Model-Validation-eos30f3

🏚️ Structure

This repository is organised by folders:

• Data: Contains the raw data, processed data and Model predictions.
• Figures: Contains a collection of visualizations presented in PNG format..
• Notebooks: Houses the jupyter notebook files where most of the developmen took place.
• Src: Contains important functions I will re-use throughout the repository, to avoid typing them each time.

Model Abstract

This model leverages the ChemProp network (D-MPNN, see original Stokes et al, Cell, 2020 for more information) to build a predictor of hERG-mediated cardiotoxicity. The model has been trained using a dataset published by Cai et al, J Chem Inf Model, 2019, which contains 7889 molecules with several cut-offs for hERG blocking activity. The authors select a 10 uM cut-off. This implementation of the model does not use any specific featurizer, though the authors suggest the moe206 descriptors (closed-source) improve performance even further.

Model Characteristics

• Input: Compound
• Input Shape: Single
• Task: Classification
• Output: Score
• Output Type: Float
• Output Shape: Single
• Interpretation: Probability of blocking hERG (cut-off: 10uM)

Installation Environment

Check here

Getting Started

1. Install Ersilia Model Hub

   • Follow this step to install the Hub.

2. Fetch the Model and Install it locally

   ersilia -v fetch eos30f3
   

3. Serve the Model

   ersilia -v serve eos30f3
   

4. Make predictions with the Processed Datasets here

   ersilia -v api run -i data/Processed/1000_Molecules.csv -o data/Model_predictions/eos30f3_output.csv
   
   

Data Procurement Process

1. ChEMBL Data Procurement:

   • A Datasets of 3592 Small Molecules Compounds that has been aproved for Use was Downloaded in a CSV format, and stored in /data/Raw.

2. DMPNN-hERG Datasets Procurement:

   • The Model was trained on 7889 compounds with well-defined experimental data on the hERG and with diverse chemical structures assembled by Cai et al in their work published in J Chem Inf Model, 2019 here, it provides target variables with various cutoffs for hERG blocking activity.

3. External Datasets Procurement:

   • This hERG data set provides 648 drugs with binary labels, it was sourced from Therapeutics Data Commons, which is a collection of datasets for drug discovery.

References

• Publication

• DPMN-hERG GitHub page

• Eos30f3 GitHub page

License

All the code in this repository is licensed under a GPLv3 License.