Validation activities and use cases
You won't be right the first time when you do the ontology work. So apply the ontology candidate in a conversion of database 1, fix it, then do database 2, etc. Errors: unmodeled things, badly modeled things, "obvious" conversions that result in falsehoods, dealing with missing information.
Proposed first test case: All of reactome, but start with a single reaction, because if you can't do one thing, you can't do 100,000. After one reaction, maybe look at influenza life cycle (again from reactome).
Aim for a large number of pathway databases converted into the new ontology, with tools to help make this efficient. Not clear what tools we'll need. SPARQL (e.g. on a DB viewed as triple store) takes you a long way, maybe, in covering a DB in layers - convert query results for one query, then another query, then another, until you've covered the entire DB.
We'll find bugs (inconsistencies) in data sources. Report these back. Check for inconsistencies across ontologies as well.