Fig. 4From: Comparison of structure- and ligand-based scoring functions for deep generative models: a GPCR case studyChemotype evolution during training, comparing the SVM-Agent and the Glide-Agent. Molecules were sampled during training at the start and after 0, 1000, 2000 and 3000 steps (at the end of training). Molecules in each batch were clustered and the centroids of the three largest clusters are shown here, alongside respective cluster size (CS). This visualizes the difference in topology and chemotype between the two approachesBack to article page