Predicting flavonoid physicochemical properties using topological indices and regression modeling.

Flavonoids, a diverse class of polyphenolic phytochemicals, exhibit multifaceted biological activities critical to human health. This study leverages degree-based topological indices (TIs) to predict six physicochemical properties of sixty flavonoids using linear, quadratic, and logarithmic regression models. Statistical validation via correlation coefficients ([Formula: see text]), Root Means Square Error (RMSE), and Mean Absolute Error (MAE) revealed robust predictive power, particularly for molar refractivity ([Formula: see text], RMSE [Formula: see text], MAE [Formula: see text]), molar volume ([Formula: see text], RMSE [Formula: see text], MAE [Formula: see text]), and enthalpy of vaporization ([Formula: see text], RMSE [Formula: see text], MAE [Formula: see text]). Quadratic models consistently outperformed linear/logarithmic approaches, indicating nonlinear relationships between TIs and properties. The methodology offers a cost-effective tool for prioritizing bioactive flavonoids in drug discovery, validated by strong agreement between predicted and experimental values for external compounds (e.g., Procyanidin B2: molar refractivity RMSE [Formula: see text]). This work bridges cheminformatics and QSPR, enabling rapid property estimation for polyphenolic systems.