(regulation of) cell-cell adhesion - what relations to use to indicate cell types?

[BarbaraCzub]

Background

We've recently been facing the issue of not knowing what (annotation extension) relations to use between 'cell-cell adhesion' GO terms (and its descendants) and cell type CL ontology terms.

We've been discussing this issue at UCL (cc @RLovering @rachhuntley) and with the Orchard Protein Function Content team at EMBL-EBI (cc @pgarmiri) and below is a summary of our conclusions, suggestions and questions.

The relations that we've been discussing include 'occurs_in', 'has_agent' and 'regulates_o_has_agent' and whether they should be used with 'cell-cell adhesion' GO term, or 'regulation of cell-cell adhesion' ('regulates_o_has_agent' is relevant here).

Details

I prepared examples in Noctua and 'borrowed' examples from Protein2GO (which provide different selections of available relations), which reflect best what we've discussed. In each of the cases below the Noctua and Protein2GO examples are based on the same experiment from the same PMID.

It would be very helpful to understand what relations other teams use for the types of annotations shown below.

If you could please share your thoughts and comment on how you annotate similar adhesion experiments (especially wrt to 'Case 3' - furthest down) this will be very helpful.

Also, could you please tag in curators that you think would be interested to comment on these annotation issues and might otherwise miss this?

Thank you,
Barbara

Case 1. 'protein binding involved in cell-cell adhesion' GO annotations

Ideal case scenario
If gene X is transfected into T-cells, and gene Y is transfected into endothelial cells, and the encoded proteins are expressed on cell surface, and bind to each other, resulting in adhesion, then we would capture this in a GO annotation as follows:

Gene product - GO term - evidence code - the 'with' field - annotation extension (relation + CL term)
Gene X - protein binding involved in cell-cell adhesion - IPI - with Gene Y - occurs_in T-cell, has_agent T-cell, has_agent endothelial cell
and
Gene Y - protein binding involved in cell-cell adhesion - IPI - with Gene X - occurs_in endothelial cells, has_agent T-cell, has_agent endothelial cell

In this example 'occurs_in' refers to the cell, in which the gene product being annotated is expressed, whereas 'has_agent' identifies the cells involved in the protein binding and the adhesion process. The use of the relation 'has_agent' for adhesion terms was previously discussed and agreed: geneontology/annotation_extensions#61

Protein2GO example from PMID:10981966
The paper shows that Cd200 is expressed in thymocytes, whereas its receptor Cd200r1 is expressed in macrophages, and the two proteins bind to each other. The interaction between Cd200 and Cd200r1 is involved in thymocyte-macrophage adhesion.

The relations used in the annotation extension align exactly with the relations from the "ideal case scenario" above.

Noctua example from PMID:10981966
The paper shows that Cd200 is expressed in thymocytes, whereas its receptor Cd200r1 is expressed in macrophages, and the two proteins bind to each other. The interaction between Cd200 and Cd200r1 is involved in thymocyte-macrophage adhesion.

The relations used in the annotation extension align with the relations from the "ideal case scenario" above, except that has_participant is used instead of has_agent (has_agent is not available in Noctua).

Case 2. 'cell-cell adhesion' GO annotations

Ideal case scenario
If gene A transfected into T-cells is involved in T-cell adhesion to endothelial cells, we would capture this in a GO annotation as follows:

Gene product - GO term - evidence code - annotation extension (relation + CL term)
Gene A - heterotypic cell-cell adhesion - IDA/IPI - occurs_in T-cell, has_agent T-cell, has_agent endothelial cell

In this example 'occurs_in' refers to the cell, in which the gene product being annotated is expressed, whereas 'has_agent' identifies the cells involved in the adhesion process. The use of the relation 'has_agent' for adhesion terms was previously discussed and agreed: geneontology/annotation_extensions#61

Protein2GO example from PMID:10981966
The paper shows that Cd200 is expressed in thymocytes, whereas its receptor Cd200r1 is expressed in macrophages. The interaction between Cd200 and Cd200r1 is involved in thymocyte-macrophage adhesion.

The relations used in the annotation extension align exactly with the relations from the "ideal case scenario" above.

Noctua example from PMID:10981966
The paper shows that Cd200 is expressed in thymocytes, whereas its receptor Cd200r1 is expressed in macrophages. The interaction between Cd200 and Cd200r1 is involved in thymocyte-macrophage adhesion.

The relations used in the annotation extension align with the relations from the "ideal case scenario" above, except that has_participant is used instead of has_agent (has_agent is not available in Noctua).

Case 3. 'regulation of cell-cell adhesion' GO annotations

Ideal case scenario
If gene B is transfected into T-cells and this causes an increase in T-cell adhesion to endothelial cells, the resulting annotation will be:

Gene product - GO term - evidence code - annotation extension (relation + CL term)
Gene B - positive regulation of cell-cell adhesion - IDA - occurs_in T-cell, regulates_o_has_agent T-cell, regulates_o_has_agent endothelial cell

The use of regulates_o_has_agent is in agreement with the use of 'has_agent' for 'cell-cell adhesion' terms (as previously decided: geneontology/annotation_extensions#61). It is also in agreement with the chain relation rule: if you are describing the regulatory process, then you should use the conventional relations (i.e. occurs_in in the example above), whereas if you are describing the GO process or function that is the target of the regulation, then you should use the chain relation (i.e. regulates_o_has_agent in the example above).

However, atm the relation regulates_o_has_agent is not available for use in either Noctua or P2GO.

So what relations should be used? What relations do you use?

I previously raised a related issue here: geneontology/noctua-models#44

Protein2GO example from PMID:20110463
The paper shows that Homo sapiens (human) microRNA hsa-miR-221-3p expression in cholangiocytes decreases the adhesion between the cholangiocytes and T-cells.

In this example the relation used to describe the location of the regulatory process is in agreement with the "ideal case scenario", but the alternative relations were used instead of regulates_o_has_agent.

Noctua example from PMID:20110463
The paper shows that Homo sapiens (human) microRNA hsa-miR-221-3p expression in cholangiocytes decreases the adhesion between the cholangiocytes and T-cells.

In this example the relation used to describe the location of the regulatory process is in agreement with the "ideal case scenario", but the alternative relations were used instead of regulates_o_has_agent. In addition, the model does not "look correct" because two relations had to be added between the regulatory process GO term and the CL term. When I had added the 'occurs_in' relation between the molecular_function and the CL term, this had not been extrapolated to indicate the location of the regulatory process in the annotation preview.

[vanaukenk]

For reference: oborel/obo-relations#244
This ticket discusses the meaning, and use, of different types of participant relations and I think is relevant to our discussion here.

[BarbaraCzub]

Hi @vanaukenk is there a 'discussion' label I could use to 'tag' this ticket for discussion during an annotation call? Or perhaps this could be scheduled to be discussed during a specific call? (cc @RLovering @rachhuntley @pgarmiri).