ZFA/Uberon issues: incompatible modelling

[gouttegd]

Many of the unsatisfiable classes found upon merging Uberon with ZFA are not due to a “local” error in one ontology or the other (e.g. one ontology wrongly asserting that an anatomical entity is part of something) or a “mapping” error (two terms being wrongly mapped even though they represent different things), but are rather the results of pretty fundamental differences between Uberon’s and ZFA’s modelling decisions. As such, they may not have an “easy” fix.

In many cases, the most fundamental root cause is the fact that ZFA tends to mostly use part of relations, whereas Uberon uses a greater variety of relations, including relations that do not imply parthood and therefore cannot violate spatial disjunctions (for example, relations such as overlaps, 'conduit for', 'contributes to the morphology of', 'has soma location', 'extends fibers into', etc.).

Spatial disjointness of brain regions

In Uberon, most brain regions (hindbrain, midbrain, etc.) are considered to be spatially disjoint. But many terms in ZFA are explicitly said to be simultaneously part of several regions.

somatomotor neuron

Directly part of both the hindbrain and the midbrain. The corresponding CL term is part of neither.

rostral mesencephalo-cerebellar tract and caudal mesencephalo-cerebellar tract

Directly part of the brainstem and indirectly (through 'rostral cerebellar tract') part of the cerebellum.

medial octavolateralis nucleus

Directly part of the cerebellum and indirectly (through 'lateral line sensory nucleus') part of the brainstem.

lateral column

Indirectly (through 'reticular formation') part of the brain and indirectly (through its mapping with Uberon’s 'spinal cord lateral column') part of the 'spinal cord'.

Spatial disjointness of central and peripheral nervous systems

This issue notably affects all cranial nerves, which in ZFA are simultaneously part of both the CNS and the PNS. In Uberon, we circumvent the problem by distinguishing between the cranial nerve proper (which is nerve and therefore part of the PNS) and the root of cranial nerve, which is part of the CNS – the two being linked by 'extends fibers into', which has no implication on parthood. But while ZFA also has a distinct cranial nerve root term, it is directly part of some cranial nerve, which negates the benefit of having two distinct terms.

Other root unsats caused by the PNS/CNS disjointness include diencephalic efferent neurons to the lateral line and rhombencephalic efferent neurons to the lateral line, which are simultaneously part of both systems. In CL they would probably be modelled as being part of the PNS and having their soma located in the brain.

Sensory system being part of the nervous system

In Uberon, (almost?) all anatomical systems are spatially disjoint with each other. Notably, nothing can be simultaneously part of the nervous system and part of the skeletal system).

However, sensory system in Uberon is not part of the nervous system, but instead merely overlaps it. This allows, for example, a term like bony labyrinth to be both part of the [skeletal system](skeletal system) (since it is part of a bone) and part of the sensory system (since it is part of the ear) without causing any disjointness violation.

But in ZFA, sensory system is part of the nervous system. As a result, several parts of sensory organs in ZFA end up being part of both the nervous and the skeletal systems, making them unsatisfiable:

• lagena (part of a bone, so skeletal system, and part of the ear, so sensory and therefore nervous system);
• annular ligament (ligament, so part of the skeletal system, and part of the eye, so sensory and therefore nervous system).

Appendage vs fin vs trunk

Uberon’s notion of appendage seems to have been designed with typical “limbs” (e.g. legs, arms) in mind. An appendage is a “major subvidision of organism that protrudes from the body”, and an organism subvidision is a “subdivision of a whole organism, consisting of components of multiple anatomical systems, largely surrounded by a contiguous region of integument”.

It’s unclear whether fins (which are currently considered as appendages) really fit within that framework. Appendages are disjoint from the trunk, which causes multiples issues in ZFA because fins (at least some of them) are located on the trunk.

The disjointness axiom between trunk and appendage has an explicit exception for ZFA, indicating that the problem is well known (in fact most of the problems between ZFA and Uberon have been known for more than 10 years...), but presumably this exception is informative only. I don’t know if the intention was that the exception should be taken into account when the mutually spatially disjoint with macro is expanded, but if it were, it has never been done: as expanded, the disjointness represented by that macro applies to all terms regardless of their provenance (ZFA terms are not excluded from the disjunction)

Material vs immaterial entities

The last category of issues is about how Uberon and ZFA represent “voids” or “spaces” (immaterial entities). There are several disagreements here. Most often, Uberon tends to represent spaces primarily by a material entity, because we want to represent the (material) structures that, by their particular disposition, create the anatomical space (if needed, we then have another term to represent the actual space created inside the structure). ZFA on the other hands, tends to represents spaces directly by immaterial entities.

For example, what Uberon considers a material anatomical junction is often represented in ZFA as either an immaterial anatomical line (e.g. corneo-scleral junction, zona limitans intrathalamica, telecenphalon diencephalon boundary) or an immaterial anatomical space (e.g. intestinal opening). Likewise, what Uberon considers a material fibrous joint is represented as an immaterial anatomical space in ZFA (e.g. fontanel). What Uberon considers a material anatomical conduit is most often represented as an immaterial anatomical space in ZFA (this notably impacts all ZFA foramina).

A notable exception is for “grooves”, where it is the other way around: groove is a (material) anatomical structure in ZFA (“the anatomical structure that delineates a furrow or depression in an organ or tissue”) whereas the corresponding surface groove is an immaterial anatomical entity in Uberon.

Slightly related to this issue:

aqueous humor

part of the anterior chamber eye, which is an immaterial anatomical space.

But mapped to aqueous humor of eyeball, which is ultimately an organism substance, which is material. So we end up with a material entity being part of an immaterial entity, which is not possible.

In ZFA, this works because aqueous humor is simply an anatomical entity with no assertion as to its “materialness” (it is neither material nor immaterial). But the mapping with its corresponding Uberon term makes it a material entity.

The simple fix here would be for ZFA to say that aqueous humor is located in, rather than part of, the anterior chamber eye. This is what we typically do in Uberon.

perilymph

Similar problem. In ZFA, this is an anatomical entity (neither material nor immaterial) that is part of the perilymphatic space, which is immaterial.

But the mapping with Uberon’s perilymph makes it a material organism substance, which cannot be part of an immaterial space.