Figure 5.

Top features (bacterial genera) in Random Forest (RF) models using (A) saliva samples of CRC (Colorectal cancer) patients and healthy controls (HC); (B) stool samples of CRC patients and healthy controls to predict sample classification into the patient and control groups. Figures also include Out-Of-Bag error and classification matrices for each model. In addition, Linear Discriminant Analysis (LDA) (C,D) between the same sample groups reveals bacterial genera which can serve as biomarkers for possible sample classification in CRC. Both approaches highlight, through different statistical methodologies, specific bacterial genera. ●.