Skip to main content

Table 2 Conditional molecular graph generation with proposed model

From: Efficient learning of non-autoregressive graph variational autoencoders for molecular graph generation

DatasetTarget conditionG-meanUnique countMolWtLogP
QM9Training set100,000122.97 ± 7.610.14 ± 1.16
Unconditional generation0.6393243123.01 ± 8.04− 0.06 ± 1.36
MolWt = 1200.5832316121.85 ± 5.110.02 ± 1.36
MolWt = 1250.5431947125.11 ± 4.56− 0.27 ± 1.22
MolWt = 1300.4821475128.98 ± 4.27− 0.41 ± 1.33
LogP = − 0.40.5762399122.97 ± 8.26− 0.40 ± 0.73
LogP = 0.20.5432099122.53 ± 8.170.19 ± 0.75
LogP = 0.80.5371989122.17 ± 8.090.83 ± 0.72
ZINCTraining set100,000357.94 ± 65.482.62 ± 1.36
Unconditional generation0.8887000366.44 ± 51.632.49 ± 1.43
MolWt = 3000.7424090313.12 ± 13.721.91 ± 1.50
MolWt = 3500.7965045356.22 ± 12.662.24 ± 1.36
MolWt = 4000.8055212400.95 ± 13.662.78 ± 1.30
LogP = 1.50.8656470352.33 ± 46.781.66 ± 0.94
LogP = 2.50.8606356366.64 ± 48.302.46 ± 0.92
LogP = 3.50.8275658381.96 ± 48.463.22 ± 0.85