Evaluation of dimedone-derived compounds as inhibitors against human colon cancer: Insights from 2D-QSAR, ADMET prediction, Osiris, Molinspiration, and molecular modeling

Abstract

For this research, we performed a 2D-QSAR analysis on a set of 34 molecules derived from dimedone with inhibitory activity against human colon cancer (HT-29). The investigation incorporated principal component analysis (PCA), multiple linear regression (MLR), multiple non-linear regression (MNLR), and artificial neural network (ANN). The evaluations of the QSAR models demonstrated high predictive power (R²_MLR = 0.884; R²_CV (MLR) = 0.86), (R²_MNLR = 0.810; R²_CV (MNLR) = 0.879), and (R²_ANN = 0.9; R²_CV (ANN) = 0.89). The reliability of the models was validated through internal, external, Y-randomization, and validation domain applicability assessments. Utilizing the QSAR model predictions, we designed four new molecular structures that exhibited superior inhibitory activity against the HT-29 human colon cancer cell line compared to the 34 previously tested compounds. Subsequently, we examined the ADMET predictions, Molinspiration, and Osiris properties of the four compounds. The results revealed excellent ADMET predictions and Molinspiration for all four compounds, while only one designed compound fulfilled all Osiris properties. Molecular docking was employed to investigate the bindings between the newly designed molecule C and the c-Met protein. The findings indicated that the newly designed compound exhibited high stability in c-Met. Finally, MD simulations were employed to assess the stability and binding modes of compound C. Based on the MD results, compound C shows promise as a potential c-Met agonist candidate. Overall, our results suggest that this investigated compound has the potential to serve as a novel inhibitor against human colon cancer.