MOA disease unique contraint failed

[korikuzma]

In #408, I am working on adding support for MOA prognostic data. This resulted in being able to harvest/transform more MOA assertions. During this transformation, I encountered a bug with MOA disease normalization resolution.

{
        "name": "Myelodysplastic Syndromes",
        "oncotree_code": "MDS",
        "oncotree_term": "Myelodysplastic Syndromes"
      }

and

{
        "name": "Myelodysplasia",
        "oncotree_code": "MDS",
        "oncotree_term": "Myelodysplasia"
      }

resolve to the same NCIt ID, however we create two separate Disease objects in MoaTransformer with the same id="moa.normalize.disease.ncit:C3247", which fails to satisfy the db uniqueness constraint.

Since MOA does not have disease "records" we will have to change how we're processing diseases. One thing I'm not sure about is which value to use for the label vs alternativeLabels. Typically, we only store normalizer information in the vicc_normalizer_data field.

[jsstevenson]

This looks like a data curation error on their part. Same oncotree code.

[korikuzma]

Ya, I was going back on forth on whether we should handle this or see if MOA could. I drafted a message to Brendan asking if they did any normalization on their side but didn't send and just ended up creating issue here and not sending. I can reach out to him to ask.

[korikuzma]

we just use oncotree for the normalization! the “name” is as written in the source text. we used an underscore in the upcoming release to differentiate “raw” values per a suggestion from James last year

[github-actions]

Closed by #413.