MCL-DTI: using drug multimodal information and bi-directional cross-attention learning method for predicting drug–target interaction

Table 5 Ablation study on combining image modal and text modal

Dataset	λ₁	λ₂	ROC–AUC	PR–AUC	Recall
Human	\(\lambda _1\)	\(\lambda _2\)	0.990	0.992	0.961
	\(\lambda _1\)	1	0.984	0.983	0.942
	1	\(\lambda _2\)	0.982	0.984	0.945
	\(\lambda _1\)	\(1-\lambda _1\)	0.982	0.981	0.936
	1	1	0.987	0.989	0.955
	1	0	0.981	0.985	0.935
	0	1	0.968	0.968	0.920
Davis	\(\lambda _1\)	\(\lambda _2\)	0.922	0.492	0.895
	\(\lambda _1\)	1	0.917	0.475	0.877
	1	\(\lambda _2\)	0.918	0.467	0.861
	\(\lambda _1\)	\(1-\lambda _1\)	0.917	0.474	0.891
	1	1	0.916	0.486	0.874
	1	0	0.918	0.469	0.891
	0	1	0.916	0.487	0.856

ISSN: 1471-2105