Reinforcement Learning

• Acegen-Open: TorchRL-based toolkit for reinforcement learning in generative chemistry
  
• ChemTSv2: An extended version of ChemTS, focusing on functional molecular design using de novo molecule generators, incorporating improvements for LogP maximization tasks and other molecular design objectives.
  
• GENTRL (Generative Tensorial Reinforcement Learning): A variational autoencoder with a rich prior distribution of the latent space, trained to find molecules with high reward, emphasizing the relations between molecular structures and their properties.
  
• MolDQN: Optimization of Molecules via Deep Reinforcement Learning
  
• REINVENT 4: A molecular design tool for various design tasks like de novo design, scaffold hopping, and molecule optimization, using a reinforcement learning algorithm.
  
• ReLeaSE: Utilizes deep reinforcement learning for de novo drug design.
  
• RL-GraphInvent: An extension using reinforcement learning for targeted molecular generation.
  

Human-in-the-loop

• REINVENT-HITL: Focuses on human-in-the-loop assisted de novo molecular design, leveraging reinforcement learning for optimizing molecules based on human feedback.