Table 1.
DDI relation types and explanations.
Fig 1.
We implemented both the one-stage and the two-stage method. (a) Data generation part. (b) One-stage method. Five-class type classifier for the one-stage method. (c) Two-stage method. The DDI detection classifier distinguishes positive DDI instances from negative instances. The DDI type classifier receives the predicted positive instances from the detection classifier as a testing set.
Fig 2.
The architecture of our recursive neural network model.
Our model is a variation of the binary tree-LSTM model. (1) The words in a sentence. The names of drug targets are underlined. (2) Vector representation of a word through the word embedding lookup process. (3) Subtree containment feature represents the importance of a node. (4) Position feature vector representing the relative distance of two target drugs from the current word position. (5) An example of the position feature vector. The current word is “accelerated.” (6) The size of the concatenated vector input x0 of our model is 10 (size of the subtree containment feature; (3) in the figure) + 20 (size of the position feature; (4) in the figure) + 200 (size of the word embedding; (2) in the figure).
Table 2.
Vector representation according to the distance between one of the target drugs and a current word.
Table 3.
Search process to find the best hyperparameters used for our model.
Table 4.
The statistics of the DDIExtraction Challenge’13 corpus after preprocessing.
Table 5.
Comparison between our proposed model and existing models.
Table 6.
Changes in our model’s performance in DDI detection by removing several features of our model.
Table 7.
The statistics from the PK DDI corpus after preprocessing.
Table 8.
Comparison of in vivo PK DDI results of our model and those of existing models.
Table 9.
Comparison of in vitro PK DDI results of our model and those existing models.
Table 10.
Performance changes of our model on the PK DDI in vivo dataset by removing features.
Table 11.
Examples of three common types of error cases.