Figure 5.

ROC curve analyses were performed to evaluate the diagnostic performance of oral microbiome for OSCC metastasis. (A, C) Boruta algorithm was used for feature selection and to assess the importance of variables. (B, D) ROC curves were constructed to predict diagnostic power of microbiome for tumor metastasis. It suggested that higher abundance of Lautropia, Asteroleplasma, Parvimonas, Peptostreptococcus, Pyramidobacter, Roseburia and Propionibacterium in the tumor tissue appeared to be associated with higher rates of tumor metastases.