An end-to-end method for predicting compound-protein interactions based on simplified homogeneous graph convolutional network and pre-trained language model

Table 5 Results of SPVec-SGCN model performance compared with machine learning- and deep learning-based models on BindingDB and PubChem test sets

Testset	Methods	Accuracy	Precision	Recall	F1-Score	AUC	AUPR
BindingDB	Ours	0.9805	0.9763	0.9847	0.9805	0.9979	0.9979
	GNB	0.6354	0.6236	0.6789	0.6500	0.6822	0.6615
	RF	0.7747	0.8072	0.7205	0.7614	0.8689	0.8745
	GBDT	0.6868	0.6524	0.7964	0.7172	0.7751	0.7733
	DNN	0.751	0.7412	0.7694	0.755	0.8222	0.807
PubChem	Ours	0.9948	0.7754	0.8262	0.8000	0.9875	0.8709
	GNB	0.5954	0.1558	0.5757	0.3708	0.6070	0.1672
	RF	0.8246	0.0466	0.6460	0.0835	0.7648	0.1940
	GBDT	0.5538	0.0207	0.7359	0.0402	0.6949	0.0244
	DNN	0.6003	0.0202	0.6411	0.0391	0.6374	0.0179

ISSN: 1758-2946