mol-opt: a Python Toolbox for Molecular Design

Please note: mol-opt is under active development and certain functionalities are yet to be fully implemented.

🚀 Installation

To get started with our package, please first ensure that PyTorch is installed with a CUDA version compatible with your device. You can find detailed installation instructions on the official PyTorch website. Once PyTorch is set up, you can install the remaining dependencies for our package by executing:

pip install .

Please note, our package has been tested and confirmed to work with Python 3.8. We recommend using this version to ensure compatibility and optimal performance.

💻 Usage (3 lines of code)

Oracle can be a string refering to a TDC oracle, e.g., qed, jnk3, gsk3b:

from molopt.graph_ga import GraphGA
optimizer = GraphGA(smi_file=None, n_jobs=-1, max_oracle_calls=10000, freq_log=100, output_dir = 'results', log_results=True) 
optimizer.optimize(oracle='qed', patience=5, seed=0)

or a callable function that takes a smiles string as input and return a real value:

from molopt.graph_ga import GraphGA
from rdkit import Chem
from rdkit.Chem import Descriptors
optimizer = GraphGA(smi_file=None, n_jobs=-1, max_oracle_calls=10000, freq_log=100, output_dir = 'results', log_results=True) 

def mol_wt(smi):
    m = Chem.MolFromSmiles(smi)
    if m is None:
        return 0
    else:
        return Descriptors.MolWt(m)

optimizer.optimize(oracle=mol_wt, patience=5, seed=0)

Note that all methods are by default desigend to maximize the oracle function. We support three hyperparameter tune strategies:

default: e.g., optimizer.optimize(oracle='qed', patience=5, seed=0).
configuration file: e.g., for graph GA, optimizer.optimize(oracle='qed', config='molopt/graph_ga/hparams_default.yaml', patience=5, seed=0).
keyword update: e.g., for graph GA, optimizer.optimize(oracle='qed', patience=5, seed=0, population_size=120).

You can follow the format of yaml file to define the hyper-parameters.

🤝 Contributing

If you wish to contribute, please refer to the CONTRIBUTING.md file. We welcome pull requests for bug fixes, feature requests, and improvements to our code.

📜 License

This project is licensed under the terms of the Apache 2.0 License.

💼 Support

If you need help with the tool, you can raise an issue on our GitHub issue tracker. For other questions, please contact our team.

📞 Contact

Please contact futianfan@gmail.com and gaowh19@gmail.com for help or submit an issue.

Cite Us

This package was developed as a spin-off from our paper (NeurIPS 2022). If you find this package useful, please consider citing:

@article{gao2022sample,
  title={Sample efficiency matters: a benchmark for practical molecular optimization},
  author={Gao, Wenhao and Fu, Tianfan and Sun, Jimeng and Coley, Connor},
  journal={Advances in Neural Information Processing Systems},
  volume={35},
  pages={21342--21357},
  year={2022}
}

wenhao-gao/mol-opt