Table 6.

GraphATT-DTA prediction performance on the Davis testing dataset vs. baseline models.

Models	Protein	Compound	Interaction	Davis MSE	Davis CI
DeepDTA	1D CNN	1D CNN	Concat	0.245	0.886
GraphDTA	1D CNN	GIN	Concat	0.229	0.890
DeepAffinity	RNN–CNN	RNN–CNN	Joint attention	0.302	0.870
ML-DTI	1D CNN	1D CNN	Mutual learning	0.222	0.891
HyperAttentionDTI	1D CNN	1D CNN	Hyperattention	0.233	0.876
FusionDTA	BI-LSTM	BI-LSTM	Fusion layer	0.203	0.911
GraphATT-DTA	1D CNN	MPNN	Interaction learning	0.204	0.904

Notes: BI-LSTM, bidirectional long short-term memory; CI, concordance index; CNN, convolutional neural network; GCN, graph convolutional neural network; GIN, graph isomorphism network; MSE, mean squared error; RNN, recurrent neural network.