Ontological analysis of controlled vocabularies used in PSI/MSI supported XML standards

Abstract

Besides a plethora of formal ontologies, the requirement for simple data annotation has led to an increased use of so called controlled vocabularies (CV) in multiple omics communities. We analyze two of those CVs from an ontological viewpoint, highlight typical modelling errors and propose more adequate solutions. Discovered errors are discussed in the light of the OOPS ontology pitfall framework and the OBO Foundry naming conventions. As a result the CVs could be improved and the OOPS catalogue could be amended and expanded with new, previously missing error categories. In an outlook we discuss potential reasons for the error prevalence and analyse what criticism is justified for CV semantics and what `errors' are more valid for formal ontologies rather than CVs. We conclude that although many design principles valid for description logics ontologies are not relevant for semantically flat CVs and in turn there is a need for CV-best-practices that are not appropriate for description logics ontologies, there is room for improvement in the analysed CVs. The scope difference between CVs and formal semantics however should affect policy providers, which should narrow down the scope of their policies, i.e. by stating for each policy the expressivity regime for which it is valid.