Table 6 Overall performance comparison: transductive vs. inductive settings.

From: PS3N: leveraging protein sequence-structure similarity for novel drug-drug interaction discovery

Model

Transductive

CS1 (New–New)

CS2 (Known–New)

ACC (%)

AUROC (%)

ACC (%)

AUROC (%)

ACC (%)

AUROC (%)

DeepDDI85

93.15

99.76

47.96

88.39

68.69

95.75

MR-GNN86

94.06

99.76

49.93

91.70

72.38

97.52

MHCADDI87

78.50

86.33

66.53

72.13

72.34

79.22

SSI-DDI21

94.47

98.38

65.02

72.38

73.35

80.92

PS3N (Ours)

97.30

99.50

70.05

62.12

69.03

67.63

  1. Bold values indicate the best performance for each metric.
Back to article page