Table 1.

Model characteristics.

	Our model (HSIC Lasso + SVM/KR)	HSIC Lasso	Lasso	SVM/KR	Random forest	Partial least squares	Sparse partial least squares	Multiple linear/logistic regression
Prediction	Performed	Not performed	Performed	Performed	Performed	Performed	Performed	Performed
Feature selection	Performed	Performed	Performed	Not performed	Not performed	Not performed	Performed	Not performed
Nonlinear association between variables	Detectable	Detectable	Not detectable	Detectable	Detectable	Not detectable	Not detectable	Not detectable

Notes: Preferable properties for handling metabolomic data are in bold.

HSIC Lasso Hilbert–Schmidt Independence Criterion Lasso, SVM support vector machine, KR kernel regression.