Decide on when to use 'located_in' vs 'part of'

[dosumis]

There are a number of cases where it may be desirable to not use 'part of' (BFO_0000050) to relate cell types to tissues/anatomical structures. Located in (or contained in?) may be more appropriate. These need to be discussed in the context of Uberon and query use cases Examples:

• Relationship of non-resident immune cells to a tissue
• Relationship of blood supply to a tissue: especially in the case of the brain
• Relationship of nervous system innovation to a tissue (afferent vs efferent)
• Relationship of tracheal system cell to tracheated(?) tissue (in arthropods)

This ticket is intended as a place to collect concrete examples & use cases for further discussion among CL/Uberon editors.

[srobb1]

Here is how we use 'part of' and 'contained in' in the PLANA ontology (planarian, flatworm).

When an entity is within some type of region, like the head region of the planarian we used 'contained in'. The head region doesn't have a super hard stop and start location, so it is a region. And all the structures are are within this region we use 'contained in'.

If a structure has definitive edges, like the reproductive system, we annotate all entities within the reproductive system with 'part of'

[cmungall]

Hi @srobb1! I don't think contained-in offers any advantage over part-of w.r.t when it comes to imprecise boundaries. I would recommend on using part-of throughout for PLANA. Are there any disadvantages for this, e.g gene expression queries?

[cmungall]

For CL my proposal is to use part-of universally. Having fewer axes of freedom means more consistency, easier for people to contribute.

The disadvantage is that we may have to construe of certain kinds of structures as being more inclusive than we had previously throught. E.g.:

• a brain is not just a mass of neurons, glial cells, and their parts; it also includes blood vessels and associated cells
  • if people really want to refer to the neural tissue subset we can make terms like 'neural tissue component of brain'
• many tissues will include immune cells like macrophages

Would this cause problems e.g for gene expression queries? Perhaps people querying for genes expressed in the brain would get more noise / reduced signal due to inclusion of immune cells, endothelial cells, ...? But isn't this the case anyway? Maybe we should be moving towards more precise cell-centered queries for gene expression anyway? cc @fbastian

Whatever we decide, the onus is on Uberon to be crystal-clear about what the makeup of any organ/tissue is; e.g. is the brain neural tissue with "holes" which blood vessels occupy, or does it include them?

Note Uberon does have a lot of axioms like (part-of some vasculature) DisjointWith (part-of some lens of camera type eye)... although they seem to be missing from current release? These are useful in making assumptions explicit.

[dosumis]

Side note - for neurons we mostly avoid part_of in favour of subproperties of overlap - with has_soma_location being used to record the region to which a neuron 'belongs'. This is orthogonal to discussion here.