A training set is the dataset that is used to train a machine learning model, in this case, a DDI prediction model. To evaluate the prediction model accuracy, we apply a specific cross-validation technique called “Leave all pairs out” in which all pairs that include at least one compound that determines the predicted effect are excluded. The values that are yielded from such procedure are IAP (Invariant Accuracy of Prediction) values, which range from 0 (theoretically, the lowest accuracy of prediction) to 1 (theoretically, the highest accuracy of prediction). For more thorough explanation, please visit this page
To get a more detailed information on a specific training set, click the link listed on the left.
Training sets were created on the basis of the data from the literature.
The detailed information about training sets can be obtained in the appropriate section.