Topological indices and machine learning techniques for quantitative structure property relationship analysis of antidiabetic drugs

Diabetes mellitus is a chronic metabolic disorder that disrupts glucose balance, leading to severe complications in various organs. Current treatments fall short of effectively managing the disease, highlighting the need for innovative drug discovery and optimization approaches. In modern quantitative structure-property relationship (QSPR) research, topological indices based on molecular structure and connectivity are crucial.

These indices predict essential physicochemical properties, aiding in drug-likeness and therapeutic potential assessment. This study investigates degree-based, degree-sum-based, and reverse-degree-based topological indices using a bond partitioning approach for antidiabetic drug molecules. The QSPR analysis shows significant correlations between these indices and key physicochemical properties, validating their predictive value.

Predictive models are built to estimate properties like boiling point, complexity, heavy atom count, molecular weight, molar refractivity, polarizability, flash point, molar volume, and enthalpy of vaporization. Advanced machine learning algorithms, such as random forest and XGBoost, improve model performance by capturing nonlinear relationships between molecular descriptors and physicochemical properties, resulting in robust and interpretable models.

The research was conducted by researchers from Loyola College in Chennai, India, and Qassim University in Buraydah, Saudi Arabia.