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

Table 7 Performance of graph based and SMILES based model on inhibitor generation, results are reported as \(Mean \pm SD\)

Condition (\({\mathbf{c}}\))	\(R_0\)	\(I_0\)	Model	% valid	\(R_{\mathbf{c}}\)	\(EOR_{\mathbf{c}}\)	Diversity
GSK-3\(\beta\)(+)	0.0008	0.806	Graph	0.939 ± 0.007	0.53 ± 0.01	666	0.783 ± 0.006
JNK3(+)	0.0008	0.806	SMILES	0.959 ± 0.003**	0.56 ± 0.01***	697	0.784 ± 0.003
GSK-3\(\beta\)(+)	0.01	0.860	Graph	0.932 ± 0.007	0.42 ± 0.01	42	0.851 ± 0.001
JNK3(−)	0.01	0.860	SMILES	0.928 ± 0.003*	0.47 ± 0.01***	47	0.854 ± 0.001**
GSK-3\(\beta\)(−)	0.0008	0.829	Graph	0.955 ± 0.003**	0.61 ± 0.00***	759	0.814 ± 0.002
JNK3(+)	0.0008	0.829	SMILES	0.944 ± 0.003	0.56 ± 0.01	698	0.821 ± 0.001***

The best performance in each metric is highlighted in italics face. Paired t-tests are carried out for the difference between the graph and SMILES based method (*** for \(p\le 0.001\), ** for \(p\le 0.01\) and * for \(p\le 0.05\))

ISSN: 1758-2946