QSAR Model for Monoamine Oxidase B (MAO-B) Inhibitors

This repository contains a QSAR (Quantitative Structure-Activity Relationship) model developed to predict the pIC50 value for Monoamine Oxidase B (MAO-B) inhibitors using a Random Forest Regressor. The model is designed to assist in the identification of potential MAO-B inhibitors, a target relevant to Parkinson's Disease and other neurological conditions related to dopamine depletion.

Project Overview

Monoamine Oxidase B (MAO-B) is an enzyme that breaks down dopamine in the brain. Inhibiting this enzyme helps increase dopamine levels, which is a critical treatment approach for Parkinson's Disease (PD). By predicting the inhibitory activity (pIC50) of chemical compounds on MAO-B, this model helps screen potential drug-like molecules.

The model was built using bioactivity data from ChEMBL and molecular descriptors calculated via PaDEL using 881 PubChem fingerprints. The machine learning model used is a Random Forest Regressor, and the model is deployed using Streamlit for user interaction.

Model Features

• Algorithm: Random Forest Regressor
• Target Variable: pIC50 value (logarithmic inhibitory concentration of a compound)
• Input Format: The model requires two columns as input:
  • SMILES notation: Structural representation of the compound.
  • ChEMBL ID: Identifier for the bioactive molecule from the ChEMBL database.

Dataset

The model is trained on bioactivity data extracted from the ChEMBL database. Molecular descriptors were calculated using PaDEL-Descriptor software, which generated 881 PubChem fingerprints for each molecule. The model predicts the pIC50 value, which is a common measure used in QSAR studies to quantify the potency of a compound as an inhibitor.

---

Model Workflow

1. Data Preprocessing:

   • The dataset is preprocessed to ensure consistency and handle missing values.
   • The molecular descriptors are calculated for each compound using PaDEL-Descriptor, which outputs PubChem fingerprints.

2. Model Training:

   • A Random Forest Regressor is used for modeling. The model was trained and optimized using cross-validation to predict the pIC50 values.

3. Deployment:

   • The model is deployed using Streamlit. Users can upload their data (containing SMILES and ChEMBL IDs) to obtain predicted pIC50 values for their compounds.

---

Literature Background

Monoamine Oxidase B (MAO-B) inhibitors have been explored extensively as therapeutic agents for Parkinson's Disease and related disorders. MAO-B is primarily responsible for the breakdown of dopamine in the brain, and its inhibition can help maintain dopamine levels, alleviating symptoms of Parkinson's. Studies have shown that selective MAO-B inhibitors reduce the oxidative stress caused by dopamine metabolism and help slow down neurodegeneration.

Several QSAR models have been proposed to predict the potency of MAO-B inhibitors based on their molecular structures. In this model, PubChem fingerprints are used as molecular descriptors, which provide a comprehensive representation of the chemical features of molecules. Random Forest is chosen as it is robust against overfitting and performs well in high-dimensional spaces, making it an ideal choice for QSAR modeling.

Dependencies

The following libraries are required to run the model:

• pandas
• numpy
• scikit-learn
• PaDEL-Descriptor
• Streamlit

Install the required libraries using pip:

pip install pandas numpy scikit-learn streamlit