Background:

EBI Gene2Phenotype contains human-curated, detailed gene-disease associations.

Our team has been working on a ingest for the current phase of Translator. The base MetaEdge is Gene (subject_form_or_variant) - Disease, with the predicate set using the "confidence" column value.

Right now, my mapping for "confidence" column values -> predicates is:

• "limited": related_to. I interpret the definition as saying there is AN association - it's just not causal (as far as we know) and it's unclear how "real"/meaningful it is. So these rows should have a predicate weaker than "causes"/"contributes to".
• "moderate", "strong", or "definitive": causes. There's moderate-definitive evidence that a gene DOES HAVE a causal role in this disease
• "refuted" or "disputed": remove rows. There's strong evidence that there ISN'T an association (negation) based on the definitions.

The data with the "limited" confidence value are therefore mapped to Gene (subject_form_or_variant) - related_to - Disease.

Question:

(1) Do you think another current predicate would be better to use? (or just keep using related_to, pending a gene-disease predicate refactor).

(2) I'm unsure of what Association class would match, and if any additions/modifications to classes are needed:

• I think the predicate pick related_to is the sticking point. gene to disease or phenotypic feature association has the predicate "affects" or descendants of, which is more specific.
• And I don't know if child Associations inherit the predicate info of the parent. So I don't know if making/modifying an child class will work.

FYI:

• This issue comes from a Translator Slack discussion (internal link). I was told to document when I use qualifiers in a way that I can't find in the Association classes, and then try to modify/add the Association classes to model this qualifier use.
• extra details on modeling/mapping, FYI only!
  • subject_form_or_variant value is set using the "molecular mechanism" column value, more info in Update gene variant qualifiers #1575.
  • I think it's okay to assume a variant is involved because every row/association has an allelic_requirement column value, and those terms are for the gene's mutations that possibly cause the disease - see Add allelic requirement property #1576 for details.)