Skip to main content
Fig. 5 | Journal of Cheminformatics

Fig. 5

From: DrugEx v2: de novo design of drug molecules by Pareto-based multi-objective reinforcement learning in polypharmacology

Fig. 5

The distribution of SA score and QED score of desired ligands and generated ligands. Shown are the distribution in the LIGAND set and of molecules generated by four different methods with PR (A, B, E and F) and WS (C, D, G and H) rewarding schemes in the multi-target case (A–D) and target-specific case (E–H). The molecules from the LIGAND set were shown in orange, and the molecules generated by DrugEx v1, v2, ORGANIC and REINVENT were shown in blue, green, red, and purple, respectively. Overall DrugEx v1 and v2 are better able to emulate the observed distributions in the training set compared to ORGANIC and REINVENT

Back to article page