Linear and nonlinear QSAR models of acute intravenous toxicity of organic chemicals for mice

Abstract

The QSAR analysis of acute intravenous toxicity for mice of 68 monofunctional chemical compounds is presented. These compounds are referred to seven chemical classes: hydrocarbons (6 chemicals), alcohols (13), amides (22), amines (12), ethers (5), ketones (7), and nitriles (3). Preliminary consideration of data for these chemical compounds showed the necessity of consideration of nonlinear toxicity — descriptor relationships in addition to linear toxicity — descriptor relationships. The linear and nonlinear QSAR models were considered for each indicated class of organic chemical compounds. Analogical models were constructed for the whole set of the monofunctional chemical compounds. The statistical parameters and robustness of nonlinear models were better than those for linear models. Replacement of a lipophilicity parameter for descriptors characterizing molecular weight and H-bond ability in nonlinear models also improved their statistical characteristics. It was concluded that in the case of nonlinear dependence between the intravenous toxicity of monofunctional chemical compounds and their lipophilicity one may expect nonlinear behavior of this dependence for compounds containing more than one functional group. This conclusion was further checked by generating linear, parabolic, and bilinear models for small clusters containing structurally related compounds with several functional groups and several clusters where nonlinear QSAR models exhibited better statistical criteria compared with linear models were found.