This repository is the implementation code for Materials discovery with extreme properties via reinforcement learning-guided combinatorial chemistry

If you find this code or idea useful, please consider citing our paper:

@article{kim2024materials,
  title={Materials Discovery with Extreme Properties via Reinforcement Learning-Guided Combinatorial Chemistry},
  author={Kim, Hyunseung and Choi, Haeyeon and Kang, Dongju and Lee, Won Bo and Na, Jonggeol},
  journal={Chemical Science},
  year={2024},
  publisher={Royal Society of Chemistry}
}

html files in result/extrapolation were visualized with ploty which is a tool for generating user interactive graph. You can see the results of materials extrapolation to hit multiple extreme target properties (Fig. 3 in the manuscript) with the user interactive graph.
The following gif example of parallel cooridates plot shows the results of generated molecules. The vertical magenta lines indicate constraints. Each horizontal line indicate a set of properties of a generated molecule and it is colored by score if it meets the all constraints simultaneously. pRMSE = 1 is a target bound for each property.

python inference_extrapolation.py data/checkpoints/extrapolation/checkpoint-750-1 --run PPO --env DrugDiscoveryEnv --episodes 1000
python inference_extrapolation.py data/checkpoints/extrapolation/checkpoint-750-2 --run PPO --env DrugDiscoveryEnv --episodes 1000
python inference_extrapolation.py data/checkpoints/extrapolation/checkpoint-750-3 --run PPO --env DrugDiscoveryEnv --episodes 1000
...
python inference_extrapolation.py data/checkpoints/extrapolation/checkpoint-750-10 --run PPO --env DrugDiscoveryEnv --episodes 1000

python inference_docking.py data/checkpoints/docking/checkpoint-80 --run PPO --env DrugDiscoveryEnv --episodes 1000

python inference_HIV.py data/checkpoints/HIV/checkpoint-250-ccr5 --run PPO --env DrugDiscoveryEnv --episodes 1000
python inference_HIV.py data/checkpoints/HIV/checkpoint-250-int --run PPO --env DrugDiscoveryEnv --episodes 1000
python inference_HIV.py data/checkpoints/HIV/checkpoint-250-rt --run PPO --env DrugDiscoveryEnv --episodes 1000