Fig. 5
From: AI-driven molecular generation of not-patented pharmaceutical compounds using world open patent data
Frequency of structural similarities of generated molecules against the drug-patent DB compounds. a Maximum similarities of generated molecules using the Rpatent (orange) and Rnot-patent (blue) rewards to the 247,738 drug-patent DB compounds used as training data of the patent RNN. b Maximum similarities of generated molecules using the Rpatent (orange) and Rnot-patent (blue) rewards to the drug-patent DB compounds. Molecular generation was performed at C = 0.1