Abstract
Developing alternative methods for carcinogenicity tests remains challenging. We have been working to develop a read-across method to predict non-genotoxic carcinogenicity by using chemical structural information and in vitro test results related to carcinogenic mechanisms. In read-across, the toxicity of the target substance is predicted from the toxicity of tested substances (i.e., source substances) with similar chemical and/or biological properties. To this end, we collected the results of two-year rat carcinogenicity tests of pesticides, and the presence or absence of tumors is set as the objective variable. The outline of the method is as follows: 1) Calculate the Euclidean distance between substances based on molecular descriptors calculated by using commercial software. 2) Select substances within a certain distance from each test substance as source substance candidates. 3) Conduct in vitro experiments related to carcinogenicity for test substances and primary source substances. 4) Select only substances whose experimental results match the test substance as the final source substance. 5) Perform read-across with the final source substances with the incidence rate of each tumor in the dataset as the criterion for positive judgment. We mainly work on tumors associated with epithelial cell damage, such as the nasal cavity, stomach, and bladder tumors, and those associated with nuclear receptor activation and drug-metabolizing enzyme induction, such as liver and thyroid tumors. In this session, I will share our recent progress of this study for discussion.
