Open cmungall opened 5 months ago
I am not sure what the action is for this issue
If I understand correctly, it could be that at least some of the type 1 expansions are correct?
Given W = P1 + P2 + ... + Pn
, if (a) there is no structure that is part of W
while overlapping several of the Pn
components (which, I agree, is clearly not the case of the “CN = PNS + CNS” case), and (b) assuming the sum is “complete” (it lists all the components that make up W
, without missing any), then the following equivalence is still correct:
(part-of some W) EquivalentTo (part-of some P1) or (part-of some P2) or ... (part-of some Pn)
So maybe we should review every sum in disjoint_union_over
(there’s something like two dozens of them), see if they meet the two criteria (a) and (b) above, and for those that do not (including those for which we are not sure that they meet the criteria, e.g. if if we are not sure that there can’t be a structure overlapping several parts), we remove the type 1 expansion.
Here are the “type 1” axioms in disjoint_union_over
.
part of some 'dermatological-mucosal system' EquivalentTo (part of some mucosa) or (part of some submucosa) or (part of some hypodermis) or (part of some skin of body) ==> Dubious? I would assume that there must be some structures that can span over both the hypodermis and the skin.
part of some 'nervous system' EquivalentTo (part of some 'peripheral nervous system') or (part of some 'central nervous system') ==> Clearly wrong as noted in the initial ticket.
part of some 'life cycle' EquivalentTo (part of some 'embryo stage') or (part of some 'post-embryonic stage') ==> Should be fine? No developmental stage can overlap both the embryonic stage and the post-embryonic stage.
part of some 'pleural sac' EquivalentTo (part of some pleura) or (part of some 'pleural cavity') ==> Should be fine?
part of some pleura EquivalentTo (part of some 'parietal pleura') or (part of some 'visceral pleura') ==> Should be fine?
part of some 'peritoneal sac' EquivalentTo (part of some 'peritoneal cavity') or (part of some peritoneum) ==> Should be fine?
part of some 'pilosebaceous unit' EquivalentTo (part of some 'sebaceous gland') or (part of some 'arrector muscle of hair') or (part of some 'hair follicle') ==> Unsure whether the sum is complete.
part of some 'hippocampal formation') EquivalentTo (part of some 'dentate gyrus of hippocampal formation') or (part of some 'Ammon’s horn') or (part of some subiculum) ==> Unsure whether the sum is complete.
part of some integument Equivalent to (part of some hypodermis) or (part of 'skin of body') ==> Dubious? Same as for 'dermatological-mucosal system' above, there could be structures spanning over more than one layer.
part of some 'hair shaft' EquivalentTo (part of some 'cortex of hair') or (part of some 'hair medulla') ==> Unsure.
part of some 'pair of lungs' EquivalentTo (part of some 'right lung') or (part of some 'left lung') ==> Should be fine. Though of dubious usefulness…
part of some 'serous pericardium' EquivalentTo (part of some 'parietal serous pericardium') or (part of some 'visceral serous pericardium') ==> Should be fine?
part of some 'abdominopelvic cavity' EquivalentTo (part of some 'abdominal cavity') or (part of some 'pelvic cavity') ==> Dubious. The “border“ between the abdominal cavity and the pelvic cavity is fuzzy, surely there can be overlapping structures.
part of some 'pair of dorsal aortae' EquivalentTo (part of some 'left dorsal aorta') or (part of some 'right dorsal aorta') ==> Should be fine?
part of some diaphragm EquivalentTo (part of some 'central tendon of diaphragm') or (part of some 'crus of diaphragm') or (part of some 'costal diaphragm') ==> Dubious? There may be vessels (blood and/or nerves) spanning over the entire diaphragm.
part of some 'feather shaft' EquivalentTo (part of some 'feather vane') or (part of some 'feather rachis') ==> No idea.
part of some 'orbital region' EquivalentTo (part of some 'eyeball of camera-type eye') or (part of some 'ocular adnexa') ==> Wrong according to the textual definition of 'orbital region', which says it is “the subdivision of the face that includes the eye (eyeball + adnexa) and the orbit of the skull plus associated parts of the face” (emphasis mine).
part of some 'vibrissal follicle-sinus complex' EquivalentTo (part of some 'vibrissa follicle') or (part of some 'blood sinus of vibrissa') ==> Wrong according to the textual definition of 'vibrissal follicle-sinus complex', which mentions that such a complex may sometimes also contain “concentric rings of epidermal tissue“ that are part of neither the follicle nor the blood sinus.
part of some 'respiratory primordium' EquivalentTo (part of some 'respiratory primordium epithelium') or (part of some 'respiratory primordium mesenchyme') ==> None of the terms are defined, so unsure.
We have some old documentation for part-disjointness: https://github.com/obophenotype/uberon/wiki/Part-disjointness-Design-Pattern
This needs translated to yaml
However, we have no documentation for the mereological sum DP
These are stored in the
disjoint_union_over.ofn
component:3180
The basic idea was to encode mereological sum axioms. These are conceptually very straightforward
W = P1 + P2 + ... + Pn
These were translated into two axioms:
(part-of some W) = (part-of some P1) or ... (part-of some Pn)
Disjoint( (part-of some P1), ..., (part-of some Pn))
And actually to be complete this needs expanded further to account for the local reflexivity of part-of in the context of this.
However, expansion
1
above is actually wrong; considerNS = CNS + PNS
If we expand to:
`(part-of some NS) = (part-of some CNS) or (part-of some PNS)
This is not right, because a structure that was within the NS as a whole but overlapped CNS and PNS
What we in fact need to write is
(grain and part-of some NS) = (grain and part-of some CNS) or (grain and part-of some PNS)
(expansion 2 is fine)
Where grain is something small enough that it doesn't span a boundary.
cell
is not safe since some cells span both parts (in particular any projecting neuron).atom
is safe. At this point the axiom ceases to be useful...I am not sure what the action is for this issue - it may be simplest if we don't try and represent msums at all and just focus on spatial disjointness. But I at least wanted to record some of the original thinking.