QSAR studies and application of genetic algorithm - multiple linear regressions in prediction of novel p2x7 receptor antagonists’ activity

Quantitative structure-activity relationship (QSAR) models were employed for prediction the activity of P2X7 receptor antagonists. A data set consisted of 50 purine derivatives was utilized in the model construction where 40 and 10 of these compounds were in the training and test sets respectively. A suitable group of calculated molecular descriptors was selected by employing stepwise multiple linear regressions (SW-MLR) and genetic algorithm-multiple linear regressions (GA-MLR) as variable selection tools. The proposed MLR models were fully confirmed applying internal and external validation techniques. The obtained results of this QSAR study showed the superiority of the GA-MLR model over the SW-MLR model. As a result, the obtained GA–MLR model could be applied as a valuable model for designing similar groups of P2X7 receptor antagonists.

Keywords:

QSAR , genetic algorithms , P2x7 receptor antagonists , Purine derivatives