Skip to main content

Table 5 Performance comparison with sufficient random seed values

From: MultiChem: predicting chemical properties using multi-view graph attention network

 

Dataset

Classification (AUROC)

Regression (RMSE)

BBBP

Tox21

ToxCast

SIDER

ClinTox

BACE

FreeSolv

ESOL

Lipo

GROVER

0.900

0.787

0.664

0.599

0.854

0.815

2.376

0.983

0.805

(Rong, et al., 2020)[50]

(0.034)

(0.041)

(0.029)

(0.036)

(0.142)

(0.046)

(0.891)

(0.175)

(0.072)

MPG

0.919

0.817

0.713

0.625

0.908

0.872

2.165

0.878

0.637

[35]

(0.025)

(0.021)

(0.016)

(0.024)

(0.031)

(0.044)

(0.594)

(0.128)

(0.043)

CD-MVGNN

0.907

0.837

0.729

0.623

0.885

0.870

2.049

0.825

0.600

(Ma, et al., 2022) [40]

(0.032)

(0.020)

(0.015)

(0.030)

(0.047)

(0.048)

(0.613)

(0.097)

(0.055)

KANO

0.951

0.820

0.719

0.614

0.910

0.874

1.970

0.798

0.618

(Fang, et al., 2023) [16]

(0.016)

(0.023)

(0.012)

(0.026)

(0.040)

(0.043)

(0.583)

(0.091)

(0.046)

Our

0.957

0.836

0.740

0.628

0.919

0.854

2.019

0.783

0.597

 

(0.017)

(0.021)

(0.014)

(0.027)

(0.037)

(0.045)

(0.532)

(0.074)

(0.038)

  1. Average AUROC and RMSE scores used 30 random seed values
  2. Bold means pretraining model
  3. The bold number and underlined number mean the highest and second-highest numbers