. 2023 Apr 7;9:11. doi: 10.1038/s41540-023-00274-9

Table 2.

Model evaluation.

Metabolite	Model	Balanced accuracy
L-Glutamate	M1	0.265
L-Glutamate	M1_Smart	0.478
L-Proline	M1	0.300
L-Proline	M1_Smart	0.325
L-Histidine	M1	0.289
L-Histidine	M1_Smart	0.316
L-Tryptophan	M1	0.250
L-Tryptophan	M1_Smart	0.413
L-Leucine^a	M1	0.000
L-Leucine^a	M1_Smart	0.000
Adenine^b	M1	0.000
Adenine^b	M1_Smart	0.469
Guanine	M1	0.406
Guanine	M1_Smart	0.544
Uracil^a	M1	0.000
Uracil^a	M1_Smart	0.000
Adenosine	M1	0.644
Adenosine	M1_Smart	0.544
PEP	M1	0.507
PEP	M1_Smart	0.382

Balanced accuracy of the different models when predicting for significant changes in metabolite flux (p-value < 0.1, Wilcoxon signed-rank test) in comparison to observed significant changes in metabolite accumulation (p-value < 0.1, linear model) for all the comparisons available in the study.

^ano correct matches done by either of the models.

^bno significant observations for all pairwise comparisons.).