The first step for developing this software has been to collect about 4,000 scientific studies of dye properties owned by the most important dye producers, which are members of the three European Dyes REACH Consortia.

Initially, three selected endpoints were designated for the statistical analysis, that is, in-vitro gene mutation study in bacteria (Ames test), skin sensitisation and acute fish toxicity. The QSAR models were calculated on a structural similarity approach with the addition of specific alerts to prevent false positives or negatives. The outcome of the models consists of the estimated endpoint for a substance (target) along with its degree of reliability quantified by the number of similar substances and the average of their similarity to the target. The models were developed and validated according to the OECD principles and REACH requirements for QSAR. The new models were evaluated in comparison with already existing QSAR models for the selected endpoints and the obtained results were satisfactory in terms of both specificity and sensitivity (85% or Ames test, 91% for skin sensitisation, and 79% for fish toxicity). In particular, the proposed models were able to broaden the Applicability Domain of QSARs towards disconnected structures that typically characterize dyes, overcoming the limitations of the existing QSAR models, which cannot predict most of the dyes, being calibrated on really different sets of molecules.

This is a unique and extremely innovative project – says Dr Michela Kahlberg, CEO of REACH&Colours Italia and manager of the three European Dyes REACH Consortia for textile, leather and paper – entirely financed by the private industry and the Bicocca University – in which we strongly believe, that will allow to better understand chemicals deeply used and widespread such as organic dyes, protecting human health, our environment and avoiding animal testing.

Further information can be found at: (