Unreferenced CHEBI term in 6606056e00002011 YAML

[kimrutherford]

I've been looking at the YAML files in tests/input while reviewing #165.

I noticed that this chemical is in objects but it doesn't seem to be referenced anywhere?:

- id: CHEBI:166824
  label: peptide antigen
  type: gocam:Object

The term is also unconnected in the YAML in the main branch test directory and in the YAML from the API:
https://live-go-cam.geneontology.io/product/yaml/go-cam/6606056e00002011.yaml

In Noctua it's connected:

• http://noctua.geneontology.org/workbench/noctua-visual-pathway-editor/?model_id=gomodel%3A6606056e00002011
• https://amigo.geneontology.org/amigo/model/6606056e00002011

[pkalita-lbl]

This is a result of not handling the is small molecule activator of (RO:0012005) relationships (for example, between peptide antigen (CHEBI:166824) and activity with MF transmembrane signaling receptor activity (GO:0004888) enabled by CD72 Hsap (UniProtKB:P21854). This will involve additional LinkML modeling to resolve.

Some initial thoughts on modeling

1. Follow the existing pattern of the has_input, has_output, has_primary_input, has_primary_output slots on Activity. Specifically: add a new has_small_molecule_activator slot with range MoleculeAssociation. An abridged example:

activities:
  - id: gomodel:0000/activity-1
    molecular_function: ...
    enabled_by: ...
    has_input:
      - molecule: gomodel:0000/chem-1  # this assumes changes from https://github.com/geneontology/gocam-py/pull/165
  - id: gomodel:0000/activity-2
    molecular_function: ...
    enabled_by: ...
    has_small_molecule_activator:
      - molecule: gomodel:0000/chem-2 

2. Make a bigger change and model molecular associations more like the Activity.causal_associations slot. Specifically: add a predicate slot to MoleculeAssociation; remove the has_input, has_output, has_primary_input, and has_primary_output from Activity; add a molecular_associations slot to Activity. An abridged example:

activities:
  - id: gomodel:0000/activity-1
    molecular_function: ...
    enabled_by: ...
    molecular_associations:
      - predicate: RO:0002233  # has input
        molecule: gomodel:0000/chem-1 
  - id: gomodel:0000/activity-2
    molecular_function: ...
    enabled_by: ...
    molecular_associations:
      - predicate: RO:0012001  # has small molecule activator
        molecule: gomodel:0000/chem-2 

@cmungall do you like one of these options or something else?

[pgaudet]

CHEBI:166824 should not be used in models.

I dont know how we can blacklist some CHEBIs - would that be possible?

[pkalita-lbl]

@pgaudet the LinkML model and code in gocam-py could help identify or even filter out such models during pipeline processing, but it can't do anything to prevent certain CHEBI terms from being used in the first place. That would be up to curation tools to do.

[kimrutherford]

CHEBI:166824 should not be used in models.

Here's an example from a different model that has an unreferenced, but valid CHEBI term:

• https://www.pombase.org/gocam/pombase-view/docs/66187e4700003150
• http://noctua.geneontology.org/workbench/noctua-visual-pathway-editor/?model_id=gomodel%3A66187e4700003150

This term is mentioned in the YAML but isn't referenced by an activity:

- id: CHEBI:17234
  label: glucose

[cmungall]

We can actually constrain CHEBI IDs in a number of ways

• by branch
• by inclusion or exclusion list (e.g. chemessence, non-ph73-protonated-forms)

But I'd separate that discussion from the data modeling until there is a clear proposal and just treat this as curation best practices for now

[cmungall]

This term is mentioned in the YAML but isn't referenced by an activity:

• id: CHEBI:17234
  label: glucose

as an aside, this ID should DEFINITELY not be used. No uncommitted stereoisoforms!

[pkalita-lbl]

@cmungall do you have a preference on the modeling options outlined in #167 (comment)?

[kimrutherford]

as an aside, this ID should DEFINITELY not be used. No uncommitted stereoisoforms!

That's above my pay grade so I'll ask @ValWood to reply when she's back from holiday.

[ValWood]

I can see how that happened, I would have done the same.
Usually we follow Rhea, but this is a glucose receptor so it has no Rhea.
@PCarme can you change to CHEBI:4167 (D-glucose)

[PCarme]

I can see how that happened, I would have done the same. Usually we follow Rhea, but this is a glucose receptor so it has no Rhea. @PCarme can you change to CHEBI:4167 (D-glucose)

That's done, I have changed it to D-glucose (CHEBI:17634)