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

Table 4 Model performance of nine model combinations between SGCN and GCN on two independent test sets

	Model	Accuracy	Precision	Recall	F1-Score	AUC	AUPR
BindingDB	GCN	0.8493	0.8370	0.8669	0.8516	0.9261	0.9276
	GCN + GCN	0.8243	0.7966	0.8697	0.8316	0.9084	0.9065
	GCN + SGCN	0.8563	0.8426	0.8754	0.8587	0.9296	0.9273
	GCN + GCN + GCN	0.8386	0.8152	0.8748	0.8440	0.9201	0.919
	GCN + SGCN + SGCN	0.8691	0.8476	0.8993	0.8727	0.9433	0.9425
	SGCN + SGCN + SGCN	0.9805	0.9763	0.9847	0.9805	0.9979	0.9979
	GCN + GCN + SGCN	0.8394	0.8129	0.8808	0.8455	0.9202	0.9162
	GCN + GCN + GCN + GCN	0.8197	0.8066	0.8399	0.8229	0.8906	0.8989
	GCN + SGCN + GCN + SGCN	0.8618	0.8369	0.8938	0.8658	0.9375	0.9358
PubChem	GCN	0.8151	0.0419	0.6208	0.0785	0.7741	0.0558
	GCN + GCN	0.8246	0.0406	0.5666	0.0758	0.7676	0.0681
	GCN + SGCN	0.8436	0.0510	0.6433	0.0945	0.8147	0.0771
	GCN + GCN + GCN	0.8211	0.0407	0.6643	0.0856	0.7934	0.0701
	GCN + SGCN + SGCN	0.8380	0.0508	0.6659	0.0944	0.8331	0.1156
	SGCN + SGCN + SGCN	0.9948	0.7754	0.8262	0.8000	0.9875	0.8709
	GCN + GCN + SGCN	0.8031	0.0431	0.6840	0.0810	0.8010	0.0998
	GCN + GCN + GCN + GCN	0.8032	0.0401	0.5632	0.0743	0.7849	0.0695
	GCN + SGCN + GCN + SGCN	0.8364	0.0537	0.6623	0.0934	0.8321	0.1121

ISSN: 1758-2946