Table 4 Performance of prediction models using various marker sets selected by several methods. The numbers in parentheses indicate the number of markers used in the prediction models.
| Â | Phylum | Genus | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Clinical information | Marker selection | Prediction model | AUCMD | AUCts | Marker selection | Prediction model | AUCMD | AUCts | |||||
Age, sex | Clinical information only | LR | 0.599 | 0.545 | Clinical information only | LR | 0.599 | 0.545 | |||||
RF | 0.859 | 0.664 | RF | 0.859 | 0.664 | ||||||||
SVM | 0.599 | 0.545 | SVM | 0.599 | 0.545 | ||||||||
DNN | 0.600 | 0.453 | DNN | 0.600 | 0.453 | ||||||||
Single (2) | LR | 0.970 | 0.964 | Single (7) | LR | 1.000 | 0.927 | ||||||
RF | 1.000 | 0.952 | RF | 1.000 | 0.938 | ||||||||
SVM | 0.984 | 0.957 | SVM | 1.000 | 0.913 | ||||||||
DNN | 0.910 | 0.954 | DNN | 1.000 | 0.988 | ||||||||
Stepwise1 (3) | LR | 0.990 | 0.934 | Stepwise1 (3) | LR | 0.996 | 0.955 | ||||||
RF | 1.000 | 0.891 | RF | 1.000 | 0.929 | ||||||||
SVM | 0.984 | 0.957 | SVM | 0.939 | 0.943 | ||||||||
DNN | 0.855 | 0.959 | DNN | 0.962 | 0.807 | ||||||||
Stepwise2 (3) | LR | 0.955 | 0.915 | Stepwise2 (3) | LR | 0.972 | 0.924 | ||||||
RF | 1.000 | 0.935 | RF | 1.000 | 0.921 | ||||||||
SVM | 0.931 | 0.947 | SVM | 0.968 | 0.929 | ||||||||
DNN | 0.858 | 0.937 | DNN | 0.910 | 0.868 | ||||||||
LASSO (10) | LR | 0.936 | 0.818 | LASSO (11) | LR | 1.000 | 0.896 | ||||||
RF | 1.000 | 0.922 | RF | 1.000 | 0.877 | ||||||||
SVM | 1.000 | 0.844 | SVM | 1.000 | 0.929 | ||||||||
DNN | 0.966 | 0.899 | DNN | 1.000 | 0.961 | ||||||||
Whole marker (20) | LR | 1.000 | 0.891 | Whole marker (296) | LR | 1.000 | 0.378 | ||||||
RF | 1.000 | 0.956 | RF | 1.000 | 0.957 | ||||||||
SVM | 1.000 | 0.889 | SVM | 1.000 | 0.980 | ||||||||
DNN | 1.000 | 0.848 | DNN | 1.000 | 0.998 | ||||||||
