Hazard classification systems alone can’t be used for carcinogenicity risk assessment without large uncertainty due to the need to consider exposure, dose, and mode of action for any given chemical. For mechanistic insight in silico and in vitro approaches can generate information for an assessment. To explore the value of some of the current methods, we investigated how such approaches can inform on modes of action and how the evidence correlated with carcinogenic potency. A dataset of 882 chemicals was created which contained IARC classifications as group 1, 2a, 2b and 3, of which 288 had in vitro mutagenicity (Ames test) and cytogenicity assay (chromosome aberration and/or micronucleus) and 173 had Gold TD50s for carcinogenicity potency. We found that for 138/173 both the mutagenicity and cytogenicity call was concordant of which the majority (116) of these were positive, however these did not easily extrapolate to high-potency carcinogens. The other 22 compounds mutually agreed on a negative call, indicating the presence of non-genotoxic pathways.
The in silico model for mutagenicity was highly predictive of the Ames test, whilst the current chromosome damage model has a high false positive rate. Furthermore, our expert-based system was able to provide mechanistic insight, such as information on the possible chemical modes of action for each chemical class. In all cases there was no simple correlation between the in vitro assay call, the IARC groupings, and TD50 ranges. We present how the use of in vitro methods can support mechanistic understanding and vice versa, they can inform what testing strategy to pursue in the absence of further in vivo data. These systems provide valuable information and can be used as weight of evidence in support of a regulatory assessment.