cmungall
commented
9 years ago Thinking about this more. SPARQL or a query language is not required for this scenario. The constraint could simply be encoded as an equivalence axiom to a probe class. This would require absolutely zero changes in any part of the stack, everything is covered in #3. The user would see the probe class (e.g. "mechanism regulator, disallowed") as an additional inferred class. Maybe a soupcon of UI configuration to color these but even this should be handled by a generic configurable mechanism.
Of course there is the ongoing issue that Elk does not handle OPEs, which is bigger now we have switched to all-indiv.
Primary work is on the Minerva side but ultimately exposed on noctua
We'd like to have configurable (at the level of the Noctua admins, e.g. GO ontology group, set using github not the UI) checks on models. Models in violation would be highlighted in a similar way to owl-inconsistent models.
The check would probably be best specified as a SPARQL query. Analogous to what James is doing in ROBOT, can crib his code. Challenge: want to do over inferred graph.
The first test would be:
no protein binding instances should regulate any instance