We have had some succession using the PHES-ODM Search Model Context Protocol (MCP) and the ontology mapping tool that @dmanuel developed, to improve the ontology linkages within the PHES-ODM to better support controlled vocabulary usage.
However, the PHES-ODM exists at a niche that isn’t well ontoglogized, I would argue (public health, wastewater surveillance), and despite the strong and valiant efforts of our colleagues over at PHA4GE and the GENEPIO ontology, there are still a lot of gaps.
Now a lot of things are close, but I am unsure if they’re appropriate to link because of the purpose of their source ontology - for example, COVOC:0050061 is a ontology entry for a school closure as a public health action, but the ontology is only for COVID-19 pandemic study and work, so not useable in other contexts. Or the particularly rich APOLLO_CV, which has entries like APOLLO_SV:00000621 for lockdowns, but is intended for epidemiological modelling and simulation, not reporting real life actions. For these items I feel like it’s a “close but no cigar” kind of scenario.
For other items, like GENEPIO:0100953 for a grab sample and UO:0000180 for mass for unit volume, they’re useful, but more as a parent to certain ideas (like parent to horizontal and vertical grab samples, or for gram per cubic meter respectively).
I am wondering:
- Is continuing to pursue better ontology alignment worthwhile?
- Should we add a new parts table column, “ontologyParentID” to use where a satisfactory match is unavailable, but a parent exists?
- Should we just do what we can, matching mostly units and genes and pathogens, and call it a day for now?
Looking forward to hearing more feedback!