Gene Ontology classification: Can we infer from term T that one of the child terms is also a correct label?

Question

[Lorrit]

Let's assume that we found out that gene G can be labeled with the Gene Ontology term T. In the Gene Ontology T has 4 children T0, T1, T2, T3, each connected with term T through a "is a" relationship.

Can we infer that since gene G is labeled as term T, it also has to be one of the child terms T0, ..., T3? In other words, is the Gene ontology complete in the sense that the child terms of a term T list all possible more specific functions / processes / components of T? So any time that term T is present, at least one (or maybe exactly one?) of its child terms has to be present?

[cpad0112]

Ontology is DAG. In general (in most of the cases), child inherits parent term, not the other way around i.e gene with T0 term is also classified T term, but a gene with T term may not be assigned T0. In short T is superset of T0..T3..Tn.

Answer 1

[Jean-Karim Heriche]

Short answer: no. First, the ontology is known to be incomplete in the sense that not all areas are covered. Since our knowledge is always incomplete and the ontology lags behind current knowledge, the ontology will likely never be complete. Second, an ontology is a directed graph in which the semantic relation flows from the leaves to the root so, using your example, T0 is a T doesn't imply T is a T0. Also the child terms T0, T1,... are not necessarily mutually exclusive. Taking a more concrete example, if a gene is involved in DNA replication (GO:0006260) it doesn't mean it is involved in positive regulation of DNA replication (child term GO:0045740). DNA replication also has child terms nuclear DNA replication and mitochondrial DNA replication. Being involved in DNA replication doesn't imply that the gene is involved in either nuclear or mitochondrial DNA replication, it can be both or none.