The Translator UI has to do complicated things to try to figure out what the specific types of a clique returned by NodeNorm are. However, NodeNorm itself knows the answer clearly:
- For non-conflated cliques, there is always a single Biolink type, which we then expand with supertypes.
- For conflated cliques, it is the set of Biolink types for the identifiers in this clique (we could also provide a list, but a set seems to make more sense to me).
We currently have a type field that provides the full Biolink tree. We could add a specific_types or biolink_types field that will be a list of one or more Biolink types that are included within this clique in the same order as the conflated cliques. Since this information is available on every query, I would suggest not having a flag to request this information but to just return it with every normalization.
Should be solved alongside or in PR #350, which adds clique_leaders -- if you use that flag, you can iterate through clique_leaders[].type to look up all of the Biolink types. But having the specific_types for all queries will also be useful, I think.
/cc @gprice1129 who will need to consume this information.
/cc @ctrl-schaff who will need to implement this in NodeNorm ES once we figure out the right way to implement both this and #340.
The Translator UI has to do complicated things to try to figure out what the specific types of a clique returned by NodeNorm are. However, NodeNorm itself knows the answer clearly:
We currently have a
typefield that provides the full Biolink tree. We could add aspecific_typesorbiolink_typesfield that will be a list of one or more Biolink types that are included within this clique in the same order as the conflated cliques. Since this information is available on every query, I would suggest not having a flag to request this information but to just return it with every normalization.Should be solved alongside or in PR #350, which adds
clique_leaders-- if you use that flag, you can iterate throughclique_leaders[].typeto look up all of the Biolink types. But having the specific_types for all queries will also be useful, I think./cc @gprice1129 who will need to consume this information.
/cc @ctrl-schaff who will need to implement this in NodeNorm ES once we figure out the right way to implement both this and #340.