concentration reduction in relationship to disease

[mmayers12]

This issue is in reference to DrugMechDB conversion to biolink, see issue #314

One relationship in DrugMechDB is Chemcial REDUCED_IN Disease. This relationship implies that a reduced concentration of a chemical is somehow associated with a disease. I've considered two slots for this edge prevents and causes, however, both of these imply some kind of causal relationship between the disease and the chemical.

Here are a few examples:

• Estrogen REDUCED_IN postmenopausal atrophic vaginitis
• Folic Acid REDUCED_IN Megaloblastic anemia (Methotrexate Treatment)
• Calcium Ion REDUCED_IN hypocalcemia

[mbrush]

FWIW, another way to think of these is as phenotypes associated with the disease. The HPO already phenotype terms representing altered levels of a specific chemical entity (e.g. Hypocalcemia - 'an abnormally decreased calcium concentration in the blood'), such that the biolnk model could represent something like "Calcium Ion REDUCED_IN hypocalcemia" as a DiseaseToPhenotypicFeatureAssociation where 'has phenotype' is the relation and the HPO term is in the the object slot.

The issue here of course is that every chemical would need to have a phenotype defining its abnormal level in the HPO. And at present there are probably only on the order of tens to hundreds of such terms in the HPO, added on an as-needed basis, out of the of tens or hundreds of thousands that would be needed/possible. @cmungall or @balhoff may be able to comment about how creation of such phenotype classes may be automated, and/or how a more flexible and expressive GO-CAM-like modeling approach might be useful to consider for these types of granular/molecular phenotypes.

[nlharris]

@balhoff any comments on this?