hubmap requirements for bilateralogs and serialogs

[cmungall]

See also #2035

To best serve hubmap we should collect prospectively what kinds of terms are likely to be needed in short/medium/long term frames

I don't anticipate adding new terms that already conform to existing uberon design patterns and general philosophy (much of which I appreciate needs better documentation). the challenge is when new kinds of terms are requested.

For example, we have avoid "symmetric bilaterelogs" and "non-differing serialogs", with some exceptions.

These could be added, but they carry a cost. E.g. adding "LEFT fingernail of pinky" as a subclass of "fingernail of pinky" is easy. This would inherit being part of the pinky which is part of the hand which is part of the forelimb. However, the structure is confusing as we might expect to see part of the LEFT pinky which is part of the LEFT hand which is part of the LEFT forelimb. Then what do we do with the phalanges? And the developmental precursors of the phalanges? And the connectivity relations between these?

Similarly we have serialogs for spinal cord segments. These are already a bit unsatisfying as with the exception of C1-7 there is no relationship between subclasses of these classes, e.g. T10 in one species may be a homolog of a region overlapping T7-8 in another.

Once we start adding individual vessel serialogs that connect to each segment (many of which have bilateralogs) we end up with a large lattice of mostly repeating structures.

This policy has served many uberon use cases thus far - e.g. gene expression. But we should evolved to meet new use cases. The challenge is balancing this with requirements and not accuring technical debt. Just adding parts of the lattice above ad-hoc could make things much harder to do things systematically later.

My first question would be: can hubmap/ASCT make do with a compositional system, where existing slots in their datamodel which are filled with atomic uberon terms and replaced by OWL class expressions or a simple data structure, e.g. a tuple of laterality (L/R), repeated unit (e.g. spinal cord segment), and element within that unit (e.g. specific vein)?

There would have to be some logic for the join points where the repeated structures join with non-repeated (e.g. spinal nerve)

If this is not possible then we need to accelarate some of the dosdp work