Fig. 5.

Disease status classification using disease-associated taxa and/or metabolites. (a, c, e) Random forest classifiers were constructed to discriminate PCOS and healthy, PCOS-LB and Healthy-LB, PCOS-HB and Healthy-HB, respectively, in the training dataset. (b, d, f) Random forest classifiers composed of bacterial and fungal genera, metabolites, predicted pathways and their combinations were constructed to discriminate PCOS and healthy, PCOS-LB and Healthy-LB, PCOS-HB and Healthy-HB in the test dataset. ROC, receiver operating characteristic curve. AUC, area under the curve. The input features were excavated on the basis of Wilcox test comparison and the mean decrease in Gini by random forest importance parameters. Data were assigned to training (80%) and test (20%) datasets after the whole dataset was shuffled