Fig. 4
From: Oral-microbiome-derived signatures enable non-invasive diagnosis of laryngeal cancers
Identification of microbial-based markers for clinical diagnosis by random forest models. A Mean decrease accuracy (MDA) of variables from the RF model using the profile of 15 selected biomarkers in tissue samples (left). ROC curves are based on the training set and test set from tissue samples (middle). The AUC is displayed in the lower right-hand corner. PCoA based on the profile of 15 selected genera in tissue samples (right). The P value is derived from PERMANOVA. B MDA of 15 selected biomarkers (left), ROC curves, and AUC (middle) based on the RF model from oral rinse samples. PCoA and PERMANOVA based on the profile of 15 selected genera in oral rinse samples (right)