FIG 4.

A universal model for AD diagnosis via skin microbiome. (a) The 25 genera with the most discriminating power in the universal diagnosis model were selected as disease markers. The bar length at each row indicates relative contribution of the genus to the RF model. Inset, relationship between the number of variables (i.e., genera) in the RF model and model performance. (b) MiSH distinguishes healthy and lesional samples. (c) Correlation between core microbiome size and model performance along the spatial scaling of geographical distance. The y coordinate (line plot) depicts the size of core microbiome. The x coordinate (bar plot) describes model accuracy (i.e., AUC). The bottom table describes the spatial scales of sampling, with “+” and “−” indicating inclusion and exclusion of samples, respectively, at each of the scales.