Table 1.

Comparison Results (Mean Scores in %).

Model	Pang et al.			BioSNAP			AdverseDDI
	F1	ACC	AUROC	F1	ACC	AUROC	F1	ACC	AUROC
RFC	79.94	81.19	89.60	74.80	74.34	82.57	77.17	75.42	83.60
DeepDDI	81.43	90.44	92.18	88.60	89.23	94.74	87.26	87.45	94.34
DeepDDS	89.05	88.65	94.86	87.55	87.93	95.21	82.45	85.34	93.07
Deepconv	97.19	97.30	98.61	97.70	97.70	99.14	88.65	88.74	95.21
LSTM	96.35	96.14	98.49	97.75	97.72	98.78	88.41	88.23	94.30
GAT	95.87	95.39	97.30	91.19	91.52	96.82	86.14	85.92	93.62
AMDE	97.33	96.68	98.90	97.37	97.38	99.19	86.57	86.33	93.54
HTCL-DDI	96.33	96.34	95.99	-	-	-	85.07	85.27	92.14
DSN-DDI	96.90	96.80	98.28	93.72	93.55	97.89	83.76	84.13	92.62
MFE-DDI	97.91	97.87	99.18	98.20	98.23	99.22	89.52	89.38	95.62

Due to equipment memory limitation, we were unable to run HTCL-DDI on BioSNAP dataset.