“…Below we summarize a few models which follow this scheme: NeVAE, 17 a model from DeepMind, 10 MolMP and MolRNN, 11 JT-VAE, 12 HierVAE, 16 and GraphINVENT. 25 For additional examples of iterative molecular graph generation, see GCPN, 14 MolecularRNN 48 (extension of GraphRNN 15 to molecules), CGVAE, 13 RL-VAE, 49 GraphAF, 24 and ScaffoldVAE. 19 Additionally, while Graph Recurrent Attention Networks (GRANs) 50 have not yet, to our knowledge, been used to generate molecular graphs, they have been used with great success to generate protein graphs, where each node represents an individual amino acid.…”