Mapping external references

UBERON conveniently provides multiple external references within UMLS Metathesaurus with each code. e.g. FMA, MSH, and often also CUIs. We are combining multiple ontologies for the HuBMAP program into a Knowledge Graph and when ingesting UBERON we find that if we use all of the external references, then each UBERON code may have multiple CUIs (the CUI(s) of the provided external reference code(s) or CUI). We suspect that with the goal of maximizing semantic interoperability we would choose to allow all these mappings to exist but the consequence is that many UBERON codes then have multiple meanings so we then also add each UBERON relationship among two UBERON codes to each of its underpinning concepts. This works for us, and seems correct and verbose and may maximize the interoperability we seek, but is there risk we are creating "invalid" relationships? So, the key question here is this: if we choose to only map the CUI provided directly with each UBERON code and not leverage other vocabulary's codes and THEIR CUIs then we lose many opportunities for "query by other vocabulary" (concept/CUI interoperability) while we gain having no multi-concept UBERON codes. Which is better to use when? Are there specific circumstances or types of codes we should watch out for in doing this work? For example when we ingest vocabularies for "application ontologies" we use strict one-code-one-CUI so the relationships behave predictably in an application but when we want to use UBERON for a "general ontology" to map concepts in a "leaky" semantic way we use all external references. As developers of UBERON does that seem appropriate to you?

No repository files, tests, or entry points are named. Start by reviewing the external-reference and CUI ingestion context described in the issue; a useful outcome would be a documented decision on when to retain direct and indirect mappings for application and general ontology use.