The computational prediction of genotoxicity

Importance of the field: The computational prediction of genotoxicity is important to the early identification of those chemical entities that have the potential to cause carcinogenicity in humans. Areas covered in this review: The review discusses key scientific developments in the prediction of Ames mutagenicity and in vitro chromosome damage over the past 4 – 5 years. The performance and limitations of computational approaches are discussed in relation to published and internal validation exercises. Their application to the modern drug discovery paradigm is also discussed. What the reader will gain: Key highlights of a review of the recent scientific literature for the prediction of Ames mutagenicity and chromosome damage and an appreciation of the factors that limit the predictive performance of in silico systems. Take home message: Current in silico systems perform well in the mutagenicity prediction of the publicly-derived data on which they are based, but their performance outside the applicability domain is considerably lower. We conclude that it is the lack of mechanistic structure–activity relationships and limited access to high quality proprietary data which are holding back computational genotoxicity from reaching higher predictive levels.