Skip to main content

Table 2 Comparison between SMILES based and graph-based generators in \(D_{KL}\)(\(\times\,10^{-3}\)) and \(D_{JS}\)(\(\times\,10^{-3}\))

From: Multi-objective de novo drug design with conditional graph generative model

Model

MW

LogP

QED

\(D_{KL}\)

\(D_{JS}\)

\(D_{KL}\)

\(D_{JS}\)

\(D_{KL}\)

\(D_{JS}\)

SMILES VAE

13.5 ± 0.6

3.6 ± 0.2

3.9 ± 0.4

0.9 ± 0.1

2.6 ± 0.4

0.6 ± 0.1

SMILES GRU1

8.6 ± 0.4

2.3 ± 0.1

3.1 ± 0.3

0.7 ± 0.0

1.5 ± 0.3

0.3 ± 0.1

SMILES GRU2

7.8 ± 0:3

2:0 ± 0.1

1.4 ± 0.2

0.3 ± 0.0

2.2 ± 0.3

0.5 ± 0.1

SMILES LSTM

6.5 ± 0.7

1.8 ± 0.2

3.4 ± 1.2

0.8 ± 0.3

1.9 ± 1.3

0.4 ± 0.3

MolMP (\(\alpha =1.0\))

11.5 ± 1.3

3.4 ± 0.4

7.0 ± 1.8

1.7 ± 0.4

5.3 ± 1.2

1.3 ± 0.3

MolMP (\(\alpha =0.8\))

8.3 ± 1.6

2.4 ± 0.5

4.3 ± 1.2

0.9 ± 0.2

2.7 ± 0.8

0.6 ± 0.2

MolMP (\(\alpha =0.6\))

8.4 ± 1.0

2.4 ± 0.3

5.0 ± 1.3

1.1 ± 0.4

3.0 ± 0.9

0.7 ± 0.2

MolRNN (\(\alpha =1.0\))

5.0 ± 0.6

1.4 ± 0.2

2.8 ± 0.5

0.7 ± 0.1

2.0 ± 0.6

0.5 ± 0.1

MolRNN (\(\alpha =0.8\))

4.1 ± 0.7

1.1 ± 0.2

1.6 ± 0.3

0.3 ± 0.1

1.0 ± 0.2

0.2 ± 0.0

MolRNN (\(\alpha =0.6\))

3.3 ± 0.2*

0.9 ± 0.1**

3.0 ± 0.4

0.5 ± 0.1

1.1 ± 0.4

0.2 ± 0.1

  1. Results are reported as \(Mean \pm SD\). The best performance in each metric is highlighted in italics face. Paired t-tests are carried out for the difference between the best and second performing methods (*** for \(p\le 0.001\), ** for \(p\le 0.01\) and * for \(p\le 0.05\)). Multiple models are highlighted if the difference is not significant