![[logo] US EPA](https://webarchive.library.unt.edu/eot2008/20090513063733im_/http://www.epa.gov/epafiles/images/logo_epaseal.gif){kind=logo}
The Virtual Embryo Project (v-Embryo™):
A computational framework for developmental toxicity
Knowledge Discovery
Information infrastructure
The knowledge-discovery element of v-Embryo™ requires specialized tools and resources for data-mining and text-mining. Components of the underlying data and metadata will be cawptured from gene expression repositories and the body of knowledge associated with developmental health and disease. Many database projects are underway to manually curate data from developmental endpoints and as such the structured data are reasonably well understood; however, unstructured data presents a different challenge. This information often holds the key to the major themes or ideas associated with the structured data. As such, patterns in the information can be analyzed and classified (e.g., document summarization and clustering) beyond simple hit-lists to formalize associative relationships (ontologies) and perform semi-automated feature mapping. This module is being developed at NCCT through ITS-ESE contract No.: 68-W-04-005, Task Order No. 058: Technical Support for Development of Developmental Systems Toxicity Network (DevToxNet) with Lockheed Martin.
Informal ontologies that include less explicit information about developmental processes and toxicities can make a useful contribution when the end-user is knowledgeable about the field. To build this on a case basis is best done through a broader network; hence, v-Embryo™ is piloting a Wiki-space for this purpose. The ability to assess efficacy in best practices must adapt constantly to advances in scientific knowledge. The Wiki will address the basic science foundation for normal embryology (embryo-formatics), a prioritization schema to maintain focus on research to support risk assessment, strategic planning of specific case examples to calibrate the system, and finally improved ways to translate the work more effectively.
Expert systems
Expert systems are computer applications that carry out a degree of logical reasoning similar to those of human beings. As such, they make subject-matter expertise available to non-experts. The two main components of an expert system are: one, a knowledgebase that holds a collection of rules, e.g. formalized truths extracted from actual experience; and two, an inference engine that draws appropriate deductions by logically compiling the set of rules triggered by an input query. A great deal of collective experience and expertise of the developmental toxicology community is likely required for a rule-based system to predict teratogenicity.
Whereas a v-Embryo™ knowledgebase can address qualitative issues such as whether a particular pathway is present at a particular stage of development, a quantitative inference engine simulator is required to predict what level of perturbation invokes developmental toxicity. This simulator draws from the knowledgebase to enable reveal implications or general principles, generalize based on specific cases, and infer the best explanation based on facts. The in silico simulation can motivate best practices by enabling ‘what-if’ kinds of experiments to systematically test hypotheses that identify moments in time at which interventions have extreme consequences (‘lever points’), and distinguish among pathways leading to structural and functional abnormalities.