Table 2.

Evaluation results for multiple drug-drug interaction relation predictions (×100 for Hits@k).

Framework	Raw					Filter
	Hits@1a	Hits@5	Hits@10	MeanRankb	Hits@1		Hits@5	Hits@10	MeanRank
Tiresias	14.23	33.18	50.61	21.89	19.21		45.29	52.94	17.93
SCNNc	12.19	26.31	39.02	37.91	16.82		27.03	40.78	37.06
MDDPd	20.95	58.66	79.48	13.53	43.19		68.57	84.12	7.85
TransE	26.61	70.23	83.97	8.01	57.88		79.99	87.27	7.02
TransR	31.33	75.80	87.63	6.89	69.58		84.01	89.01	6.25
PRDe	45.11	85.57	91.01	6.11	75.11		88.60	92.85	5.45

^aHits@x: accuracy of real values contained in the top x rank.

^bMeanRank: the average rank of all correct relations.

^cSCNN: Syntax Convolutional Neural Network.

^dMDDP: multitask dyadic drug-drug interaction prediction.

^ePRD: Predicting Rich Drug-Drug Interaction.