Fig. 4From: Improved detection of disease-associated gut microbes using 16S sequence-based biomarkersROC curves and ROC-AUC scores for phenotype prediction using random forest classifiers and different aggregation strategy groupings as features. Both ROC and ROC-AUC confidence intervals were using a 80%/20% train-test split for 100 iterations. a ROC curves for prediction of the autism phenotype using various types of aggregation methods to extract features b ROC curves for prediction of the obesity phenotype using various types of aggregation methods to extract featuresBack to article page