This is a cheminformatics tool developed in the Brennan Group at the Centre for Medicines Discovery in the University of Oxford for the discovery of novel aromatic heterocyclic bioisosteres. HCIE identifies and ranks bioisosteric replacements using shape and electrostatic similarity (ESP) to a query molecule. It supports one-vector and two-vector (hash-based) alignments, and uses an RDKit-based backend and parallel processing for scalability.
Given a query molecule with one or two defined exit vectors, HCIE searches a database of aromatic heterocycles and returns the top-scoring matches by:
- Aligning probe molecules to the query via exit vector matching.
- Scoring the alignment based on shape and ESP similarity.
- Sorting and returning the top alignments.
- Exporting results to
.txt,.sdf, and.pngfor further analysis.
This enables rapid identification of novel scaffold replacements that preserve key spatial features important for bioactivity
The Python dependencies are listed in requirements.txt, and are best satisfied using a Conda installation (miniconda or anaconda)
Clone this repo
git clone https://github.com/BrennanGroup/HCIEChange directory to the cloned repo
cd HCIENote: If you are unsure about changing directories using the command line, follow instructions here).
It is highly advisable to create a new virtual environment for this package. If using conda, this can be achieved using
conda env create -f environment.ymlOnce all the packages have installed, activate the environment with
conda activate hcie_env
When in the folder, run the following to install HCIE into the virtual environment.
pip install .Once the package has been installed into the virtual environment, it can be run from any directory as long as the virtual environment is activated.
Searches are based on a SMILES string representation of the query molecule, with one or two attachment points
indicated by dummy atoms. This tool
is very useful for generating SMILES expressions. Within a Python script (or session in the command line -
python) run the following:
from hcie import VehicleSearch
search = VehicleSearch('<INSERT SMILES HERE>', name='<INSERT NAME HERE>')
search.search()
This will start a search, which will take anywhere from 4 to 15 minutes depending on the search type (one-vector searches tend to take longer, as a larger number of probe ligands are explored). The results will be deposited in a directory NAME_hcie_results, with the following structure:
├── NAME_hcie_results
│ ├── NAME_aligned_results.sdf
│ ├── NAME_results.png
│ ├── NAME_results.txt| File Name | Description |
|---|---|
NAME_aligned_results.sdf |
The 3D structures of the top 50 highest-scoring ligands, aligned to the query ligand in the alignment of highest similarity. |
NAME_results.png |
Grid image of the structures of the top 50 returned molecules for easy inspection. |
NAME_results.txt |
Tab-delimited file containing scores (shape, ESP, total), conformer info, and SMILES for each ligand in the searched library, ordered by total score. The exit-vectors are illustrated in the SMILES strings for ease of downstream enumeration. |
from hcie import VehicleSearch
search = VehicleSearch('[R]c1ccccn1', name='2-pyridine')
search.search()
from hcie import VehicleSearch
search = VehicleSearch('[R]c2ccn1ncc([R])c1n2', name='repotrectinib')
search.search()
We welcome contributions to HCIE!
If you'd like to contribute:
-
Report Bugs / Request Features
Open an issue on GitHub to report bugs, suggest improvements, or request new features. -
Submit Code Improvements
- Fork the repository.
- Create a new branch for your changes.
- Make your edits with clear commit messages.
- Submit a pull request with a brief explanation of your changes.
-
Add New Molecules or Datasets
If you’d like to expand the searchable space or add more chemical series, please reach out or submit an issue describing the source and scope of your additions. -
Documentation and Examples
Contributions to improve the documentation or add example notebooks / use cases are also greatly appreciated.
If you have used HCIE in your research and would like to share your feedback or results, we’d love to hear from you!
The scoring methodologies in HCIE are based largely on those described in EspSim by Esther Heid and colleagues, and we gratefullly acknowledge their contribution to this project.
This project is licensed under the MIT License.
You are free to use, modify, and distribute this software with proper attribution.
See the LICENSE file for full terms and conditions.