Some entities are showing up in both proteins and chemicals

[gaurav]

In some cases we have incomplete cliques that are chemicals and proteins:

• Mesh Proteins as chemicals #200
• UMLS:C0004609 "baclofen" not cliquing with PUBCHEM.COMPOUND:2284 "Baclofen" #232
• BDNF has no gene or protein interactions NCATSTranslator/Feedback#150
• multiple nodes for Glatiramer NCATSTranslator/Feedback#576
• Insulin, receptor from MESH:D011972 not correctly normalized NCATSTranslator/Feedback#613
• Multiple entries for same result (Baclofen may treat Cerebral Palsy) NCATSTranslator/Feedback#623
• Harmonizing Ingests to support UI Collapsing of Edges and Presentation of Consolidated Evidence NCATSTranslator/Feedback#743 (comment)

Note that this sometimes results in the same identifier showing up twice -- once as a chemical and once as a protein:

• Filgrastim is still being assigned the wrong preferred identifier #215
• An ID can be present in multiple cliques (CURIE duplication) #276
• Some UMLS-MeSH connections are broken in Babel 2024jul13 #308
• NN failures: JAK genes NCATSTranslator/Feedback#821
• There are two different results for Bevacizumab within this query. They have different paths and evidence. NCATSTranslator/Feedback#970

[gaurav]

Given the structure of Babel and NodeNorm, the easiest way to fix this would be by creating a hidden conflation that combines chemicals and proteins that we understand to mean the same thing into a hybrid Protein-Chemical clique. So that reduces this problem to:

• Where can we store this hidden conflation in our databases (we could create a new database, but that would be annoying)
• How do we identify that a particular CHEBI and a particular Protein refer to the same thing?