Taxonomic Ranks

From phenoscape
Revision as of 04:44, 10 April 2008 by Peteremidford@yahoo.com (talk | contribs) (Developing an Ontology for Taxonomic Ranks)

Developing an Ontology for Taxonomic Ranks

When the Teleost Taxonomy Ontology (TTO) was submitted to OBO, the suggestion was made that the terms for taxonomic ranks (e.g., Family, Genus, Species) should be broken out. At present, taxonomic ranks are included in the TTO and cross referenced to similar terms in the NCBI taxonomy ontology. Although the process of constructing an ontology of rank terms is straightforward, there are some semantic issues that need to be resolved.

The current implementation can be diagrammed as follows:


       Cyprinidae   -------------------->  Family
            ^             has_rank           
            |                                
       is_a |                                
            |                                
         Davario    -------------------->  Genus
            ^             has_rank           
            |                                
       is_a |                                 
            |                                
 Davario aequipinnatus ---------------->  Species
                         has_rank


Presently, the rank terms are simply subclasses of taxonomic_rank. There is no relation defined between the rank terms. The has_rank relation, as defined in the NCBO ontology is a meta_data relation (c.f. OWL annotation properties), which means it is intended to be ignored by any reasoner.

As there is clearly an ordering among the rank terms, it would be worthwhile to define an ordering relation between terms so that the term 'family' is indicated as 'larger' or 'more inclusive' than the term 'genus.'

There has been some question as to the nature of this ordering relation. There appear to be two points of view:

  1. A special relation exists between taxonomic ranks. It is certainly transitive and antisymmetric.
  2. The relation is simply part_of. Part_of is transitive and antisymmetric.


Chris Mungall's response to this question was:

 3. What about imposing an ordering among ranks and, if so, is part_of the appropriate relation?
 
 so with any ontology you should ask "what are the instances". In this case, the instances are best considered to be terms/classes/categories rather
 than something tangible in nature. This takes us close to weird metaclass modeling territory.
 
 but not to worry. I would say reserve part_of for "real" part_of relations, between objects and processes. I would just go with a custom relation for 
 ranks. I don't have strong opinions on what you name it - above/below? more_ancestral_than?
 
 Declare the relation transitive



In order to resolve whether the rank ordering relation is simply part_of, it may be necessary to clarify the property 'has_rank' which is used to relate a particular taxon to its rank. This diagram may clarify where these relations and properties fit in the taxonomic ontologies.

According to the documentation for the NCBI taxonomy ontology, which shares this structure, 'has_rank' is a metadata relation. As such, it does not have any meaning for a reasoner. The assertion that 'has_rank' is a metadata relation might be subject to revision.

The is_a hierarchy in the left-side column indicates that taxa are sets, a minority view among philosophers of biology. If taxa are individuals, the is_a hierarchy should be replaced by a part_of hierarchy. However, that does not require the 'rank ordering' relation for level terms to be part_of because has_rank is a metadata relation.

There is another property of the 'rank ordering' that may be called, by analogy with the rational numbers, density. The rational numbers are considered dense because between any two rational numbers a third rational number can be found. Likewise, within limits, between any two rank terms another term can be inserted: for example by using sub- and super- prefixes.