Fig. 3

Classification of myasthenia gravis (MG) status by relative abundance of candidate microbial marker species. A Distribution from five trials of 10-fold cross-validation error in a random forest classification of MG, with increasing species number used in each model. The model was trained using the relative abundance values for MG-associated species. The gray lines represent the cross-validation error for the five trails. The black curve represents the average of the gray lines. The red line marks the number of species in the optimal marker species set. B The probability of MG in the discovery cohort. C Receiver operating characteristic (ROC) in the discovery cohort. D Classification of the validation samples and their predicted probability of MG. E The probability of MG in the validation cohort. F ROC for the validation cohort samples