Update build process to include DrugBank identifiers from RxNorm

[hoganwr]

We have been discussing at the DrOn monthly meeting the possibility of providing links to DrugBank. RxNorm does include DrugBank identifiers, I believe on IN term types, that we could incorporate somehow. I am kicking off this discussion to talk about how we do it. Annotation? If so which Annotation Property? Do we follow standard OBO dbXref process? (which I am not a big fan of, personally)

[matentzn]

I think this would be a very valuable addition, and I am personally totally fine with xref to start with!

[hoganwr]

It would be important to do it that way to stay consistent with how ChEBI classes that we import are annotated. Also, we note that ChEBI database_cross_reference annotations have CURIEs as values, and that Protege recognizes them and makes them clickable (which is neat, I am wondering how Protege does it since there is no DrugBank prefix in the OWL).

[hoganwr]

While thinking on it, it occurred to me that ChEBI already has them, and thus numerous classes in DrOn imported from ChEBI already have DrugBank IDs. And sure enough they do! For example CHEBI:28971 database_cross_reference DrugBank:DB00415

So we would want to engineer the process so that it only adds DrugBank IDs to classes with DrOn PURLs.

Then if for some reason we replace a DrOn class PURL with a ChEBI class PURL, we would want to only have the DrugBank ID from the ChEBI import and not also from DrOn natively after the replacement.

[hoganwr]

We could either (1) strip off annotations other than label when we replace DrOn ingredients with ChEBI terms or (2) just leave them there. I don't think they hurt much other than file size, but we don't have too many of htem (some months, RxNorm update results in zero new Dron-Chebi mappings).

[hoganwr]

The first step would be curating an Rxcui-DrugbankId mapping from RxNorm. We can add that to the templates, databases along the way, etc. One question is what happens when Rxcuis get obsoleted.