Prediction of cellular toxicity of halocarbons from computed chemodescriptors: a hierarchical QSAR approach.

Abstract

A hierarchical quantitative structure-activity relationship (HiQSAR) approach was used to estimate toxicity and genetic toxicity for a set of 55 halocarbons using computed chemodescriptors. The descriptors consisted of topostructural (TS), topochemical (TC), geometrical, semiempirical (AM1) quantum chemical, and ab initio (STO-3G, 6-31G(d), 6-311G, 6-311G(d), and aug-cc-pVTZ) quantum chemical indices. For the two toxicity endpoints investigated, ARR and D(37), the TC indices gave the best cross-validated R(2) values. The 3-D indices also performed either as well as or slightly superior to the TC indices. For the four categories of quantum chemical indices used for the development of predictive models, the AM1 parameters gave the worst performance, and the most advanced ab initio (B3LYP/aug-CC-pVTZ) parameters gave the best results when used alone. This was also the case when the quantum chemical indices were used in the hierarchical QSAR approach for both of the toxicity endpoints, ARR and D(37). The models resulting from HiQSAR are of sufficiently good quality to estimate toxicity of halocarbons from structure.