Skip to main content
Fig. 12 | Journal of Cheminformatics

Fig. 12

From: Molecular optimization by capturing chemist’s intuition using deep neural networks

Fig. 12

The overlap of the generated molecules with desirable properties from HierG2G, Seq2Seq and Transformer given each starting molecule in Test-Original. Each circle represents the set of molecules with desirable properties for each model. Transformer generated more desirable molecules than HierG2G and Seq2Seq, with the biggest circle. Most generated molecules with desirable properties from three models given the same starting molecule are different from each other, with only 2,937 identical molecules. Transformer and Seq2Seq have more identical desirable molecules (12,151) compared with (HierG2G, Seq2Seq) and (Transformer, HierG2G)

Back to article page