Skip to main content

Table 1 Performance of Graph-CNN on metastatic event prediction, depending on normalization

From: Explaining decisions of graph convolutional neural networks: patient-specific molecular subnetworks responsible for metastasis prediction in breast cancer

Method

Std

100*AUC

Accuracy, %

F1-weighted, %

Graph-CNN

-

82.57 ±1.25

76.07 ±1.30

75.82 ±1.33

Random Forest

-

81.27 ±1.66

74.23 ±1.73

73.47 ±1.84

Graph-CNN

+

82.16 ±1.25

76.18 ±1.36

75.86 ±1.35

Random Forest

+

81.40 ±1.76

74.74 ±1.67

74.00 ±1.82

glmgraph

+

80.88 ±1.37

75.14 ±1.30

74.73 ±1.39

  1. Std stands for standardization of features (genes)
Back to article page

Genome Medicine

ISSN: 1756-994X

Contact us