Table 2.

Predicted performances of the selected biosignatures

	Metric	In Sample	Out Sample	95 % Confidence Interval
Metabolites	R2	0.997	0.840	[ 0.335, 0.972 ]
	Mean Absolute Error	0.228	2.503	[ 0.783, 4.060 ]
	Mean Squared Error	0.2104	12.180	[ 0.964, 24.916 ]
Transcription factors	R2	0.983	0.825	[ 0.565, 0.939 ]
	Mean Absolute Error	0.897	2.658	[ 1.009, 4.950 ]
	Mean Squared Error	1.239	13.315	[ 3.743, 29.804 ]
Integrated list	R2	0.9965	0.909	[ 0.721, 0.986 ]
	Mean Absolute Error	0.465	1.779	[ 0.487, 3.919 ]
	Mean Squared Error	0.2639	6.927	[ 0.469, 18.832 ]

Performances are reported in terms of the determination coefficient R², Mean Absolute Error (MAE) and Mean Squared Error (MSE, see text for more details on these metrics). The in-sample performances quantify the fitness of the predictive models on the training data, while the out-of-sample values estimate the expected performance on new data. Confidence intervals are calculated using a bootstrap approach