Leveraging AI and Linear Regression for Advancements in Cheminformatics

Abstract

Accurate prediction of molecular solubility is essential in fields such as drug discovery, environmental chemistry, and material science. This study investigates the effectiveness of Linear Regression for predicting solubility based on a large dataset of molecular structures and corresponding solubility values. By applying feature selection and data preprocessing techniques, we optimized the model to achieve strong performance, with a Mean Squared Error (MSE) of 0.00 and a perfect coefficient of determination (R2) of 1.00. Linear Regression proved to be both accurate and interpretable, offering valuable insights into the molecular properties that influence solubility. When compared to other machine learning methods, it demonstrated competitive predictive power while maintaining simplicity and transparency. These findings highlight Linear Regression as a reliable and practical tool for solubility prediction, with broad applications across scientific and industrial domains.