Curcumin and derivatives against human colon cancer (HCT-116). Computationnel

In order to look for new cancer drugs, the Quantitative Structure-Activity Relationship (QSAR) is a tool used to build models capable of helping us to propose and accurately predict the activity of new candidates. In this context, a series of curcumin analogues have been studied against human colon cancer HCT-116. For QSAR models, we used the DFT method with the functional B3LYP and the 6-31 G(d) basis to calculate the quantum chemical descriptors. To calculate the constitutional, physico-chemical and topological descriptors, the molecular mechanics method (MM2) was used. To establish reliable models, multiple linear regression (MLR), multiple non-linear regression (MNLR) and neural network (ANN) have been used. The values of the determination coefficients show the predictive quality of the proposed models, internal validation, external validation, randomization test and applicability domain confirm robustness and predictive power of established models for new compound proposal