LAUSR

Pancreatic cancer is an aggressive cancer, usually with poor prognosis, as it is mostly discovered at an advanced stage of development where treatment is challenging. Using principal components analysis (PCA) of variable selection, multiple linear regression (MLR), multiple non-linear regression (MNLR) and the artificial neural network (ANN), 2D-QSAR models for the anti-pancreatic cancer activity are developed from a set of twenty three molecules of 1, 2, 4-triazole derivatives to build the QSAR models. The well generated MLR and MNLR models exhibit the cross validation coefficients Q2 of 0.51 and 0.90, respectively. Moreover, the predictive ability of those models has been evaluated by the external validation using a test set of four compounds with predicted determination coefficients R2test of 0.936 and 0.852, respectively. The artificial neural network (ANN) method has shown a correlation coefficient of 0.896 with an architecture 3-2-1. The obtained results indicate the validation and the good quality of the 2D-QSAR models.