BMC Bioinformatics

Table 3 Performance values for different CV schemes using DrugBank v4 and KEGG as reference DDIs

From: Evaluation of knowledge graph embedding approaches for drug-drug interaction prediction in realistic settings

		Traditional CV				Drug-wise CV				Pairwise CV
Reference Data	ML Method	AUPR	F1	AUC		AUPR	F1	AUC		AUPR	F1	AUC
DrugBank 4	Logistic Regression	0.75	0.70	0.76		0.70	0.65	0.72		0.66	0.60	0.67
	Naive Bayes	0.71	0.66	0.72		0.70	0.65	0.70		0.68	0.63	0.69
	Random Forest	0.90	0.84	0.91		0.78	0.69	0.79		0.69	0.52	0.70
KEGG	Logistic Regression	0.77	0.71	0.78		0.69	0.63	0.70		0.62	0.54	0.62
	Naive Bayes	0.73	0.67	0.74		0.71	0.63	0.71		0.66	0.58	0.67
	Random Forest	0.85	0.79	0.86		0.78	0.68	0.79		0.67	0.38	0.67

The embedding generated by RDF2Vec with SkipGram was used in the experiments. (Bold: best score)

Back to article page

ISSN: 1471-2105

Contact us

General enquiries: journalsubmissions@springernature.com