DGDTA: dynamic graph attention network for predicting drug–target binding affinity

BMC Bioinformatics

Table 4 Comparison with the state-of-the-art methods

Dataset	Methods	CI	MSE	\({\textbf{r}}_{\textbf{m}}^{2}\)
Davis	DeepDTA [8]	0.878	0.261	0.631
	DeepCDA [35]	0.891	0.248	0.652
	MATT-DTI [34]	0.890	0.229	0.688
	GraphDTA(GAT) [26]	0.892	0.232	0.689
	GraphDTA(GAT-GCN) [26]	0.881	0.245	0.667
	CPInformer [6]	0.874	0.277	0.621
	DeepGLSTM [33]	0.893	0.236	0.679
	DGDTA-CL (ours)	0.889	0.237	0.672
	DGDTA-AL (ours)	0.899^*	0.225	0.707
KIBA	DeepDTA [8]	0.863	0.194	0.673
	DeepCDA [35]	0.889	0.176	0.682
	MATT-DTI [34]	0.889	0.150	0.762
	GraphDTA(GAT) [26]	0.866	0.179	0.738
	GraphDTA(GAT-GCN) [26]	0.891	0.139	0.789
	CPInformer [6]	0.867	0.183	0.678
	DeepGLSTM [33]	0.890	0.143	0.789
	DGDTA-AL (ours)	0.881	0.162	0.762
	DGDTA-CL (ours)	0.902	0.125	0.809

ISSN: 1471-2105