Table 4 Model of metabolites in MDD with full remission using different types of machine learning algorithm.

Model metabolite^a	Machine learning model	Training model				Testing model
Model metabolite^a	Machine learning model	AUC	P_{permutation test}^b	Predictive accuracy	P_{permutation test}	Predictive accuracy	Sensitivity	Specificity	Positive predictive value	Negative predictive value
Succinic acid Proline Acetic acid Creatine Glutamine Glycine Pyruvic acid Histidine	Linear SVM	0.784	0.007	0.707	0.011	0.846	0.846	0.846	0.733	0.917
	PLS-DA	0.779	0.003	0.705	0.011	0.846	0.923	0.808	0.706	0.955
	Random FOREST	0.738	0.007	0.677	0.029	0.821	0.769	0.846	0.714	0.880
	Logistic regression	0.772	0.004	0.715	0.005	0.821	0.769	0.846	0.714	0.880

MDD major depressive disorder, AUC area under the receiver operating characteristic curve, SVM support vector machine, PLS-DA partial least squares-discriminant analysis.
^aMetabolites for which p < 0.05 were selected.
^b1000 random permutations were performed for validation testing.

Quick links

Search