Closed ianwdunlop closed 8 years ago
I would allow all those fields to be indexed (but maybe exclude the @en). On the drugbank page http://www.drugbank.ca/drugs/DB01112, the identifiers are also given in other languages, maybe you could include those as well? There are also many more brand names available there. Also, it is good that the Drugbank identifier would find results (maybe not necessarily with the full uri, but just DB and number).
Is there some way to preserve in the elastic search index the fact that some text is a label, synonym or brand name? That information would be useful for ranking results.
@AlasdairGray You can boost the result rankings depending on the field type.
@danidi It needs a really deep dive into the drugbank RDF to figure out what is available. It seems to have changed a lot recently.
Now indexing:
drugbank:
compound:
graph: http://linkedlifedata.com/resource/drugbank
type: drugbank:drugs
properties:
- drugbank:brandName
- drugbank:genericName
- drugbank:chemicalIupacName
- drugbank:synonym
- drugbank:swissprotName
target:
graph: http://linkedlifedata.com/resource/drugbank
type: drugbank:targets
properties:
- drugbank:geneName
- drugbank:synonym
- drugbank:swissprotName
enzyme:
graph: http://linkedlifedata.com/resource/drugbank
type: drugbank:enzymes
properties:
- drugbank:geneName
- drugbank:name
See also Drugbank properties
@stain do you have an example with actual data? I'm wondering if the swissprotName in drugbank/compound contains the name of the protein the drug is acting on. Do we want to find this?
Also, is drugbank:synonym from compound different to drugbank:synonym from target?
For drugbank:Drug see https://github.com/openphacts/OPS_LinkedDataApi/blob/1.5.0/api-config-files/01_01_compoundInfo.ttl#L142 for the fields that are returned in the API call. Based on this here is the first version of SPARQL to grab the fields for Elastic Search to index
Sample response: