🧪 Molecular Solubility Prediction using Machine Learning

This project uses machine learning to predict the logarithmic solubility (logS) of molecules from their structural descriptors using two models: Linear Regression and Random Forest Regressor. The entire implementation is contained within a Jupyter Notebook viewable directly on GitHub. 📁 Project Structure

. ├── hypothesis.txt # Step-by-step methodology explanation ├── visuals_for_ML/ # Folder containing generated plots and visuals ├── Solubility.csv # Dataset with molecular descriptors ├── machine.ipynb # Jupyter Notebook with complete implementation └── README.md # You're reading it!

📌 Note: The code is written in a Jupyter Notebook (machine.ipynb).
You can preview and read the full code, output, and graphs directly on GitHub.

🎯 Objective

To predict the solubility (logS) of chemical compounds using four key molecular features:

MolLogP – Lipophilicity

MolWt – Molecular Weight

NumRotatableBonds – Flexibility of the molecule

AromaticProportion – Proportion of aromatic atoms

The target variable (y) is logS. ⚙️ Environment Setup ✅ Step 1: Create and Activate Virtual Environment

Create virtual environment named 'inv'

python -m venv inv

Activate it:

Windows:

inv\Scripts\activate

macOS/Linux:

source inv/bin/activate

✅ Step 2: Install Required Packages

pip install pandas numpy matplotlib scikit-learn jupyter

💻 Run the Notebook in VS Code

If you're using Visual Studio Code:

Open the project folder.

Press Ctrl+Shift+P → Select Python: Select Interpreter → Choose your inv environment.

Install the Python and Jupyter extensions if not already installed.

Open machine.ipynb.

Run the notebook cells interactively.

You can also open machine.ipynb directly on GitHub to preview the code and results.

📊 Dataset Overview

The file Solubility.csv includes: Column Description MolLogP LogP (lipophilicity) MolWt Molecular weight NumRotatableBonds Bond flexibility indicator AromaticProportion Ratio of aromatic atoms logS Solubility (target variable) 🔁 Project Workflow

1. Data Preparation

   Load Solubility.csv using pandas

   Separate X (first 4 features) and y (logS)

   Visualize feature columns using .drop() method

2. Data Splitting

   Use train_test_split from sklearn.model_selection

   80% training data (915 rows), 20% test data (229 rows)

3. Model 1: Linear Regression

   Fit a linear regression model

   Predict logS on training and test sets

   Evaluate using:

    Mean Squared Error (MSE)
   
    R² Score (R-squared)
   

4. Model 2: Random Forest Regressor

   Train a RandomForestRegressor model

   Predict and evaluate performance using the same metrics

   Compare to Linear Regression results

5. Formatting & Visualization

   Use DataFrame.transpose() and .columns for tidying results

   Create graphs and visuals using matplotlib and numpy

   All plots saved in visuals_for_ML/

📈 Sample Visuals Linear Regression Random Forest

(Replace with actual filenames from your visuals_for_ML/ folder) 📊 Final Comparison Metric Linear Regression Random Forest MSE Moderate Lower R² Score Acceptable Higher Visual Fit Less precise More accurate

🎯 Conclusion: Random Forest Regressor yields better predictions and is more robust for this problem. 🧰 Tech Stack

Python 3.x

Jupyter Notebook

VS Code (recommended IDE)

Libraries:

    pandas

    numpy

    matplotlib

    scikit-learn

    jupyter

🚀 How to Run This Project

Clone the repository

git clone https://github.com/your-username/your-repo-name.git

cd your-repo-name

Setup virtual environment

python -m venv inv source inv/bin/activate # or inv\Scripts\activate on Windows

Install dependencies

pip install -r requirements.txt # (optional if you have a requirements file)

Launch Jupyter Notebook

jupyter notebook machine.ipynb

Or, open machine.ipynb directly in VS Code or GitHub to explore the full implementation.