Fig. 5.

ReLeaSE (reinforcement learning for structural evolution). General pipeline of reinforce learning system for the generation of novel compounds. This technique uses database such as chEMBL, having chemical and biological information regarding available drug molecules, and uses these data to identify the initial model of a new compound which is likely to have a pharmacological affect. Later through a process of reinforcement learning, the initial model is converted into a de novo design, and later converted into the smiles of the bioactive molecule