Table 3. Binary classification for the 7 metabolite model.
These values were derived from multivariate exploratory ROC analysis with PLS-DA or RF as the final classification methods. The 7 metabolite model is composed of tyrosine, tryptophan, threonic acid, pseudouridine, taurine, xylitol, and lyxitol. Sens, sensitivity; Spec, specificity. When both training and testing time points are listed, the model is trained using data from training time point and used to predict the testing time point classifier (i.e treated vs control).
| Classification Method | Training Time Point | Testing Time Point | True pos. | False neg. | True neg. | False pos. | Sens. | Spec. |
|---|---|---|---|---|---|---|---|---|
| PLS-DA | 1 | NA | 3 | 2 | 3 | 3 | 60% | 50% |
| PLS-DA | 2 | NA | 5 | 0 | 6 | 0 | 100% | 100% |
| PLS-DA | 3 | NA | 5 | 0 | 6 | 0 | 100% | 100% |
| PLS-DA | 4 | NA | 5 | 0 | 6 | 0 | 100% | 100% |
| PLS-DA | 5 | NA | 5 | 0 | 5 | 1 | 100% | 83.33% |
| RF | 1 | NA | 2 | 3 | 2 | 4 | 40% | 66.67% |
| RF | 2 | NA | 4 | 1 | 6 | 0 | 80% | 100% |
| RF | 3 | NA | 5 | 0 | 5 | 1 | 100% | 83.33% |
| RF | 4 | NA | 5 | 0 | 6 | 0 | 100% | 100% |
| RF | 5 | NA | 5 | 0 | 5 | 1 | 100% | 83.33% |
| PLS-DA | 2 | 2 | 4 | 0 | 4 | 0 | 100% | 100% |
| PLS-DA | 4 | 3 | 0 | 4 | 3 | 1 | 0% | 75% |
| PLS-DA | 5 | 4 | 4 | 0 | 2 | 2 | 100% | 50% |
| RF | 2 | 2 | 3 | 1 | 3 | 1 | 75% | 75% |
| RF | 4 | 3 | 0 | 4 | 2 | 2 | 0% | 50% |
| RF | 5 | 4 | 3 | 1 | 3 | 1 | 75% | 75% |