DeepMPF: deep learning framework for predicting drug–target interactions based on multi-modal representation with meta-path semantic analysis

Table 5 AUC values of comparing with state-of-the-art methods on gold-standard datasets

Model view	Method	Enzyme	GPCR	Ion channel	Nuclear receptor
Lower-view	Zhan et al. [69]	0.9532	0.8882	0.9349	0.8199
	Li et al. [70]	0.9288	0.8856	0.9171	0.9300
	Pan et al. [30]	0.9498	0.8775	0.9270	0.7755
	SAR [73]	0.9486	0.8902	0.9428	0.8822
	MLCLE [74]	0.8420	0.8500	0.7950	0.7900
	RFDT [75]	0.9150	0.8450	0.8900	0.7230
	DeepDTIs [31]	0.9067	0.8603	0.9417	0.8043
Higher-view	DASPfind [26]	0.9291	0.8810	0.9068	0.8527
	DT‑Hybrid [71]	0.8980	0.8387	0.9200	0.6995
	NRWRH [72]	0.9289	0.8493	0.9156	0.7390
	CMF [76]	0.8785	0.8244	0.8974	0.7637
	BRDTI [77]	0.8834	0.8487	0.9234	0.7962
	lNeuRank [78]	0.9539	0.8615	0.9686	0.7832
	DeepMPF (our)	0.9645 ± 0.0046	0.8782 ± 0.0236	0.9762 ± 0.0015	0.8272 ± 0.0894

ISSN: 1479-5876