Fig. 2

Comparison of different metabolite panels based on receiver operating characteristic (ROC) curves. The curves correspond to metabolites having the highest area under the curve (AUC) and the highest accuracy values in predicting ACO from asthma and COPD. a ACO versus asthma: serine, AUC = 0.753 ± 0.06 (95% confidence interval (CI) 0.634–0.849); ethanolamine, AUC = 0.767 ± 0.06 (95% CI 0.650–0.861); glucose, AUC = 0.767 ± 0.06 (95% CI 0.650–0.861); 2-palmitoylglycerol, AUC = 0.780 ± 0.06 (95% CI 0.664–0.871); D-mannose, AUC = 0.913 ± 0.03 (95% CI 0.820–0.967); succinic acid, AUC = 0.770 ± 0.06 (95% CI 0.653–0.863), b ACO versus COPD: glucose, AUC = 0.722 ± 0.07 (95% CI 0.597–0.826); 2-palmitoylglycerol, AUC = 0.784 ± 0.06 (95% CI 0.664–0.876); lactic acid, AUC = 0.750 ± 0.06 (95% CI 0.627–0.849); linoleic acid, AUC = 0.798 ± 0.05 (95% CI 0.680–0.888); D-mannose, AUC = 0.791 ± 0.06 (95% CI 0.673–0.882); succinic acid, AUC = 0.710 ± 0.07 (95% CI 0.584–0.816)