Skip to main content
Fig. 6 | Journal of Cheminformatics

Fig. 6

From: Augmented Hill-Climb increases reinforcement learning efficiency for language-based de novo molecule generation

Fig. 6

Centroid of the 5 largest clusters for the top 100 molecules according to docking score against DRD2, OPRM1, AGTR1 and OX1R receptors. Cluster size (CS), centroid docking score (DS) and the average cluster docking score (AvDS) is annotated below. In each case Augmented Hill-Climb generates clusters with lower (better) docking scores, while maintaining reasonable chemotypes that are indistinguishable to those generated by REINVENT. Note that protonation states, tautomers and stereoisomers are enumerated by the docking protocol (see Methods)

Back to article page