The OBO Corpus

In the following we will describe the OBO corpus in general. The OBO corpus contains 164 ontologies,

• 9 of which failed to download: bto idomal mirnao miro ogi ogsf sep tads tgma,
• 6 of which failed to be parsed by ROBOT (or were empty): dinto nomen omrse rnao sepio xlmod
• 0 of which failed to be processed by the data collection pipeline (timeout 60min):
• 2 of which failed were manually excluded: ncbitaxon pato

resulting in a total of 147 valid ontologies for the following analysis. Full breakdown:

How important is the PATO for OBO foundry ontologies?

We say that an axiom ax is PATO-dependent if the axiom contains one or more PATO-classes. An ontology is PATO-dependent if it contains at least one PATO-dependent axiom. We say that an axiom is p-dependent if it contains a particular PATO class p. In the following, we will quantify the importance of PATO for OBO ontologies by (1) counting the number of PATO-dependent ontologies and (2) determining the prevalence of PATO-dependent axioms. If a TBox axiom (“SubClassOf”, “EquivalentClasses”,“DisjointClasses”,“DisjointUnion”) contains a set of classes [C1,C2..,Cn], all of which are PATO classes (i.e. having a PATO iri), we make the assumption that this is a PATO axiom. Those PATO axioms are removed from the axiom analysis (but not from the ontology level analysis; if the PATO class is mentioned, it is counted as mentioned). This mitigates overcounting PATO-dependent axioms that are imported from PATO and then phyiscally merged into the release artefact (which happens relatively often).

How important are the individual PATO classes?