Machine Learning-based Prediction of Physicochemical Properties of Curcumin Derivatives using Topological Indices

Abstract

Curcumin derivatives have emerged as promising bioactive compounds in pharmaceutical and medicinal chemistry. Predicting their physicochemical properties is essential for improving drug development and molecular optimization. This conference paper presents a machine learning framework that utilizes topological descriptors to estimate physicochemical properties of curcumin compounds. Multiple supervised learning algorithms, including Linear Regression, Random Forest Regression, Support Vector Regression, and Gradient Boosting methods, were implemented and evaluated. The study demonstrates that graph-based molecular descriptors can effectively capture structural information relevant to property prediction. Experimental results indicate that ensemble learning models achieve superior predictive accuracy and robustness compared to traditional regression methods. The proposed approach provides a computationally efficient and cost-effective strategy for cheminformatics-based molecular property prediction.