Computational chemistry solutions supporting chemical safety assessment: Lessons learned for using in silico approaches

Computational methods have become increasingly important for evaluating consumer safety to chemicals. Approaches to predict complex endpoints such as chronic toxicity require a robust mechanistic basis, such as anchoring to relevant Adverse Outcome Pathways (AOPs). The EU COSMOS Project (www.cosmostox.eu) within the SEURAT-1 cluster of projects (www.seurat-1.eu) has developed computational in silico tools to support safety assessment for repeated dose toxicity for cosmetics-related substances. A number of fundamental issues for computational toxicology have been identified from the COSMOS Project which can be applied beyond the field of cosmetics: (1) Knowledge creation requires high quality data. A well curated database associated with highly defined chemical structures has been created in the freely available COSMOS Database (http://cosmosdb.cosmostox.eu). Importantly, it provides not only access to existing test data, but can facilitate the creation of new knowledge through data mining, e.g. new chemotypes for organ level toxicity and the development of thresholds of toxicological concern. (2) Biokinetic processes must be taken into account. Biokinetic models are essential to extrapolate from in vitro experiments to in vivo target organ concentrations (IVIVE) and in vivo bioavailability. Physiologically based kinetic (PBK) models can assist in extrapolating route-to-route exposures. (3) Established modelling techniques can be modified for use in risk assessment. Molecular modelling approaches, from drug discovery, have been successfully transferred to predictive toxicology, changing from hit identification with a minimum of false positives to screening chemicals to evaluate potential effects with a minimum of false negatives. (4) It is essential to make models easily accessible and usable. To facilitate the uptake of computational modelling methods, COSMOS models were implemented in a user-friendly, transparent and freely available form as KNIME workflows. In this way they are useful to support safety assessment, not as black box software approaches, but also available for adaption to the specific applications. The funding from the European Community's 7th Framework Program (FP7/2007-2013) COSMOS Project (Grant Agreement N° 266835) and Cosmetics Europe is gratefully acknowledged