QSPR probing of Na+ complexation with 15-crown-5 ethers derivatives using artificial neural network and multiple linear regression

Abstract

A quantitative structure–property relationship (QSPR) study is performed to develop a model, relating to Na+ complex stability constant (log K) and the structure of 74 derivatives of 1,4,7,10,13-pentaoxacyclo-pentadecane ethers (15C5). Stepwise Multiple Linear Regression (SMLR) and Artificial Neural Network (ANN) methods have been exploited as linear and nonlinear methods, respectively to build the QSPR model. MOPAC software embedded in ChemOffice 2004 package was used for the minimizing energy using semi-empirical AM1 method. The optimum structures have been applied to generate more than 50 descriptors using available servers in ChemOffice 2004. The five most important constitutional, steric, electronic, thermodynamic and molecular descriptors were selected using the common preselection method combined by SMLR method. SMLR and ANN models were constructed based on the five selected descriptors. Both proposed models efficiently predict log K of 15C5 complexes. However, the results of ANN were more effective respect to SMLR model. This phenomenon reveals that log K of 15C5 complexes have a deviation from linear behavior related to the selected descriptors.