Drug properties versus substance properties / Entity normalization

[southalln]

If you filter to principle substances that are antiviral, oseltamivir is NOT included, but it is included if you list all substances.
https://drugs.ncats.io/substances?q=(comments:%22Pharmacologic%20Substance%5BC1909%5D%7CAnti-Infective%20Agent%5BC254%5D%7CAntiviral%20Agent%22)%20AND%20(%22oseltamivir%22)&facet=Development%20Status%2FUS%20Approved%20Rx&page=1
https://drugs.ncats.io/substances?q=(comments:%22Pharmacologic%20Substance%5BC1909%5D%7CAnti-Infective%20Agent%5BC254%5D%7CAntiviral%20Agent%22)%20AND%20(%22oseltamivir%22)&facet=Development%20Status%2FUS%20Approved%20Rx&facet=Substance%20Form%2FPrincipal%20Form&page=1

The principle substance for that drug is actually K6106LV5Q8
When we filter on principle substance AND classification text such as antiviral --- this record is excluded because K6106LV5Q8 doesn't inherit the same classifications as its members 4A3O49NGEZ ... GSRS is generally pretty good about including the classifications into multiple records, but not in this specific case.

You might see this as a GSRS error ... or ... we should have stitcher gather all of the classifications (and other info) and spew a canonical set of them in the same way we do for development status ... basically all 'drug' properties come from stitcher/additional-data. All 'substance' properties come from original GSRS records.