Drug Discovery Using Machine Learning

This project focuses on applying machine learning techniques in the field of bioinformatics to aid in drug discovery.

The aim is to accelerate the identification of potential drug candidates using predictive modeling, ultimately reducing the cost and time of early-phase drug development.

Objective

To build and evaluate machine learning models that can predict the bioactivity of chemical compounds against a biological target.

Project Structure

├──notebook/
└──CDD_ML_Part_1_bioactivity_data.ipynb                                        # data collection
└──CDD_ML_Part_2_Exploratory_Data_Analysis.ipynb                               # EDA      
└──DD_ML_Part_3_Acetylcholinesterase_Descriptor_Dataset_Preparation.ipynb      # data modelling
└──CDD_ML_Part_4_Acetylcholinesterase_Regression_Random_Forest.ipynb           # machine learning
└──CDD_ML_Part_5_Acetylcholinesterase_Compare_Regressors.ipynb                 # model evaluation+comparison

Dataset

The dataset contains molecular descriptors of various chemical compounds and their corresponding biological activity. It is sourced from public bioactivity databases like ChEMBL.

• Features: Molecular fingerprints, physicochemical properties
• Target: Bioactivity classification (Active/Inactive)

Requirements

Install the dependencies using:

pip install -r requirements.txt

How to Run

1. Clone the repository

git clone https://github.com/veydantkatyal/drug-discovery.git
cd drug-discovery

2. Launch Jupyter Notebook:

jupyter notebook

3. Open /notebooks and run the cells within python notebook.

Future Work

• Incorporate deep learning (Graph Neural Networks) for molecular representation

• Use of additional descriptors (e.g., SMILES)

• Hyperparameter tuning with Bayesian Optimization

Acknowledgements

• ChEMBL for providing the dataset

• RDKit for cheminformatics tools (future integration planned)

License

This project is open-source and licensed under MIT License