The QSAR Paradigm to Explore and Predict Aquatic Toxicity

The necessity to follow the environmental fate of chemicals and in particular the aquatic toxicity not only of existing compounds, but also of those to be developed, has challenged the application of quantitative structure–activity relationships (QSAR) already at their genesis. The present chapter provides an overview of the historical evolution of environmental QSAR and its development as an autonomous discipline. It highlights the progress in the essential elements of QSAR methodology, such as molecular representation, statistical algorithms, suitable to treat the growing amounts of multiple endpoints, the strengthening in the criteria for model validation and for the definition of the applicability and tries to track the inherent philosophy rather, and then compile the numerous QSAR models published for the variety of toxicity endpoints assessed in different aquatic species at the three trophic levels. The huge diversity in the chemicals space resulted in the formulation of structural alerts for classification purposes in chemical categories, an essential requirement for the construction of a robust model or for choosing the appropriate model among the existing ones for toxicity predictions. Narcotics obey to rather simple rules with hydrophobicity being the single physicochemical parameter in the relevant QSAR models. On the other hand reactive and specifically acting chemicals, which exhibit "excess toxicity" demand the application of more elaborated statistical tools and the exploitation of the big arsenal of molecular descriptors. Hydrophobicity is still a major partner, however other parameters mostly related to electrophilicity and hydrogen bonding, as well structural and topological descriptors may show important contribution. The QSAR paradigm is facilitated by the existing and continuously being updated databases for ecotoxicological endpoints, as well by the development of software or web platforms either, freely or commercially available, which incorporate a variety of statistical tools and models for different endpoints and chemical classes. Commonly used software tools are presented in this chapter. The most important fact is that the QSAR paradigm in aquatic toxicity is proved to be a success story, since it has been adapted by the regulatory organizations, which have developed their own software and in silico platforms or support existing ones. Read-across techniques for filling data gaps are also implemented in many of platforms. In this aspect, the goal for the construction of any new model is not only to be more efficient than existing ones leading to more reliable and accurate predictions but also to meet the requirements of regulatory authorities. More to the point, the last years emphasis has been given to the integration of the results from multiple modeling tools and read across approaches for improved reliability of predictions. Several integration techniques have been suggested for this purpose. Moreover, the development of quantitative activity–activity relationships (QAAR) for interspecies toxicity predictions would reduce the vast toxicity space, as shaped by the multiple endpoints at a variety of aquatic species.