300 12 Building Bioinformatics Ontologies
12.6.1 Uniform Hierarchy
To help understand large hierarchies one should try to make them as uniform
as possible. While uniformity is a subjective notion, there are some objective
criteria that one can use to help make a taxonomy more uniform:- Every level of a hierarchy should represent the same level of generality.
In other words, all subclasses of one class should be at the same concep-
tual level. Of course, this depends on the purpose of the ontology. One
way to classify animals would be to divide them into three subclasses: hu-
mans, domesticated animals, and wild animals. For people in everyday
life, these three subclasses are on the same conceptual level. However,
this is not the case when animals are classified genetically. - Every class that has subclasses should be subdivided into at least two and
no more than a dozen subclasses. Subdividing into a single class suggests
that the ontology is either incomplete or that the subclass is superfluous.
Subdividing into a large number of subclasses makes it difficult for a per-
son to understand or to navigate the taxonomy.
Unfortunately, these two criteria can conflict with each other. The taxon-
omy of living beings is a good example of this. The most general concept is
subdivided into domains, which are subdivided into kingdoms, which are
subdivided into phyla, continuing until one reaches individual species. The
notion of a phylum, for example, serves to identify a level in the taxonomy,
and every phylum represents the same level of generality throughout the hi-
erarchy. However, the price that one pays for this uniformity is that some
subclassifications consist of a single subclass while others consist of a large
number of subclasses.
When the number of subclasses is large, one can introduce new levels into
the hierarchy. In the taxonomy of living beings, additional levels are some-
times used to reduce the number of classes in a subclassification, such as
“subphyla,” “superphyla,” “suborders,” “superfamilies,” and so on. Unfor-
tunately, there is no easy way to deal with classes that have only one subclass.
In the case of the taxonomy of living beings, one can argue that the single
subclass is the only one that is currently known, leaving open the possibility
that others may have existed in the past or may be discovered in the future.
The speciesH. sapiensis the only species in the genusHomo. However, there
were other species in this genus in the past.