹H NMR metabolomics models to detect Bovine Respiratory Disease in feedlot cattle defined with six reference diagnosis methods.

Diagnosis method	Dataset	Sensitivity	Specificity	Accuracy	AUCa	N leavesb	N componentsc	Component Number Metabolite ID
Visual Diagnosis	Training	0.81	0.93	0.87	0.87	2	1	92
	Validation	0.82	0.87	0.85	0.85	2	1	Unknown (Singlet at 5.39 ppm)
Temperature Diagnosis	Training	0.76	0.88	0.85	0.82	2	1	34
	Validation	0.52	0.77	0.69	0.65	2	1	Phenylalanine
Lung Auscultation Diagnosis	Training	0.80	0.73	0.77	0.76	2	1	123
	Validation	0.77	0.45	0.64	0.61	2	1	Lactate
Clinical Diagnosis	Training	0.77	0.85	0.79	0.81	2	1	227
	Validation	0.79	0.54	0.70	0.67	2	1	Glutamine
Visual-Clinical Diagnosis	Training	0.99	0.88	0.93	0.94	4	3	55, 211, 158
	Validation	0.88	0.74	0.81	0.83	4	3	Tyrosine, Citrate, Hydroxybutyrate
Lung Lesion Diagnosis	Training	0.76	0.97	0.92	0.90	6	5	219, 130, 292, 305, 25
	Validation	0.38	0.89	0.74	0.71	6	5	Citrate, Unknown, Unknown, Leucine, Unknown

^aAUC = Area Under the Curve.

^bN leaves = number of leaves in final pruned tree.

^cN components = number of components or metabolite components selected in final pruned tree by the classification tree analysis.