Table 4:
Top performing models and their CV-AUCs based on models with one of ID50, ID80 and IIP as predictors, in addition to epitope and challenge virus, based on the subset of 218 animals with IC80 information. See also Tables S4 and S5.
| Neutralization Activity Measure* | Algorithm** | CV-AUC | 95% CI | |
|---|---|---|---|---|
| None | SL.glm | 0.719 | 0.638–0.800 | |
| ID50 | SuperLearner | 0.875 | 0.822–0.928 | |
| ID80 | SuperLearner | 0.879 | 0.827–0.931 | |
| IIP | SL. randomF orest | 0.833 | 0.772–0.894 |
*
In addition to epitope and challenge virus
**
Algorithms are listed by the functions used in the SuperLearner R package. An exception is “SL.naivebayes”, which was a custom wrapper designed to use the naiveBayes function from the e1071 package. The SL.glmnet package was used with the lasso penalty. All tuning parameters are set to the default values of the SuperLearner package, except SL.xgboost, which we modified to fit decision stumps rather than trees.