Table 10. Performance changes of our model on the PK DDI in vivo dataset by removing features.

	Removed Features	P (%)	R (%)	F (%)
Ensemble	+ All Features	85.0	82.6	83.8
Single	+ All Features	82.1	83.3	82.6
Single	+ All Features—Subtree feature	78.8	84.2	81.3
	- Position feature	55.7	74.4	62.8
	- Static Word Embed	51.8	70.7	59.5

Note: P, R and F denotes Precision, Recall and F1 score, respectively. The same experiment was repeated five times and the results were averaged.