Fig. 4

Area under the receiver-operating characteristic (AUROC) for classifying case versus control status from the URT microbiome profile across studies. AUROC values are shown for each study and sampling site (N = 30 data sets, 4706 samples), based on random-forest classifiers constructed using fivefold cross-validation for data from each study, separately. Values less than 0.5 are not shown. Sample count for each study is shown (range = 12–1021). Per-study disease type is shown via color encoding. Shaded background indicates the URT sampling site of each study (nasopharynx = pink; oropharynx = blue)