<正>Carcinogenicity is an important toxicological endpoint that poses the highest concern for human health.In this study,we develop a method that extracts structural alerts(SAs) and modulating factors of carcinogens based on pure statistical analyses.Firstly,Gaston algorithm,as a frequent subgraph mining method,detects efficiently substructures of any size,shape,and level of chemical details.Then,building and pruning of a molecular fragments tree select SAs with statistical significance in binomial test.Finally,modulating factors that are able to destroy the carcinogenicity potential of SAs are extracted by three self-defining rules.The results highlight that this method is a preferable method with higher accuracy,and the selected SAs are usable for prediction as well as interpretation.Moreover,our method can extract SAs from database using an automated and unbiased manner which has no use of a priori knowledge of mechanism of action, and are conveniently employed by a user.