Biology Reference
In-Depth Information
Ashburner was already aware of the work on ontologies in artifi cial
intelligence and medicine, and his proposal included the suggestion that
the consortium's gene ontology (GO) be compatible with, or at least
translatable to, more generalized ontologies.
18
He argued that the GO
had to be more sophisticated than a mere list of terms: “The advantage
of a structured graph over a fl at keyword list is that you could have a
representation of part and whole, it's easy to maintain, it's easy to use,
and you have information built into the structure of the graph, whereas
with fl at keywords, there is only one thing you can do with that: sort
it alphabetically.”
19
A structured graph looks somewhat like a fl ow dia-
gram in which terms are linked together by arrows. Figure 4.2 shows a
small, simplifi ed section of the GO structured graph. The arrows show
how the GO terms are logically linked to one another—in this case,
they show that a cell comprises various parts and subparts. “Cell” has
two parts, “cytoplasm” and “nucleus.” “Nucleus,” in turn, has three
subparts, “nucleolus,” “nucleoplasm,” and “nuclear membrane.” This
structure can be represented in a computer fi le in XML format with the
nodes as GO terms and the connecting arrows as GO relationships. A
stripped-down version might look like this:
<go:term>
<go:name> cell </go:name>
</go:term>
<go:term>
<go:name> nucleus </go:name>
<go:part_of> cell </go:part_of>
</go:term>
<go:term>
<go:name> nucleolus </go:name>
<go:part_of> nucleus </go:part_of>
</go:term>
<go:term>
<go:name> nucleoplasm </go:name>
<go:part_of> nucleus </go:part_of>
</go:term>
<go:term>
<go:name> nuclear_membrane </go:name>
<go:part_of> nucleus </go:part_of>
</go:term>
Using the <go:part_of> tag, a simple computer program (a parser)
could read off the hierarchy from this XML. In other words, it could
