3,852 interactions from DGIdb could not be mapped

[saramsey]

There are 3,852 interactions in DGIdb that could not be mapped into KG2 because no controlled identifier for the drug was available. Is there a way to map them into the KG2, without severely compromising modularity of the build process?

[dkoslicki]

@saramsey close due to RTXteam/RTX#788?

[saramsey]

nope

[saramsey]

Reminder to myself: @ericawood described a potential partial fix for this issue, in RTXteam/RTX#449. Maybe there is a way to have dgidb_tsv_to_kg_json.py also take umls-ncit.json as an optional input file?? It would break modularity a bit but would potentially (partially) fix this issue.

[ecwood]

Looking into this again, the interaction breakdown has changed significantly:

{
    "GuideToPharmacology": 8618,
    "DrugBank": 2965,
    "NCI": 2798,
    "TdgClinicalTrial": 1186,
    "PharmGKB": 914,
    "JAX-CKB": 601,
    "CIViC": 211,
    "TEND": 120,
    "TALC": 109,
    "ClearityFoundationClinicalTrial": 85,
    "MyCancerGenome": 76,
    "MyCancerGenomeClinicalTrial": 73,
    "ChemblInteractions": 71,
    "CGI": 65,
    "FDA": 37,
    "ClearityFoundationBiomarkers": 34,
    "DTC": 19,
    "CancerCommons": 15,
    "OncoKB": 14,
    "COSMIC": 5,
    "DoCM": 4
}

Based on this, I am going to focus on mapping the DrugBank identifiers, since those appear to be the easiest to map.

[saramsey]

@ecwood OK to close, based on commit 58d58b1?