Skip to main content

Table 5 The performance of different types of features on ChCh-Miner dataset

From: Multi-type feature fusion based on graph neural network for drug-drug interaction prediction

Method

AUROC

AUPRC

F1

S

90.17 ± 0.04

90.27 ± 0.18

89.14 ± 0.08

M

92.87 ± 0.74

92.55 ± 0.40

90.93 ± 0.56

I

93.23 ± 0.01

92.74 ± 0.15

90.28 ± 0.31

S+I

96.01 ± 0.83

96.89 ± 0.76

94.99 ± 0.23

S+M

95.49 ± 0.72

95.33 ± 0.54

95.02 ± 0.16

M+I

96.25 ± 0.05

97.23 ± 0.02

94.87 ± 0.05

S+M+I

97.02 ± 0.25

98.45 ± 0.06

96.94 ± 0.39

  1. The best results are highlighted in bold
  2. S SMILES sequence, M molecular graph, I interaction