Identification of potential natural product derivatives as CK2 inhibitors based on GA-MLR QSAR modeling, synthesis and biological evaluation

Abstract

Protein kinase CK2 is a validated target for cancer therapy. Many natural products have shown inhibitory activity against CK2 as potential anti-cancer drug candidates. A compatible quantitative structure-activity relationship (QSAR) model of natural products is necessary to identify the structural determinants related to their biological activities and provides valuable clues for the discovery of natural leads as anticancer drugs. In this study, genetic algorithm (GA) and multiple linear regression (MLR) methods, combined with preferred molecular descriptors, were employed to build QSAR models of CK2 natural product inhibitors. The best model, composed of eight molecular descriptors, yielded Q²_Loo = 0.7914 and R² = 0.8220 for the training set and Q²_ext = 0.7921 and R²_ext = 0.7998 for the test set, indicating the model’s robust reliability and high predictability. As a proof of concept, a true external test set, distinct from the training and test sets, was synthesized and tested in vitro to verify the predictive ability of this model. The predicted pIC₅₀ values of 13 compounds showed less than 30% relative error (including 10 compounds with relative errors less than 20%), further validating the predictive performance of this model. And compound M18, M24, and M26 were identified as potential CK2 inhibitors with the predicted pIC₅₀ values of 11.29, 8.79, and 12.03 respectively. Furthermore, the underlying structural mechanisms through which key molecular descriptors influenced their inhibitory activities against CK2 were elucidated. All these results provide valuable information for the discovery of CK2 inhibitors.

Graphical Abstract