Research Article

Identification of Microbial and Proteomic Biomarkers in Early Childhood Caries

Table 1

Performance statistics of three (multivariate) classification models built for the microbial data. The models were optimized for the average misclassification error (zero-one loss). The statistics include averages and standard deviations of test errors, sensitivities, and specificities of respective classifiers. The averages and standard deviations were calculated across 40 different train/test obtained using the random subsampling approach.

ClassifierTest errorSensitivitySpecificity

“SVM” 1 5 . 6 5 % ± 3 . 8 7 % 8 1 . 9 8 % ± 6 . 6 5 % 8 6 . 2 4 % ± 5 . 1 2 %
SVM 20 WLCX 1 1 . 7 7 % ± 2 3 . 6 7 4 % 8 6 . 0 5 % ± 7 . 3 6 % 9 0 . 1 1 % ± 4 . 5 9 %
“RF” 8 . 3 1 % ± 3 . 1 5 % 8 7 . 5 1 % ± 6 . 5 1 % 9 4 . 9 1 % ± 3 . 2 3 %

SVM: linear support vector machine.
SVM on the top 20 Wilcoxon peaks.
Random forest.