Fig. 3

OTU feature selection results. A The classification performance of the most important bacterial microbiota (OTUs) selected by each of the feature selection approaches used in the study: AM, CS, IG, MRMR, and SU are based on ensemble feature selection; RF and SVM are wrapper-based feature selection techniques; Union5, Union10, and Union20 combine the feature selection results. Under four classifiers, the selected OTUs can distinguish asthmatic subjects with nearly 99% AUC accuracy. B A subset of the most important 57 bacteriaal microbiota (OTUs) for asthmatic diagnosis developed by the feature selection methdology (Union20). The heat map shows the relative abundance of the 57 most important bacterial microbiota (OTUs) for asthmatic diagnosis