European Chemicals Agency dossier submissions as an experimental data source: refinement of a fish toxicity model for predicting acute LC50 values.

As a result of the stringent data requirements of the Registration, Evaluation, Authorisation, and Restriction of Chemicals (REACH) regulation, a vast amount of ecotoxicological data has become available through the dissemination portal of the European Chemicals Agency (ECHA). As of April 2014, the database contained 12,439 unique substances from 47,909 dossiers. This vast database could be used to refine existing, or to create new, non-testing methods, such as quantitative structure-activity relationships (QSARs). Acute fish toxicity data were mined from the ECHA database using the eChemPortal; after filtering for single organic substances, 1159 experimental data points remained, representing 564 compounds. To evaluate the quality and accessibility of this data, the authors used the data to refine and improve an existing QSAR. The reliability of the data submitted to the ECHA database, as well as the effectiveness of the Klimisch scoring system, were assessed by comparing the refined QSAR with established QSAR benchmarks. The model developed meets all Organisation for Economic Co-operation and Development principles, has strong internal (leave-one-out internally cross-validated correlation coefficient [Q(2)(LOO)] = 0.91) and external (external coefficient of determination (predicted vs experimental [test set])) validation statistics, and can provide reliable fish median lethal concentration (LC50) predictions for non-polar narcotics. Although some issues with dossier misinformation were discovered, it was found that the ECHA dissemination portal is a valuable and reliable data source. When queried using the eChemPortal, chemical dossiers containing reliable data could be found quickly. The ECHA dissemination portal holds great potential for future QSAR development and improvement, such as updating QSARs within the Ecological Structure-Activity Relationships (ECOSAR) program.