Improved QSARs for predictive toxicology of halogenated hydrocarbons

In our continuing efforts to provide a predictive toxicology capability, we seek to improve QSARs (quantitative structure-activity relationships) for chemicals of interest. Currently, although semi-empirical molecular orbital methods are hardly the state of the art for studying small molecules, AM1 calculations appear to be the method of choice when calculating quantum-chemical descriptors. However, with the advent of modern computational capabilities and the development of fast algorithms, ab initio molecular orbital and first principles density functional methods can be expeditiously applied in current QSAR studies. We present a study on halogenated alkanes to assess whether more accurate quantum methods result in QSARs that correlate better with experimental data. Furthermore, improved QSARs can also be obtained through development of new descriptors with explicit physical interpretations that should lead to better understanding of the mechanisms involved in the toxic response. We show that descriptors calculated from chemical intermediates may be useful in future QSARs.