Journal of Cheminformatics

Table 3 Performance of predicting drug‒drug interactions via the GIN of the group graph and other baseline models

From: Group graph: a molecular graph representation with enhanced performance, efficiency and interpretability

Tasks	BIOSNAP			DrugBankDDI
Metric	AUC_ROC↑	PRC_AUC↑	F1↑	AUC_ROC↑	AUC_PRC↑	F1↑
SimNN	0.8530 _±0.0010	0.8480 _±0.0010	0.7140 _±0.0010	0.7860 _±0.0030	0.7530 _±0.0030	0.7090 _±0.0040
DeepDDI	0.8860 _±0.0010	0.8710 _±0.0070	0.8170 _±0.0070	0.8440 _±0.0030	0.8280 _±0.0020	0.7720 _±0.0060
CASTER	0.9100 _±0.0050	0.8870 _±0.0012	0.8430 _±0.0050	0.8610 _±0.0050	0.8290 _±0.0030	0.7960 _±0.0070
PretrainGNN	0.9948 _±0.0002	0.9939 _±0.0001	0.9607 _±0.0022	0.9716 _±0.0003	0.9668 _±0.0004	0.9172 _±0.0007
GIN(FGS-pretraining)	0.9957 _±0.0005	0.9940 _±0.0008	0.9795 _±0.0010	0.9792 _±0.0002	0.9755 _±0.0006	0.9343 _±0.0010
GIN(group graph)	0.9962 _±0.0002	0.9950 _±0.0003	0.9793 _±0.0015	0.9865 _±0.0004	0.9842 _±0.0007	0.9498 _±0.0011

The average values and 95% confidence intervals from three independent runs are reported

Back to article page

ISSN: 1758-2946

Contact us

Submission enquiries: journalsubmissions@springernature.com