Background REACH (Regulation (EC) 1907/2006 on the registration, evaluation and authorisation of chemicals), which came into force in 2007, sets out a new regulatory regime for chemicals in Europe. It requires manufacturers and importers of chemicals to submit hazard assessments and safe-use information, and it is based on the principle that the use of chemical substances should be permitted only if they have been evaluated as safe for the intended use. It is expected that some 40 000 chemical substances will eventually be registered under REACH. REACH also promotes the use of non-animal testing of substances wherever possible. Non-animal test methods include computational models to assess chemical properties. Annex XI of REACH defines the criteria for models used for this purpose, known as quantitative structure-activity relationship models (QSAR). Objectives The aims of the 'CALEIDOS' project are: To provide practical guidance to the users of non-animal test methods such as QSAR, for the evaluation of chemicals under REACH; To organise a statistically sound evaluation of the results of 25 QSAR methodologies, by using them to predict the properties of chemicals registered under REACH, and by comparing the predicted values with the values derived from experiments, as included in the substance registration dossiers; To evaluate if read-across (applying results from one chemical to a similar chemical) and grouping (assuming similar properties in similar substances) methods can improve the predictions, by comparing the results from read-across and grouping exercises with those obtained from QSAR models for a set of selected chemicals; To develop and make freely available a web tool for predicting chemical properties; and To establish a network of experts from regulatory bodies, industry and scientific institutions, who will work to improve standardisation of the use of non-animal testing methods. Expected results The project expects to achieve the following quantified results: Production of guidelines on the use of non-testing methods (QSAR, read-across and grouping), including case studies, performance statistics for the non-animal testing methods, and chemicals safe-use indications; A tool, with 12 QSAR models freely available, for prediction of substance properties relevant for REACH; A 20% increase in the use of QSAR data for the assessment of substances under REACH; and The completion of an international exercise involving experts, on the reproducibility of computationally supported/non-supported read-across and grouping methods for the prediction of chemical properties.