This repository explores the chemical space associated with SARS-CoV-2 according to manually curated data at UB-CeDD (Buea, Cameroon). The project is developed in the context of a research visit of Prof. Fidele Ntie-Kang.
An important component of UB-CeDD's work is the manual curation of synthetic and natural product compounds targetting a specific pathogen. UB-CeDD has assembled a team of curators who have catalogued compounds of potential SARS-CoV-2 activity. In this project, we set up a chemoinformatics pipeline to expand the chemical space around these manually-curated compounds using tools that can work in low-resource settings. Importantly, while SARS-CoV-2 may not be a top priority globally, the pipeline is conceived such that in can be applied to other pathogens and disease areas. Therefore, this repository should be viewed as a proof-of-concept. Central to the pipeline is the Ersilia Model Hub, our resource of open source AI/ML models for drug discovery.
The pipeline consists of the following steps:
- Manual curation of SARS-CoV-2 related compounds at UB-CeDD. Both synthetic and natural products are annotated.
- Automated characterisation of the manually-curated compounds using the Ersilia Model Hub. This includes calculation of interesting features such as synthetic-accessibility and natural-product-likeness, as well as ADME properties.
- Ultra-large scale similarity search against Zinc and Enamine REAL databases. For this, CHEESE is used to query using 2D and 3D similarities.
- Post-processing and aggregation of similarity search results.
- Characterisation of the resulting chemical space based on SARS-CoV-2 predictors as provided by the REDIAL-2020 suite of models, as available from the Ersilia Model Hub.
All data used in the project is publicly available. The manually curated molecules can be found under data/original and a compilation of both Natural Products and Synthetic Derivatives is available in all_molecules.csv. Additional datasets downloaded from their public sources (DrugBank, ChemDiv Coronavirus) are also in the /data folder. In addition to manually curated and publicly available data, the folder contains the calculations of several descriptors and molecular properties for each of the datasets, obtained via the Ersilia Model Hub. Each dataset is referenced with an eos identifier.
The /scripts folder contains the data processing pipelines and the necessary scripts to precalculate all descriptors and molecular properties using Ersilia's models. The Python scripts (.py) can be executed one-by-one following their numerical order. The ersilia_models_* scripts can be executed assuming all the necessary models have been fetched using the ersilia command-line interface. The /results folder contains the molecules obtained during the ultra-large scale library screening using CHEESE.
You will need a .env file with the CHEESE_API_KEY. You need to place this file in the sars-cov-2-chemspace directory. The file should have the following information:
CHEESE_API_KEY="your_api_key"
To obtain a this key, please follow instructions from the CHEESE documentation. CHEESE is a remarkable resource developed by a small team. Please acknowledge this resource if you find it useful and be mindful of resource consumption when querying their tool.
The notebooks folder contains a relatively unstructure set of Jupyter notebooks showcasing data analysis recipes and publication-ready figure generation. Figures are stored in the figures folder.
We are currently working on a draft related to this project. Feel free to read this draft in this online document.
The code in this repository is available under a GPLv3 license and the data and figures under a CC-BY-4 License.
The Ersilia Open Source Initiative is a tech non-profit research organisation aimed at developing data science capacity in the global south.