Table 3.

Predictive performance of the top-ranked single-metabolite models.

Time Point	Metabolite	AUC [95% CI]	Accuracy [95% CI]	F1 Score [95% CI]	Association with High Stress
Rest	AICAR	0.688 [0.542 – 0.825]	0.693 [0.531 – 0.829]	0.576 [0.318 – 0.800]	↑
	Ile	0.688 [0.513 – 0.870]	0.667 [0.359 – 0.875]	0.708 [0.353 – 0.900]	↓
	Agmatine	0.680 [0.491 – 0.833]	0.636 [0.375 – 0.829]	0.595 [0.105 – 0.846]	↑
Post-UK	Tetrahydrocortisone	0.747 [0.525 – 0.901]	0.706 [0.359 – 0.875]	0.710 [0.146 – 0.895]	↓
	25 – OHD	0.746 [0.279 – 0.953]	0.705 [0.375 – 0.906]	0.730 [0.448 – 0.914]	↑
	Androstenediol	0.734 [0.543 – 0.889]	0.723 [0.498 – 0.860]	0.702 [0.425 – 0.850]	↓
Recover	Estrone	0.794 [0.590 – 0.933]	0.762 [0.563 – 0.891]	0.757 [0.463 – 0.914]	↓
	Hypoxanthine	0.729 [0.578 – 0.849]	0.683 [0.421 – 0.906]	0.700 [0.444 – 0.916]	↓
	AICAR	0.723 [0.565 – 0.861]	0.694 [0.515 – 0.829]	0.629 [0.259 – 0.811]	↑
Post-WM	N6,N6,N6-trimethyllysine	0.778 [0.638 – 0.921]	0.743 [0.469 – 0.891]	0.702 [0.462 – 0.874]	↓
	Choline	0.741 [0.559 – 0.881]	0.737 [0.515 – 0.875]	0.765 [0.571 – 0.911]	↓
	DHEA	0.740 [0.540 – 0.868]	0.692 [0.469 – 0.844]	0.698 [0.448 – 0.875]	↓
All	Choline	0.691 [0.537 – 0.809]	0.645 [0.457 – 0.766]	0.671 [0.446 – 0.799]	↓
	Ile	0.681 [0.499 – 0.855]	0.615 [0.370 – 0.778]	0.672 [0.378 – 0.834]	↓
	Estrone	0.666 [0.537 – 0.780]	0.607 [0.387 – 0.754]	0.595 [0.380 – 0.756]	↓

The table shows the performance of predictive models constructed using each metabolite as a single feature. The top 3 models at each time point, ranked by the highest area under the receiver operating characteristic curve (AUC), are presented.

Association with High Stress: ↑ indicates that higher values are associated with the high-stress group; ↓ indicates that lower values are associated with the high-stress group.

UK, Uchida-Kraepelin test; WM, working memory test; CI, confidence interval.

25-OHD, 25-hydroxyvitamin D; AICAR, 5-aminoimidazole-4-carboxamide ribonucleotide; DHEA, dehydroepiandrosterone; Ile, isoleucine.