Within the framework of the European REACH regulation1, a large  number of (eco)-toxicological but also physico-chemical properties are  required for the registration of substances in the perspective of  industrial scale production. For reasons of time, cost, availability of  laboratories and ethics (in case of experiment on animals), companies  are encouraged to present alternative testing strategies, such as  computational chemistry and molecular modelling tools. Indeed, the REACH  regulation recommended the use of alternative methods to experiment and  in particular QSPR methods that can predict properties using only the  molecular structure of substances. These methods also benefit from  recognized validation principles, established by OECD, to ensure their  transparency, reproducibility and accuracy, in terms of fitting,  robustness and predictive capabilities and to allow their use in a  regulatory context.

Beyond this regulatory context and  considering the fast development and availability of computers, such  models allow industries to qualify new hazardous substances in a risk  assessment process and developers of new chemicals to integrate safety  considerations in the early steps of their development, even before the  synthesis of identified molecules of interest. It can facilitate the  rational design of new or alternative processes and products, for  example when testing is expensive or impractical or when ‘green’  alternatives are a must. In this way, R&D can focus on the most  promising candidates and this will cut costs and development time,  allowing timely elimination of future failures. So, the integration of  predictive and experimental knowledge will aid decision-making.   While such models have already been developed for some physico-chemical  properties, only few of them followed the different OECD validation  criteria, in particular for hazardous physico-chemical properties  (flammability and explosivity). First attempts were made to predict  reactivity hazards of organic peroxides using the QSPR approach but a  huge need is still growing to enhance the safety knowledge of these  reactive substances. This proposal addresses the development of two  complementary predictive methods, QSPR and small-scale tests, for  investigating hazardous properties of these chemicals based on solid  experimental databases.