Figure 5

Performance of classification models using DEPs and DEGs. (A) Accuracies of the DEP- and DEG-based models. The accuracy of all models was estimated by the leave-one-out test. (B) Sensitivities and specificities of the DEP- and DEG-based models. (C) Confusion matrix of DEP- and DEG-based models in (A) (UI, Uninfected; AT, Acute; NP, Non-progressive; CN, Chronic; DT, Decision tree; NN, Neural network). This matrix shows the actual stages of samples and their predicted stages by classification methods. Each column represents a predicted stage, and each row represents an actual stage. The count represents the result for DEGs/that for DEPs. (D) Accuracy of SVMs according to cutoff values for selecting DEPs. (E) Accuracies according to classification models and the number of principal components used for building models. DEPs showed higher accuracy than did DEGs regardless of the classification model or the number of used principal components.