Predicate issues around DisGeNET

[colleenXu]

Is your feature request related to a problem? Please describe. See the DisGeNET associations (OWL classes) hierarchy here and here for Gene -> Disease and Variant -> Disease relationships

In trying to "convert" or "map" these into relations (biolink or otherwise), various issues come up:

• Biomarker (SIO:001121) encompasses both "correlations" and "causation" (and affects to some extent). Most ontologies seem to separate these as top-level siblings, so a modeler has no choice but to go with the most general/upper-most relation.
• GeneticVariation (SIO:001122) is used in DisGeNET for both Variant-(correlates_with)-Disease and Gene-(has-variant-that-correlates-with)-Disease relationships.
  • The latter doesn't have a clear predicate/relation to map to (biolink or otherwise).
  • Note that in the TSV data dump for DisGeNET gene-disease associations, they do not include the variant for these kinds of relationships.
• It's not easy to find relations for the inverse relationships: Disease -> Gene and Disease -> Variant.

What working group (or team) did this request originate from? Service Provider / Exploring Agent

Describe the solution you'd like

• Have an "associated_with" parent for affects, correlates_with, causes, has biomarker / biomarker of. This is a child of related to. This could also help with the nebulous "predicted" relationships that are predicted using literature co-occurrence, similarity measures, graph module membership, etc.
  • This fuzzy maps with SIO:001403 "is_associated_with"
• Have a child of "correlated_with": "has_variant_correlated_with" and its inverse "is_correlated_with_variant_of". This could work for gene-disease shortcut relationships (where the situation is really that gene has gene variant/abnormal expression that is correlated with the disease), protein-disease shortcut relationships (where the situation is really that the protein has a protein variant/abnormal action that is correlated with the disease)
  • this maps to the GeneticVariation in SIO class above
• Have a child of "associated_with": "has_variant_associated_with" and its inverse "is_associated_with_variant_of".
  • this maps to the narrower NCIT:R176 "Disease Mapped to Gene". Description: "A role used to assert a direct relationship between a disease, disorder or finding and a gene. This restriction can be used when a polymorphism or an abnormality in a gene is either a clinical marker for, a causative event for, or predisposes a subject to a disease."
  • also maps to the narrower NCIT:R39 "Gene Is Biomarker of": Description: "A role used to assert that expression or alteration of a gene is correlated with a particular disease or disease state or is predictive of the disease or disease state"
• try to have symmetrical predicates or match each predicate with an inverse.

Additional information to support this request (optional)

• An interesting idea is from NCIT where they have separate relations for something that is definitely present with X VS something that is only sometimes present with X: compare "Disease has molecular abnormality" to "Disease may have molecular abnormality"