Skip to main content

Table 3 Comparison of model performance in other cell lines. The p value compared with GCAN + LSTM is added in brackets

From: Novel deep learning-based transcriptome data analysis for drug-drug interaction prediction with an application in diabetes

Cell

Method

Macro-F1

Macro-recall

Macro-precision

A357

Original + DNN

85.3% ± 3% (0.001)

86.9% ± 3.5% (0.0003)

86.4% ± 2.8% (0.005)

 

GCAN + DNN

88.8% ± 2% (0.03)

89.9% ± 2% (0.035)

89.8% ± 2.1% (0.029)

 

Original + LSTM

89.2% ± 2.7% (0.005)

90.5% ± 3.6% (0.012)

89.5% ± (0.004)

 

GCAN + LSTM

92.8% ± 2.5% (–)

94.4% ± 2.7% (–)

92.4% ± 2.4% (–)

A549

Original + DNN

87.4% ± 1.2% (0.001)

88.2% ± 1.4% (0.001)

89% ± 1.3% (0.01)

 

GCAN + DNN

89.8% ± 1.6% (3.7E − 05)

90% ± 2.1% (0.0002)

91.5% ± 1.9% (0.112)

 

Original + LSTM

90.4% ± 1.1% (0.003)

91.9% ± 1.5% (0.011)

90.2% ± 0.8% (0.63)

 

GCAN + LSTM

92.7% ± 1.6% (–)

94.1% ± 2.4% (–)

92.3% ± 1.6% (–)

HA1E

Original + DNN

86.4% ± 2% (0.0007)

87.9% ± 1.9% (0.0004)

87.4% ± 2.2% (0.003)

 

GCAN + DNN

90.8% ± 1.4% (0.001)

91.3% ± 1.7% (0.002)

91.8% ± 1.4% (0.002)

 

Original + LSTM

91.6% ± 1.3% (0.012)

92.3% ± 1.7% (0.01)

91.9% ± 1.1% (0.021)

 

GCAN + LSTM

94.5% ± 0.8% (–)

95.9% ± 0.7% (–)

94.1% ± 0.9% (–)

MCF7

Original + DNN

88.9% ± 1.3% (0.001)

89.3% ± 1.4% (0.001)

90.5% ± 2.2% (0.021)

 

GCAN + DNN

92.9% ± 1.2% (0.005)

93.5% ± 1.4% (0.0004)

93.5% ± 1.2% (0.741)

 

Original + LSTM

93% ± 1.1% (0.01)

94.9% ± 1.8% (0.011)

92.3% ± 1.1% (0.104)

 

GCAN + LSTM

94.8% ± 1.6% (–)

96.7% ± 1.9% (–)

93.6% ± 1.6% (–)

  1. Bold indicates the best prediction performance