Data-driven decision-making for precision diagnosis of digestive diseases

Table 5 Data-driven precision diagnosis in digestive diseases based on metabolomics

First author, year	Disease	n	Data source and specific task	ML method	Diagnostic performance	Refs.
Jiménez, 2013	CRC	26	Metabolic profiles of tumor and adjacent tissues by NMR spectroscopy; classification based on discriminant metabolites	OPLS–DA	AUC: 0.910	[134]
Yuan, 2021	ESCC	525	Serum lipidomics data; classification based on a panel of 12 lipid biomarkers, age and gender	SVM/PCA	AUC: 0.818–0.966	[135]
Takis, 2018	Diffuse abdominal pain	64	Serum metabolomics data by NMR spectroscopy; classification by metabolomics fingerprint	OPLS–DA/PCA	Accuracy > 90%	[136]
Wang, 2023	ESCC	1104	Serum metabolomics data by LC–MS; classification based on digital images of metabolome profiles	CNN	AUC: 0.950	[137]
Yang, 2022	CRC	99	LC–MS-based plasma lipidomics data; classification based on 14 lipids	PLS/RF/SVM/KNN	Accuracy: 72.6–100%	[138]
Huang, 2021	GC	400	Untargeted metabolomics data of plasma; classification based on 6 metabolites with clinical indicators	LR/RF	AUC: 0.830	[139]
Yu, 2023	GC	301	Serum metabolomics data by MS; classification based on 12 differential metabolites	PCA/SVM/RF/LASSO	AUC: 0.893	[140]
Matsumoto, 2023	GC	101	Hydrophilic metabolites quantified by LC–TOFMS; classification based on 3 metabolites	SVM	AUC: 0.885–0.915	[141]
Pan, 2022	GC	280	Target bile acid metabolomics data of serum; classification based on 6 bile acids	RF/LASSO/OPLS–DA	AUC: 0.940–1.000	[142]
Zhao, 2022	ESCC	239	Multi-platform metabolomics data of serum; classification based on 5 metabolites	RF/LASSO/PCA	AUC: 0.873 (95% CI, 0.825–0.925)	[143]

Full names of abbreviations are given in the Abbreviations section of the manuscript

ISSN: 1475-925X