Computational toxicology encompasses database creation, approaches for processing large amounts of data to discover novel associations and formulate hypotheses, and the development of in silico models to simulate chemical perturbations of biological systems leading to toxicological outcomes. This research is being fueled by the creation of structured, computable databases (e.g., bibliographic, factual, knowledgebase) and sophisticated informatics capabilities for processing and mining available biological and chemical data. Applied to developmental toxicology, the overarching objective of this research is to improve our ability to predict the means by which chemicals and chemically induced changes in biological systems are linked to adverse developmental outcomes in animals or humans. Toward this goal, public databases are being built using ontologies and controlled vocabularies; these are being linked to standardized representations of chemical structure for merging and expanding existing data inventories, and bioinformatics and visualization approaches are being tailored to the challenge of data-driven discovery relevant to developmental toxicology. For illustration, we explored chemically induced toxicity phenotypes in two distinct, nonoverlapping developmental toxicity databases (one environmental and the other pharmaceutical). We found that chemicals in the two distinct sets caused common phenotypes as well as distinct, nonoverlapping phenotypes in the same target organs. These observations underscore the need to build databases that span a large, diverse chemical space so as to broadly sample all relevant mechanisms for adverse developmental outcomes, as well as to enable chemoinformatics analysis and structure–activity-based approaches to detect common chemical features or properties linked to particular phenotypes. Automated text mining of the peer review literature, bioinformatics analysis, and sophisticated visualization tools can both confirm and expand upon existing knowledge of developmental systems to convey associations across planes of information. Such combined approaches are building libraries of features linking chemicals to genes to phenotype and are proving essential to supporting the interpretation of the new high-content and high-throughput biological screening data being generated for large numbers of chemicals. Ultimately, these new data, knowledge sources, and approaches will feed the development of virtual tissues/embryos that will enrich our understanding of system-level responses and help fulfil the vision of toxicology in the twenty-first century.