search

patrick-lloyd-ray / Allen-Institute-Ontology

Public-facing repository for versions of The Allen Institute Application Ontology.
0 stars 0 forks source link

Feature requests, iteration of specimen-level modeling #1

Open jimmymathews opened 3 years ago

jimmymathews commented 3 years ago

Feature requests for AIO. I was working from your Oct 5 release.

  1. Shall we just use "genotype information" directly, rather than "donor genotype information"? Similarly, shall we just use "allele information" directly, rather than "donor allele information"? I don't see what we gain by having this extra class. We can just make sure that the donor to genotype information relationship is "is about". Similarly: "disease of human donor". Is this on purpose (question for group)? Do we only want the human donors to be able to have reported disease states? There is already "disease" class. Could just make our instance data include a table of pertinent diseases.

  2. There doesn't seem to be an object property for "colocation", meaning that locations overlap. There is "located in" but that is too strict. Any suggestion here? We want to record cells and things as being potentially only partially in some regions.

  3. Can we make an "anatomical atlas" class, and a subclass of "material anatomical entity", something like "registered anatomical region", which is "part of" some "anatomical atlas"?

  4. There was already a class called "specimens derived from shared ancestor", which is a subclass of "collection of specimens". I think it does the work we need it to in order to encode a hierarchy of specimen derivation.

  5. In the course of my dataset-by-dataset review, I'll note whether it is sensible to say that the specimens were collected by means of a process belonging to one of the classes: 'collecting specimen from organism' biopsy lavage 'collecting specimen with swab' 'environmental material collection process' 'histological sample preparation' 'material sampling process' 'specimen set collection process'

Most likely there will be desired differentia, which should be enough to define further subclasses pertinent to our dataset set. The creation of these classes does not block the data model, since they only need to be named in the data values themselves. I'll need to make sure that enough information is given to recapitulate whatever is lost by reducing the items listed below in (6) to the material entity part only.

  1. I think we need to work together to replace these classes: 'brain specimen study design' 'brain slice study design' 'cell body study design' 'cell in slice study design' 'cell nucleus study design' 'tissue sample study design' 'whole brain study design' 'whole cell study design' with material entities... plus what is described above in (5). I guess "brain", "cell", "tissue sample"... these are already in there.

  2. Is there no proper biological species class? The "species" class seems to be the general philosophy notion. We need something controlled by NCBI taxonomy I think.

patrick-lloyd-ray commented 3 years ago

  1. Yes, I agree. The use of those classes was just to meet the need for explicit representation of those items as a desiderata in the program goals.

  2. I think 'overlaps' (x overlaps y if and only if there exists some z such that x has part z and z part of y) will work for this case of spatial colocation. If that doesn't work, we might use 'partially overlaps' (x partially overlaps y iff there exists some z such that z is part of x and z is part of y, and it is also the case that neither x is part of y or y is part of x). I might be missing something here though with the use case. We should talk about it further as it might involve creating a new mereotopological relation.

  3. Yes, we can do that. We'll probably have to maintain many non-interoperable anatomical classification schemas.

  4. Yes, I think so too.

  5. I can add more classes too, so this will be an iterative process, I'm sure.

  6. Yes, those are going to have to be changed at minimum to meet our requirements. I think the big problem is that there are no corresponding processes with those material participants now in the ontology. That's the top priority for the ontology development work w/r/t the data model at this point.

  7. NCBI doesn't have 'species' -- 'species' is a rank in that ontology where 'rank' just means something like 'grainedness'. The definition was intended to capture that usage so that we could add NCBI classes into our ontology as object aggregates. I don't know if this is the correct solution in the long-term, but we can discuss it.

jimmymathews commented 3 years ago

As for (7), I'm talking about this: https://www.ncbi.nlm.nih.gov/Taxonomy/Browser/wwwtax.cgi?mode=Info&id=10090&lvl=3&lin=f&keep=1&srchmode=1&unlock e.g. NCBI:txid10090

patrick-lloyd-ray commented 3 years ago

As for (3), I've decided to create 'anatomical atlas part' as a subclass of 'fiat object part'.

Def.= A fiat object part that is a fiat anatomical region demarcated from the rest of the organism by a two-dimensional continuant fiat boundary.

We can add subclasses here and even group them (we might need an intermediate class at some point, if we decide on some grouping), but the basic idea is that brain regions are fiat anatomical regions but we need to have those particular brain region schemas grouped together somehow under a single class. [Since anatomical atlas parts or regions may overlap, we'll probably add them as instances to avoid disjointness problems for now.] At least for now these atlas regions have a "landing spot."

Non-fiat regions and anatomical entities are already handled without problem.

patrick-lloyd-ray commented 3 years ago

As for (6), I've added and annotated the corresponding '[x] study design execution' classes, which are the corresponding processes that realize those study designs. Material entities can be linked to those classes as participants in those processes.