FIG 3.

City-specific diagnosis models for AD. (a to d) The α diversity (a) and β diversity (b to d) of skin microbiota was significantly changed between healthy and lesional samples in each of the three cities. (e) Heatmap showing the relative abundances of 18 shared genera for building the city-specific AD diagnosis models. Among the three cities, the change patterns between healthy and diseased samples were consistent for 12 genera, which are highlighted with red font. (f) Performance of cross-city prediction using each city-specific AD diagnosis model, as assessed via the area under the ROC curve (AUROC). The ROC curve of 10-fold cross-validation was marked as blue lines and the ROC curve of the prediction as red lines.