ED2Mol

<div align=center> <img src="./figs/Fig1.png" width="50%" height="50%" alt="TOC" align=center /> </div>

Section 1: Setup Environment

You can follow the instructions to setup the environment

We use mamba here. You can install it using conda install mamba. If you encounter any errors with mamba, please use conda instead.

1. you can install manually (cuda 11.7)

mamba create -n ed2mol
mamba activate ed2mol
mamba install python=3.8.19 pytorch=1.13.1 pytorch-cuda=11.7 rdkit=2023.03.2 plip biopython cctbx-base scikit-learn -y -c pytorch -c nvidia -c conda-forge
pip install torch_geometric==2.3.0

2. or you can install via conda yaml file

mamba env create -f ed2mol_env.yml -n ed2mol
mamba activate ed2mol

Section 2: Weights and Datasets

The model weights can be downloaded at the release page.

wget https://github.com/pineappleK/ED2Mol/releases/download/v1.1/weights.zip
unzip weights.zip

The main data used for training is from PDB or ZINC database.

For tranining ED extraction model, you can download the source data and unzip it.

For training fragment extension model, you can download the processed data and unzip it.

Section 3: Generation

Run ED2Mol on the test examples

1. Please setup the env dependencies
2. Just change to the base directory and run the Generate.py with prepared yml file

for denovo generation

CUDA_VISIBLE_DEVICES="0" python Generate.py --config ./configs/denovo.yml

for hit optimization

CUDA_VISIBLE_DEVICES="0" python Generate.py --config ./configs/hitopt.yml

Run ED2Mol on your own targets

For the denovo generation task, prepare your receptor PDB file and modify the example denovo.yml file.

• setting the output_dir and receptor parameters to your designated output path and receptor file.
• specifying the x, y, and z parameters as the center coordinates of the pocket of interest.

For the lead optimization task, you need also prepare a reference ligand core SDF file (Hint: You can use PyMOL to edit, delete, or export the file as an SDF) and modify the example leadopt.yml file.

• updating the reference_core parameter accordingly.

Then, you can run ED2Mol with your yml file.

CUDA_VISIBLE_DEVICES="0" python Generate.py --config path-to-your.yml

Section 4: Citation

If you find this work interesting, please cite

Li, M., Song, K., He, J. et al. Electron-density-informed effective and reliable de novo molecular design and optimization with ED2Mol. Nat Mach Intell 7, 1355–1368 (2025). https://doi.org/10.1038/s42256-025-01095-7

@article{li2025electron,
  title={Electron-density-informed effective and reliable de novo molecular design and optimization with ED2Mol},
  author={Li, Mingyu and Song, Kun and He, Jixiao and Zhao, Mingzhu and You, Gengshu and Zhong, Jie and Zhao, Mengxi and Li, Arong and Chen, Yu and Li, Guobin and others},
  journal={Nature Machine Intelligence},
  volume={7},
  number={8},
  pages={1355--1368},
  year={2025},
  publisher={Nature Publishing Group UK London}
}

Section 5: License

MIT