Identification of \u27structural alerts\u27 and associated mechanisms of action of mammary gland carcinogens in female rodents

A new structure-activity relationship (SAR) approach to modeling was utilized to study mammary gland carcinogens. A set of chemicals tested for mammary tumorigenesis that have been analyzed in the Carcinogenic Potency Database (CPDB) were subjected to several computational analyses in an attempt to predict each chemical’s carcinogenic potential. A total of six learning sets (rat and mouse mammary gland carcinogen, CPDB rat and mouse, and female-specific rodent models) were developed and validated using a SAR modeling algorithm called categorical-SAR (cat-SAR). The predictive cat-SAR program evaluates active and inactive compounds of known biological activity and predicts their biological activity from this categorical data. Overall, this study demonstrates the usefulness of cat-SAR and its successful application in developing ‘structural alerts’ to breast carcinogenicity. The resulting rat and mouse mammary carcinogen models achieved an 82.0% (sensitivity 76.7%; specificity 87.5%) and 80.6% (sensitivity 80%; specificity 81.8%) concordance between experimental and predicted results, respectively. Likewise, the general CPDB mouse and rat models were both 70% predictive. Corresponding sensitivity and specificity values were 74.2 and 66.7% and 70.4 and 68.5%, respectively. The analyses indicate the capability of cat-SAR in identifying molecular fragments that potentially interact with cellular components present only in the targeted cell type (e.g., breast tissue cells). Moreover, this method is expected to help pre-determine structural alerts to carcinogen-induced mammary cancer. Identification of these ‘structural alerts’ can assist in understanding mechanisms involved in making a normal breast cell cancerous. Using the results of these analyses, it is possible to classify and rank structurally diverse chemicals as to their potential to induce mammary gland cancer