FIG 5.

Classifier used to distinguish ECC children from healthy controls. (a) The mRMR method was used to identify the ECC-associated markers. Sequential subsets were generated at five-species intervals. For each subset, the error rate was estimated using a leave-one-out cross-validation of a linear discrimination classifier. Using only the seven marker species as predictors, the SVM model exhibited predictive performance that was already comparable to the performance of the model derived from the optimum (highest value of the Matthews correlation coefficient) subset. (b) The relative abundances of seven marker species among the H2H group, the H2C group, and the caries group are shown in a box plot. (c) Receiver operating characteristic (ROC) curves for the ECC group and healthy controls; 95% confidence intervals (CIs) are indicated by error bars. (d) The probability of caries determined by the classifier among the H2H group, H2C group, and caries group is shown in a box plot. (e) ROC curves for the H2H group and H2C group.