Struct2Graph: a graph attention network for structure based predictions of protein–protein interactions

BMC Bioinformatics

Table 3 Fivefold cross-validation performance analysis of deep-learning based machine learning methods on unbalanced dataset (1:2)

Method	Performance (%) on unbalanced training set—1:2
Method	Accuracy	Precision	Recall	Specificity
DeepPPI	97.40 ± 0.44	98.64 ± 0.61	93.52 ± 1.64	99.35 ± 0.30
DeepFE-PPI	98.91 ± 0.09	99.00 ± 0.32	97.71 ± 0.33	99.51 ± 0.16
Struct2Graph	99.03 ± 0.24	99.13 ± 0.25	98.11 ± 0.58	99.53 ± 0.13

Method	MCC	F1-score	ROC-AUC	NPV
DeepPPI	94.16 ± 0.97	96.00 ± 0.72	99.19 ± 0.21	96.85 ± 0.76
DeepFE-PPI	97.54 ± 0.20	98.35 ± 0.13	99.56 ± 0.08	99.86 ± 0.16
Struct2Graph	97.87 ± 0.51	98.62 ± 0.32	99.47 ± 0.20	98.97 ± 0.34

ISSN: 1471-2105