QSPR analysis of physicochemical properties of anti-cancer drugs using degree based graph theoretical descriptors

Abstract

Through topological invariants, chemical graph theory provides a rigorous mathematical framework for describing molecular structure and deciphering chemical behavior. This study builds quantitative structure property relationship (QSPR) models for a group of anticancer medications using ten degree-based descriptors, including the Zagreb indices, atom bond connectivity index, and Randić index. To determine the most predictive descriptors for important physicochemical properties, such as molecular weight, heavy atom count, molar refractivity, boiling point, molar volume, polarizability, and polar surface area, linear, logarithmic, quadratic, and cubic regression models are methodically examined. F-statistics, significance levels, and \(R^2\), adjusted \(R^2\) standard error are used to evaluate the model’s performance. The findings show that a number of degree-based indices have substantial associations with various drug features, with different regression types and descriptors yielding the best accurate predictions. Overall, the results show that graph-theoretical invariants are useful and effective techniques for predicting the physicochemical properties of anticancer drugs, providing important information for computational drug design.