Externally validated QSPR modelling of VOC tropospheric oxidation by NO3 radicals

The troposphere is the principal recipient of volatile organic chemicals (VOCs) of both anthropogenic and biogenic origin. The persistence of these compounds in the troposphere is an important factor for the evaluation of their fate, and the possible negative effects to the environment and human health. In this study, the tropospheric lifetime of 166 VOCs, in terms of night-time degradation rates with nitrate radical (NO3), was modelled by the quantitative structure-property relationships (QSPR) approach. The multiple linear regression method was applied, in combination with the genetic algorithm-variable subset selection procedure, to a variety of theoretical molecular descriptors, calculated by the DRAGON software. The models were developed according to the OECD principles for regulatory acceptance of QSARs, with particular attention to external validation and applicability domain (AD). The external validation was performed on an unbiased external test set or by splitting the available data using self-organized maps or the random by response approach. The best QSPR models presented in this study showed good internal (range of : 89-92%) as well as external predictivity (range of Qext²: 75-89%). The AD of the models was analysed by the leverage approach, and was represented graphically in the Williams graph.