Common important features between the GBM and ANN asthma models
Unlike the logistic models, each ML method used in this study, i.e., GBM and ANN, incorporated all microbial EV genera as features. Therefore, feature importance evaluation was conducted after developing the ML models to determine the genera with the greatest impact on both asthma diagnostic models. Twenty-one of the 50 highest ranked features in both the GBM and ANN asthma models were shared between both methods (Table 3 ). At the phylum level, Firmicutes accounted for the most significant features, with a total of 9 genera belonging to the Firmicutes phylum, followed by Proteobacteria with 7 significant features. At the genus level, Ruminococcaceae UCG-014 ,Lachnospiraceae UCG-008 , Pseudomonas ,Acinetobacter , Eubacterium hallii group , Blautia ,Bifidobacterium , Collinsella , Paracoccus , andHoldemanella were the 10 most important features in the GBM and ANN asthma models.