search

obophenotype / bio-attribute-ontology

source files for OBA (Ontology of Biological Attributes)
https://obophenotype.github.io/bio-attribute-ontology
Creative Commons Zero v1.0 Universal
27 stars 11 forks source link

Add ratio related patterns and examples #312

Closed rays22 closed 3 months ago

rays22 commented 8 months ago
  1. The purpose of pattern template compound_attributes is to extend PATO with compound qualities, i.e. qualities that are defined by using primitive (I mean existing PATO) qualities. The two-component quality is then used in the entity_attribute pattern to define a new term, e.g. digit length ratio.
  2. When there is a ratio-type quality between two other unrelated qualities, like the proportionality of mass and length in BMI, then I propose the use of proportion_of_attributes template to define such OBA attributes. Fix #272 .
cmungall commented 8 months ago

It would be good to align TO and OBA here. Ratios are very common for plant traits. cc @marieALaporte @cooperl09

marieALaporte commented 8 months ago

We can certainly align the pattern that we have in TO and what you are proposing here. Are the patterns for ratio validated yet? When it is, I can update what we have in TO. Might be good at some point to review the rest of the patterns that we are using in TO to align with standard patterns

rays22 commented 8 months ago

Are the patterns for ratio validated yet

Yes, they are formally valid.

rays22 commented 8 months ago

I have found two TO examples of ratio-type traits that also have logical definitions:

The two classes above appear to defined by the following EQ pattern:

(proportionality to
    and (TO_has_dividend_quality some TRAIT-1)
    and (TO_has_divisor_quality some TRAIT-2)

, where proportionality to, TRAIT-1 and TRAIT-2 are qualities.

The two relationships used in the above TO pattern are defined outside of the OBO Relations Ontology:

  1. TO:has_dividend_quality
  2. TO:has_divisor_quality

I am proposing the pattern

'proportionality_to'
    and ('has_part' some TRAIT-1)
    and ('has_part' some TRAIT-2)
    and ('characteristic_of_part_of' some ENTITY)

or more tersely:

'proportionality_to'
    and ('has_part' some (TRAIT-1 and TRAIT-2)
    and ('characteristic_of_part_of' some ENTITY)

In my suggested pattern I am using the standard RO has part relationship in place of TO:has_dividend_quality and TO:has_divisor_quality. Please note, that in my model the ratio quality is indifferent about the direction the calculation of a mathematical ratio. For example, the 'sodium to potassium' or 'potassium to sodium' ratios would be part of the same category.

cmungall commented 8 months ago

This is definitely wrong:

'proportionality_to'
    and ('has_part' some (TRAIT-1 and TRAIT-2)
    and ('characteristic_of_part_of' some ENTITY)

if this doesn't show up as unsatisfiable we are doing something wrong

This has cryptic incoherencies:

'proportionality_to'
    and ('has_part' some TRAIT-1)
    and ('has_part' some TRAIT-2)
    and ('characteristic_of_part_of' some ENTITY)

It loses directionality so {increased, decreased} x {A:B, B:A} become indistinguishable

rays22 commented 8 months ago

This is definitely wrong:

'proportionality_to'
    and ('has_part' some (TRAIT-1 and TRAIT-2)
    and ('characteristic_of_part_of' some ENTITY)

if this doesn't show up as unsatisfiable we are doing something wrong

This has cryptic incoherencies:

'proportionality_to'
    and ('has_part' some TRAIT-1)
    and ('has_part' some TRAIT-2)
    and ('characteristic_of_part_of' some ENTITY)

It loses directionality so {increased, decreased} x {A:B, B:A} become indistinguishable

@cmungall, thanks for your feedback. I gather from your comment that we need directionality in the trait ratio pattern to be interoperable with uPheno phenotypes. I am wondering if I should request new RO terms equivalent to

  1. TO:has_dividend_quality
  2. TO:has_divisor_quality and use an EQ based on the TO pattern above or do you have something else in mind that would work better?
cmungall commented 8 months ago

It’s less about interoperability and more about coherency. It’s really important the logical definition matches the label and the text definition. If my argument doesn’t make sense we should explore further, and it’s possible I’m wrong!

As to my preferred solution, yes I think we need to distinguish the roles of the different sub quantities and specific relations seem easiest

On Fri, Jan 26, 2024 at 5:35 AM Ray Stefancsik @.***> wrote:

This is definitely wrong:

'proportionality_to' and ('has_part' some (TRAIT-1 and TRAIT-2) and ('characteristic_of_part_of' some ENTITY)

if this doesn't show up as unsatisfiable we are doing something wrong

This has cryptic incoherencies:

'proportionality_to' and ('has_part' some TRAIT-1) and ('has_part' some TRAIT-2) and ('characteristic_of_part_of' some ENTITY)

It loses directionality so {increased, decreased} x {A:B, B:A} become indistinguishable

@cmungall https://github.com/cmungall, thanks for your feedback. I gather from your comment that we need directionality in the trait ratio pattern to be interoperable with uPheno phenotypes. I am wondering if I should request new RO terms equivalent to

  1. TO:has_dividend_quality https://www.ebi.ac.uk/ols4/ontologies/to/properties/http%253A%252F%252Fpurl.obolibrary.org%252Fobo%252FTO_has_dividend_quality?lang=en
  2. TO:has_divisor_quality https://www.ebi.ac.uk/ols4/ontologies/to/properties/http%253A%252F%252Fpurl.obolibrary.org%252Fobo%252FTO_has_divisor_quality?lang=en and use an EQ based on the TO pattern above or do you have something else in mind that would work better?

— Reply to this email directly, view it on GitHub https://github.com/obophenotype/bio-attribute-ontology/pull/312#issuecomment-1912079892, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAAMMOLP3HKUN6HKXP6FYSTYQOWLBAVCNFSM6AAAAABCCFIGB6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSMJSGA3TSOBZGI . You are receiving this because you were mentioned.Message ID: @.***>

rays22 commented 7 months ago

There are two term requests in RO that look relevant to the modelling of qualities emerging from two qualities that are in a ratio/proportion relationship:

rays22 commented 7 months ago

The resolution of this PR depends on https://github.com/oborel/obo-relations/pull/792 .

matentzn commented 6 months ago

We need this pattern now for another project, is there anything left to finalise it @rays22 ?

rays22 commented 6 months ago

We need this pattern now for another project, is there anything left to finalise it @rays22 ?

@matentzn , The resolution of this issue depends on a new RO release.

matentzn commented 6 months ago

@anitacaron can you make one soon?

anitacaron commented 6 months ago

yes, I just need to finish reviewing some PRs, and I'll make one.

matentzn commented 5 months ago

@rays22 can you bring the PR to conclusion soon? We need it for a major project coming up in OBA in the next weeks

anitacaron commented 5 months ago

I'm still waiting for another person to approve the RO release.

cmungall commented 5 months ago

approved

On Wed, May 1, 2024 at 5:44 AM Anita Caron @.***> wrote:

I'm still waiting for another person to approve the RO release.

— Reply to this email directly, view it on GitHub https://github.com/obophenotype/bio-attribute-ontology/pull/312#issuecomment-2088413153, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAAMMOKTNMEOOIJV5POEL7DZADPRVAVCNFSM6AAAAABCCFIGB6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDAOBYGQYTGMJVGM . You are receiving this because you were mentioned.Message ID: @.***>