CAT-DTI: cross-attention and Transformer network with domain adaptation for drug-target interaction prediction

BMC Bioinformatics

Table 3 Cross-domain performance comparison of CAT-DTI and other baselines on BindingDB and BioSNAP Datasets

Dataset	Method	AUROC	AUPRC	F1	Accuracy
BindingDB	SVM [10]	0.490 ± 0.015	0.460 ± 0.001	0.162 ± 0.158	0.531 ± 0.009
	RF [11]	0.493 ± 0.021	0.468 ± 0.023	0.109 ± 0.029	0.535 ± 0.012
	GraphDTA [23]	0.536 ± 0.015	0.496 ± 0.029	0.668 ± 0.001	0.472 ± 0.009
	TransformerCPI [29]	0.597 ± 0.041	0.562 ± 0.031	0.670 ± 0.005	0.490 ± 0.027
	MolTrans [30]	0.554 ± 0.024	0.511 ± 0.025	0.668 ± 0.001	0.470 ± 0.004
	DrugBAN [31]	0.576 ± 0.023	0.535 ± 0.014	0.668 ± 0.002	0.471 ± 0.012
	DrugBAN\(_{\text {CDAN}}\) [31]	0.604 ± 0.027	0.570 ± 0.047	0.675 ± 0.004	0.509 ± 0.021
	CAT-DTI	0.636 ± 0.013	0.573 ± 0.020	0.688 ± 0.004	0.553 ± 0.024
	CAT-DTI\(_{\text {CDAN}}\)	0.678 ± 0.005	0.626 ± 0.021	0.690 ± 0.004	0.572 ± 0.016
BioSNAP	SVM [10]	0.602 ± 0.005	0.528 ± 0.005	0.400 ± 0.122	0.513 ± 0.011
	RF [11]	0.590 ± 0.015	0.568 ± 0.018	0.018 ± 0.010	0.499 ± 0.004
	GraphDTA [23]	0.618 ± 0.005	0.618 ± 0.008	0.672 ± 0.003	0.535 ± 0.024
	TransformerCPI [29]	0.645 ± 0.022	0.642 ± 0.032	0.681 ± 0.009	0.558 ± 0.025
	MolTrans [30]	0.621 ± 0.015	0.608 ± 0.022	0.675 ± 0.006	0.546 ± 0.032
	DrugBAN [31]	0.630 ± 0.007	0.622 ± 0.018	0.671 ± 0.004	0.537 ± 0.034
	DrugBAN\(_{\text {CDAN}}\) [31]	0.685 ± 0.044	0.713 ± 0.041	0.677 ± 0.010	0.565 ± 0.056
	CAT-DTI	0.708 ± 0.008	0.718 ± 0.009	0.695 ± 0.008	0.618 ± 0.031
	CAT-DTI\(_{\text {CDAN}}\)	0.729 ± 0.010	0.733 ± 0.016	0.699 ± 0.008	0.633 ± 0.021

ISSN: 1471-2105