ANN and GMDH Algorithms in QSAR Analyses of Reactivation Potency for Acetylcholinesterase Inhibited by VX Warfare Agent

Abstract

Successful development of novel drugs requires a close cooperation of experimental subjects, such as chemistry and biology, with theoretical disciplines in order to confidently design new chemical structures eliciting the desired therapeutic effects. Herein, especially quantitative structure-activity relationships (QSAR) as correlation models may elucidate which molecular features are significantly associated with enhancing a specific biological activity. In the present study, QSAR analyses of 30 pyridinium aldoxime reactivators for VX-inhibited rat acetylcholinesterase (AChE) were performed using the group method of data handling (GMDH) approach. The self-organizing polynomial networks based on GMDH were compared with multilayer perceptron networks (MPN) trained by 10 different algorithms. The QSAR models developed by GMHD and MPN were critically evaluated and proposed for further utilization in drug development.