Retrosynthetic Planning with Retro*

Retro: Learning Retrosynthetic Planning with Neural Guided A Search

@inproceedings{chen2020retro,
  title={Retro*: Learning Retrosynthetic Planning with Neural Guided A* Search},
  author={Chen, Binghong and Li, Chengtao and Dai, Hanjun and Song, Le},
  booktitle={The 37th International Conference on Machine Learning (ICML 2020)},
  year={2020}
}

1. Setup the environment

1) Download the repository

git clone git@github.com:binghong-ml/retro_star.git
cd retro_star

2) Create a conda environment

conda env create -f environment.yml
conda activate retro_star_env

2. Download the data

1) Download the building block molecules, pretrained models, and (optional) test data

Download and unzip the files from this link, and put all the folders (dataset/, one_step_model/ and saved_models/) under the retro_star directory.

3. Install Retro* lib

Install the retrosynthetic planning library with the following commands.

pip install -e retro_star/packages/mlp_retrosyn
pip install -e retro_star/packages/rdchiral
pip install -e .

4. Reproduce experiment results

To plan with Retro*, run the following command,

cd retro_star
python retro_plan.py --use_value_fn

Ignore the --use_value_fn option to plan without the learned value function.

You can also train your own value function via,

python train.py

5. Example usage

See example.py for an example usage.

from retro_star.api import RSPlanner

planner = RSPlanner(
    gpu=-1,
    use_value_fn=True,
    iterations=100,
    expansion_topk=50
)

result = planner.plan('CCCC[C@@H](C(=O)N1CCC[C@H]1C(=O)O)[C@@H](F)C(=O)OC')