Entropy-weighted topological descriptors in QSPR modeling of antituberculosis drugs

In this study, we develop quantitative structure-property relationship (QSPR) model using entropy-weighted topological descriptors to predict the physicochemical properties of antituberculosis drugs. The molecular structures of fifteen drugs used to treat tuberculosis were taken from PubChem, and their conventional degree-based descriptors were enhanced via Shannon entropy to account for both structural magnitude and distributional irregularity. These entropy-augmented indices were incorporated into quadratic, cubic, exponential, and logarithmic regression models to describe structure-property relationships. Model validation was performed using unsupervised (Entropy-Property Concordance Index, EPCI) and supervised (Local Regression Concordance, LRC) methods. The results demonstrate that entropy-weighted descriptors significantly improve interpretability, predictive accuracy, and structural validation in QSPR modeling. This approach offers a robust computational framework that can be extended to the rational design and property prediction of macrocyclic ligands and supramolecular assemblies, supporting advances in host-guest chemistry and targeted drug delivery systems.