Multi-hop query for AKU - Homogentisic Acid

[andrewsu]

Issue #48 asks for ChemicalSubstances that cause AKU (MONDO:0008753). And while several ARAs get the right answer of Homogentisic acid, it only seems to come through SemMedDB. But as we know, the needles found in SemMedDB can often come in very big haystacks. For context, there are over 10,000 ChemicalSubstance - causes - DiseaseOrPhenotypicFeature edges in SemMedDB 1, including gems like Sodium Chloride - causes - Periphlebitis (C0031129) and Toxin - causes - Vitreous degeneration (C0155366). SemMedDB also reports that are 15 other DiseaseOrPhenotypicFeatures that are caused_by Homogentisic acid 2.

@vdancik raised the possibility whether we should also look at multi-hop queries to find the link between Homogentisic acid and AKU (MONDO:0008753), especially since no one was aware of resources that manually curate ChemicalSubstance - causes - DiseaseOrPhenotypicFeature edges. This issue is meant to track any efforts to design and execute such a multi-hop query.

[andrewsu]

Given that AKU is an inborn error of metabolism, one could imagine a metapath like AKU - GeneOrGeneProduct - Pathway or Process - ChemicalSubstance.

The first edge AKU - GeneOrGeneProduct seems easy -- OMIM (https://omim.org/entry/203500) has the link between AKU and the gene for HGD homogentisate 1,2-dioxygenase.

The second and third edges for GeneOrGeneProduct - Pathway or Process - ChemicalSubstance seem to depend on having a Pathway/Process database that includes both genes and compounds. Suggestions on today's standup call included Rhea, HMDB, and Reactome.

• Rhea: https://www.rhea-db.org/rhea/15449 includes homogentisate CHEBI:16169 as a reactant, which in turn links to Q93099 HGD_HUMAN via https://www.uniprot.org/uniprot/?query=annotation:(type:%22catalytic%20activity%22%20rhea:15449)&fil=reviewed%3Ayes+AND+organism%3A%22Homo+sapiens+%28Human%29+%5B9606%5D%22
• HMDB: https://hmdb.ca/proteins/HMDBP00842/metabolite_protein_links shows a link between Homogentisic acid (HMDB0000130) and Homogentisate 1,2-dioxygenase (HMDBP00842) under "metabolite protein links" (so even bypassing that intermediate Pathway or Process node). Both those entries also list related pathways, and some are shared. Note also that HMDB does list Alkaptonuria under "Associated Disorders and Diseases", but it's unclear whether Alkaptonuria is referenced by a Disease identifier, and it does not include the more specific predicate of causes.
• Reactome: reactome calls Homogentisic acid as "HGTA" https://reactome.org/content/detail/R-ALL-30337. It links to a reaction https://reactome.org/PathwayBrowser/#/R-HSA-8963691&SEL=R-HSA-71164&PATH=R-HSA-1430728,R-HSA-71291&FLG=16169, which in turn links to "HGD hexamer" which links to UniProt:Q93099 HGD

So there do exist data paths to get from AKU to Homogentisic acid via curated links. Now comes the question of how to operationalize those queries in the context of Translator, and whether the specificity of that result is better than via SemMedDB...

[andrewsu]

Also relevant is @brettasmi's post using SPOKE at #48 (comment)

[sierra-moxon]

Discussion is also ongoing in the biolink-model repo: biolink/biolink-model#478 (note comments about MolecularActivity). We plan to do a bit of a write up and be sure to discuss further with interested folks.