Application of Genetic Algorithm - Multiple Linear Regressions to Predict the Activity of Rsk Inhibitors Publisher



Avval ZM¹ ; Pourbasheer E¹ ; Ganjali MR² ; Norouzi P^{2, 3}
Source: Journal of the Serbian Chemical Society Published:2015

Abstract

This paper considers the development of a linear quantitative structure-activity relationship (QSAR) model for predicting the ribosomal S6 kinase (RSK) inhibition activity of some new compounds. A dataset consisting of 59 pyrazino[1,2-α]indole, diazepino[1,2-α]indole, and imidazole derivatives with known inhibitory activities was used. The multiple linear regressions (MLR) technique combined with stepwise (SW) and the genetic algorithm (GA) methods as variable selection tools was employed. For more checking of the stability, robustness and predictability of the proposed models, internal and external validation techniques were used. Comparison of the obtained results, indicate that the GA-MLR model is superior to the SW-MLR model and that it is applicable for designing novel RSK inhibitors.