Water quality objectives for mixtures of toxic chemicals: problems and perspectives.

The need to develop water quality objectives not only for single substances but also for mixtures of chemicals seems evident. For that purpose, the conceptual basis could be the use of the two existing biometric models: concentration addition (CA) and independent action (IA), which is also called response addition. Both may allow calculation of the toxicity of mixtures of chemicals with similar modes of action (CA) or dissimilar modes of action (IA), respectively. The joint research project Prediction and Assessment of the Aquatic Toxicity of Mixtures of Chemicals (PREDICT) within the framework of the IVth Environment and Climate Programme of the European Commission, provided the opportunity to address (a) chemometric and QSAR criteria to classify substances as supposedly similarly or dissimilarly acting; (b) the predictive values of both models for the toxicity of mixtures at low, statistically nonsignificant effect concentrations of the individual components; and (c) the predictability of mixture toxicity at higher levels of biological complexity. In this article, the general outline, methodological approach, and some preliminary findings of PREDICT are presented. A procedure for classifying chemicals in relation to their structural and toxicological similarities has been developed. The predictive capabilities of CA and IA models have been demonstrated for single species and, to some extent, for multispecies testing. The role of very low effect concentrations in multiple mixtures has been evaluated. Problems and perspectives concerning the development of water quality objectives for mixtures are discussed.

Traditional versus WHIM molecular descriptors in QSAR approaches applied to fish toxicity studies.

Quantitative structure-activity relationships (QSAR) were performed on nine congenetic aromatic hydrocarbons. Acute response was evaluated in freshwater fish species. QSAR were built by Hansch's approaches and weighted holistic invariant molecular (WHIM) indices. The prediction power of QSAR from both approaches was evaluated. Single regression analysis derivated by Hansch's approach seem suitable for non-polar compounds. However, for all species, it has not a high predictive power (Q2(LOO)) of the biological activity from only K(ow) as molecular descriptor. Multiple regression analysis obtained from WHIM descriptors showed Q2(LOO) higher than 80%, indicating that molecular descriptors have a prediction power greater than K(ow).

QSAR and chemometric approaches for setting water quality objectives for dangerous chemicals.

In order to evaluate environmentally safe levels of dangerous chemicals, there is the need for a set of toxicological data on organisms representative of the ecosystems, which is often unavailable or inadequate. In this article, a predictive approach was applied to a set of 125 chemicals (derived from the European priority list in compliance with Directive 76/464/EEC), for which water quality objectives were available. Toxicological data on organisms representative of the aquatic environment (algae, Daphnia, and fish) were taken from the literature or predicted by means of quantitative structure--activity relationships. This provided toxicological data on all three organisms for 97 of 125 chemicals and on at least two organisms (Daphnia and fish) for the whole data set. Principal Component Analysis was applied in order to perform an a priori classification of chemicals based on toxicity data. Then several classification models, based on traditional and nontraditional molecular descriptors, were applied. Classification models gave results in agreement with the a priori classification as well as with the original water quality objectives classification. The behavior of some outliers was explained. The approach described appears to be a useful tool for the preliminary classification of chemicals that are dangerous to the aquatic environment for which toxicological data are inadequate.

QSAR approach to POPs screening for atmospheric persistence.

The environmental behaviour of a persistent organic pollutant (POP) is mainly controlled by its persistence, its tendency to undergo long-range transport (LRT) and its physicochemical properties. Atmospheric half-life is one of the criteria commonly used to study air persistence and LRT potential. For the 12 UNEP POPs and another 48 possible POPs, the mean and maximum half-life estimations for degradation in air are modelled using different molecular structure descriptors (atom and fragment counts, topological and WHIM descriptors), selected by Genetic Algorithm, in QSAR regression models. Both values are modelled to obtain an average estimate and a precautionary value for ranking and screening purposes. The models, validated for their predictivity, could be applied to predict unavailable data. Principal component analysis (PCA) was then used to explore the half-life data in addition to the physicochemical properties that are most relevant to atmospheric mobility; the aim has been to screen and rank POPs with regard to their tendency towards atmospheric persistence and mobility, and to obtain a persistence index in air and an LRT index. These indexes were also modelled by molecular descriptors, thus allowing a preliminary screening of new compounds.

QSAR approach for the selection of congeneric compounds with a similar toxicological mode of action.

The selection of compounds with a similar toxicological mode of action is a key problem in the study of chemical mixtures. In this paper, an approach for the selection of chemicals with similar mode of action, based on the analysis of structural similarities by means of QSAR and chemometric methods, is described. As a first step, a complete representation of chemical structures for examined chemicals (phenylureas and triazines) by different sets of molecular descriptors allows a preliminary exploration of similarity using multi-dimensional scaling (MDS). The use of genetic algorithm (GA) to select the most relevant molecular descriptors in modeling toxicity data makes it possible to develop predictive toxicity models. The final step is a similarity analysis, based again on MDS, using selected molecular descriptors, really relevant in describing the toxicological effect.