ML LogS Prediction

This project is my first machine learning (ML) project using Python.
It focuses on predicting the aqueous solubility (logS) of molecules based on molecular descriptors.

The workflow includes:

1. Data Loading – Fetches the Delaney solubility dataset from a public CSV.
2. Data Preparation – Splits the dataset into features (MolLogP, MolWt, NumRotatableBonds, AromaticProportion) and target (logS), and performs train-test splitting.
3. Model Building – Implements and trains:
   • Linear Regression
   • Random Forest Regression
4. Model Evaluation – Calculates MSE and R² for both training and test datasets to compare model performance.
5. Data Visualization – Plots predicted vs. experimental logS for model assessment.

Requirements

• Python 3.10+
• pandas, numpy, matplotlib, scikit-learn

Usage

Run the notebook in Google Colab or Jupyter. The notebook automatically loads the dataset from a public URL.

You can experiment with different regression models or hyperparameters to improve prediction accuracy.

Dataset

The Delaney solubility dataset contains 1,145 molecules with features describing their molecular properties and experimental solubility values (logS).

Goal

The main goal of this project is to practice ML workflow including data preparation, model building, evaluation, and visualization, while gaining insights into molecular solubility prediction.