Computer-based prediction models in regulatory toxicology

Abstract

The increasing regulatory safety demands for the submission and registration of chemicals, pesticides, and pharmaceuticals as well as several amendments in animal protection legislation have exacerbated the dilemma of regulatory toxicology, where on the one hand the required scientific contributions for the protection of workers, consumers, or patients are constantly augmented while on the other hand the number of experimental animal studies should be reduced for ethical and economic reasons. One way to resolve this dilemma could be the use of computer-assisted systems to predict toxic effects. These “in silico” tools have experienced improvements in their performance and predictive power over the past three decades. They could therefore contribute to hazard identification and risk assessment at least for some toxicological endpoints. The number of areas of application of in silico tools and systems will undoubtedly increase in the future. However, knowledge of how these systems work, on the underlying data quality, and their respective limitations are prerequisites for a sensible application.