Figure 5. Detection and testing of salivary microbial markers. The four differential bacteria differential bacteria were applied for the succeeding random forest classification, and the ROC curve was used to evaluate the accuracy of sample classifications. The results showed that the accuracy rate of application of the differential bacteria as a biomarker to distinguish naïve diabetes patients from healthy people (A) and patients treated with metformin (B) or combined medication (C) could reach to 83.3%, 75%, and 75%, respectively.