Essential and desirable characteristics of ecotoxicity quantitative structure-activity relationships

Quantitative structure-activity relationships (QSAR) developed and applied in the prediction of ecotoxic potencies far out number those in other areas, such as health effects. There are yet to be any formal guidelines for the development of ecotoxicological QSARs. Despite this, the depth and breadth of our knowledge of QSARs as they apply to ecotoxicology, especially short-term aquatic toxicity, allow for the formulation of characteristics that appear to be essential and/or desirable for high-quality QSARs. The three components of a QSAR are the biological activity, the property/structural descriptors, and the statistical methodology. Problems may arise from all three components and may be compounded by interactions between them. In an effort to minimize any tribulations associated with development and application of ecotoxic QSARs, a number of essential or desirable characteristics have been identified. Ecotoxicological data used in formulating the QSAR must be reliable, of high quality, and reflect a well-defined and continuous endpoint; this dataset should be diverse both in terms of potency and chemical structure (i.e., property). Descriptors used in formulating the QSAR should be of high quality, reproducible, of a number and type consistent with the endpoint being modeled, and when possible allow for a mechanistic interpretation of the QSAR. The statistical process used in formulating a QSAR should be as rigorous as possible, appropriate for the endpoint being modeled, and allow for the development of as easily interpretable (i.e., transparent) QSARs as possible. The resultant QSAR should be validated, only used within the descriptor space and chemical domain of the model, and relied on in relation to the total weight of evidence; precision of the QSAR and expectations from its application need to be related to the error in the original ecotoxicological and descriptor measurements. Finally, development of QSARs should be through the interaction of a group of multidisciplinary experts.