Table 2 Transductive model evaluation based on training dataset with negative samples generated by random sampling. The three values separated by double vertical bars are AUROC, Spearman coefficient correlation (ρ) between the degree of positive samples and their predicted interaction probability, and that between the degree of negative samples and their predicted interaction probability. Results represent mean of n = 15 independent runs

From: Negative sampling strategies impact the prediction of scale-free biomolecular network interactions with machine learning

		Noise-RF	Seq-RF	Seq-Deep
LPI	NPInter4.0	0.993\|0.912\|0.170	0.993\|0.826\|0.144	0.994\|0.576\|0.333
LPI	RAIDv2.0	0.997\|0.859\|0.416	0.995\|0.772\|0.247	0.995\|0.861\|0.095
PPI	InBioMap	0.930\|0.867\|0.825	0.936\|0.845\|0.758	0.971\|0.621\|− 0.003
	STRING	0.844\|0.935\|0.865	0.862\|0.911\|0.734	0.935\|0.761\|0.093
	BioGRID	0.815\|0.738\|0.538	0.857\|0.819\|0.662	0.874\|0.739\|0.411
	HuRI	0.865\|0.716\|0.579	0.892\|0.747\|0.649	0.904\|0.618\|0.339
DTI	DrugBank	0.808\|0.742\|0.599	0.888\|0.765\|0.497	0.894\|0.674\|0.346
DTI	DrugCentral	0.864\|0.845\|0.667	0.920\|0.865\|0.454	0.934\|0.738\|0.223

Back to article page

ISSN: 1741-7007

Contact us

General enquiries: journalsubmissions@springernature.com