Informal ontologies that include less explicit information can make a useful contribution when the end-user is somewhat knowledgeable about the field (Bard, 2005). For example, mapping gene expression identifiers (GeneIDs) by stage, tissue and region in development and extracting this information for a sensitive period of development to a particular chemical or class of chemicals can provide information about pathway-level responses to exposure. An informal ontology defining target tissue can then include detailed tissue geometry and morphogenetic boundary conditions drawn from conventional histology (Bard, 2005). Interoperability can be built with ontology tools such as Protégé.
id: MP:0001297 ! microphthalmia
intersection_of: EMAP:304 !TS12, eye
vulnerability_start: EMAP:304 !TS12
vulnerability_end: EMAP:3003 !TS18
associated_with: Pax6, Fgf8
The distribution of a particular phenotype or combination of features can be summarised by ‘frequency’ and ‘redundancy’. We can define frequency as any reference to the term in a document, which may be positive (exposure-related), negative (mentioned but not observed), or noise (not exposure-related). We can define redundancy as the number of occurrences for each record. Using redundancy as a quantitative metric, we can apply multivariate clustering to give information on the association of a particular organ system with a chemical, group of chemicals, or animal model. Essentially it is a measure of sensitivity. The pattern of terms appearing together (co-occurrences) can give information on syndromes for a chemical or species as a measure of specificity. Thus, some pairwise statistics would be useful to assess how often two particular terms appear jointly in an experimental condition or dose group.