Table 1

List of metabolites selected based on the multivariate elastic net analysis comprising the biomarker signature and their analysis of variance results

Metabolite name	Trainings set			Test set
Metabolite name	Fold change	p Value	FDR	Fold change	p Value	FDR
CA19-9	18.36	6.89E-09	7.65E-06	14.27	3.17E-09	1.07E-06
Proline	0.69	2.24E-05	0.0027	0.75	0.0001	0.0082
Sphingomyelin (d18:2,C17:0)	1.15	0.005612	0.0400	1.15	0.0119	0.0696
Phosphatidylcholine (C18:0,C22:6)	1.26	8.59E-05	0.0034	1.06	0.2091	0.4619
Isocitrate	1.26	0.008074	0.0518	0.99	0.9159	0.9377
Sphinganine-1-phosphate (d18:0)	0.79	0.025867	0.1175	0.85	0.0705	0.2430
Histidine	0.77	0.000324	0.0073	0.79	0.0004	0.0109
Pyruvate	0.93	0.367408	0.6114	0.97	0.6479	0.7976
Ceramide (d18:1,C24:0)	0.79	0.001509	0.0167	0.80	0.0087	0.0583
Sphingomyelin (d17:1,C18:0)	1.36	4.86E-05	0.0029	1.37	4.61E-05	0.0078

Univariate statistical analysis was done by a linear model on log10-transformed data with disease, gender, body mass index, age and storage time as fixed effects. Multiple testing was addressed by calculating the false discovery rate (FDR) described by Benjamini and Hochberg.