Abstract
This paper selects 60 types of resveratrol analogues were selected and discusses the relationship between thermodynamic properties and the molecular descriptors, including the topological indexes of principal quantum number (0P2, 0P4, 0P3) and the sum total of them (∑0P). For the sake of simplicity, ∑0P is confirmed as the fundamental variable for the models through comparative analysis. Based on the characteristics of sine series, sin(k∑0P) (k=1, 2, 3, …) are incorporated into the models to eliminate the residuals, resulting in the 200 maximization of adjusted coefficient of determination (Radj ). The introduction of sin(5∑P) and sin(8∑P) successfully eliminates the residuals and enhances the predictive power. The global correlation coefficients are greater than 0.99, and even close to 1. The stability and predictive power of each model are tested with additional resveratrol analogues other than the test samples by the cross-validation method. The test reveals that the QSPR model has a satisfactory stability and good predictive power.
