DRUGBANK:category predicate remap should be fixed?

[saramsey]

I think maybe the predicate remapping entry for DRUGBANK:category (which is currently mapping to biolink:related_to) is not sufficiently precise. I think biolink:subclass_of would be a better fit. A couple of examples should be checked to corroborate this and then we can fix the entry in predicate-remap.yaml.

https://github.com/RTXteam/RTX/blob/6fa597807f3f91ce294b7b4660f563950d201f54/code/kg2/predicate-remap.yaml#L569

[ecwood]

Here's some examples: (the images are from the DrugBank browser)

DRUGBANK:DB00001 (Lepirudin):

• DRUGBANK:DB00001--DRUGBANK:category->MESH:D000602 (Amino Acids, Peptides, and Proteins)
• DRUGBANK:DB00001--DRUGBANK:category->MESH:D000925 (Anticoagulants)
• DRUGBANK:DB00001--DRUGBANK:category->MESH:D058833 (Antithrombin Proteins)
• DRUGBANK:DB00001--DRUGBANK:category->MESH:D000991 (Antithrombins)
• DRUGBANK:DB00001--DRUGBANK:category->MESH:D002317 (Cardiovascular Agents)
• DRUGBANK:DB00001--DRUGBANK:category->MESH:D004791 (Enzyme Inhibitors)
• DRUGBANK:DB00001--DRUGBANK:category->MESH:D050299 (Fibrin Modulating Agents)
• DRUGBANK:DB00001--DRUGBANK:category->MESH:D006401 (Hematologic Agents)
• DRUGBANK:DB00001--DRUGBANK:category->MESH:D010455 (Peptides)
• DRUGBANK:DB00001--DRUGBANK:category->MESH:D011480 (Protease Inhibitors)
• DRUGBANK:DB00001--DRUGBANK:category->MESH:D011506 (Proteins)
• DRUGBANK:DB00001--DRUGBANK:category->MESH:D015842 (Serine Protease Inhibitors)
• DRUGBANK:DB00001--DRUGBANK:category->MESH:D015843 (Serpins)

DRUGBANK:DB00004 (Denileukin diftitox):

• DRUGBANK:DB00004--DRUGBANK:category->MESH:D036002 (ADP Ribose Transferases)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D000602 (Amino Acids, Peptides, and Proteins)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D036002 (ADP Ribose Transferases)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D000970 (Antineoplastic Agents)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D001427 (Bacterial Toxins)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D001685 (Biological Factors)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D016207 (Cytokines)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D004798 (Enzymes)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D045762 (Enzymes and Coenzymes)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D016695 (Glycosyltransferases)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D036341 (Intercellular Signaling Peptides and Proteins)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D007378 (Interleukins)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D008222 (Lymphokines)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D010430 (Pentosyltransferases)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D010455 (Peptides)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D011506 (Proteins)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D011994 (Recombinant Proteins)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D014118 (Toxins, Biological)
• DRUGBANK:DB00004--DRUGBANK:category->MESH:D014166 (Transferases)

[saramsey]

Thank you, @ericawood. What do you think of using biolink:subclass_of? It's a bit odd, since normally we use subclass_of within an ontology and not cross-ontology. But on the other hand, it would provide more information than related_to, right?

[ecwood]

Thank you, @ericawood. What do you think of using biolink:subclass_of? It's a bit odd, since normally we use subclass_of within an ontology and not cross-ontology. But on the other hand, it would provide more information than related_to, right?

I agree that it seems a bit odd. However, given the wide variety of meaning of related_to, I think that more specific would be helpful in this instance to give actual meaning to the edges.

[saramsey]

Maybe let's go for it? We can always change our mind in the future, if something better than subclass_of becomes available in Biolink.

[kvarforl]

looks fixed in kg2.6.0 :)