A major goal of the emerging field of computational toxicology is the development of screening-level models that predict potential toxicity of chemicals from a combination of mechanistic in vitro assay data and chemical structure descriptors. In order to build these models, researchers need quantitative in vitro and ideally in vivo data for large numbers of chemicals for common sets of assays and endpoints. A number of groups are compiling such data sets into publicly available web-based databases. This article (1) reviews some of the underlying challenges to the development of the databases, (2) describes key technologies used (relational databases, ontologies, and knowledgebases), and (3) summarizes several major database efforts that are widely used in the computational toxicology field.