Table 4.

Performance of DeepMHADTA and the ablation models using structures or sequence alone on Davis dataset.

Models	CI	${\mathit{r}}_{\mathit{m}}^2$	AUPR	MSE
Without Structures	0.893	0.699	0.754	0.253
Without Sequence	0.853	0.588	0.698	0.358
DeepMHADTA	0.895	0.701	0.766	0.244