Publicly supported compilations of mutagenicity and carcinogenicity data are available for a significant number and variety of environmental and industrial chemicals and, to a lesser extent, pharmaceutical chemicals. These datasets represent tremendous past investment in

in vivo


in vitro

chemical toxicity testing, primarily driven by government regulatory concerns. These datasets are also the historical informational basis from which virtually all past structure-activity relationship (SAR) models of mutagenic and carcinogenic activity have been derived and from which mechanism-based SAR inferences pertaining to these endpoints have been gleaned. It follows that the nature, representation, and availability of these data exert a governing influence on the success of derived SAR models. Less appreciated, however, is the role that SAR modeling, itself, can play in assessing data quality, consistency, and completeness. Furthermore, SAR modeling can offer objective means for assessing information content as a function of how these data are pooled, classified, or otherwise interpreted by toxicologists and regulators. In this sense, existing representations of mutagenicity and carcinogenicity data constitute the working interface between toxicologists and SAR modelers.