Add relations for process boundaries

[cmungall]

In the RO temporal text definitions we have α(x) and ω(x) denoting start and end points.

Note we can unroll expressions like α(x) <= ω(y) to property chains like starts-at o before-or-equal o inv(ends-at), to derive property chain axioms that can be partially expressed in OWL.

See: http://biorxiv.org/content/early/2014/06/27/006650

Rough draft (incomplete)

ends-at-point o inv(starts-at-point) -> immediately-precedes
ends-at-point o precedes o inv(starts-at-point) -> precedes

Requested for https://github.com/obophenotype/uberon/issues/1172

[Public-Health-Bioinformatics]

Hi, Has more work been done on this? I wanted to follow up, esp. with regard to how >= and <= translate to english. For GenEpiO I was looking for temporal boundary relations, so threw in:

necessarily_follows
necessarily_is_or_follows
necessarily_is_or_precedes
necessarily_precedes

as relations. Seems like this is what you are talking about above? (I'd say that these relations aren't restricted to temporal dimension; they can map events, objects etc. to ordinal positions ore number line too). I was trying to anticipate how to describe a life cycle, briefly,

date of conception
date of birth
date of death

date of birth necessarily follows date of conception
date of death necessarily is or follows date of birth (to allow for stillbirth etc.)

One tricky thing is that the granularity of measurement or display entails that at one scale (days say) events look like singular points in time, whereas at other scales (minutes say) events have duration. One world normally want to claim a "follows" or > relation, e.g. stillbirth must always be proceeded by birth; but in reality e need to use a >= "is or follows" relation occurs when only dates are recorded. A reasoner would likely have trouble dealing with a data granularity proviso though!

[cmungall]

The relations you need are likely in RO already:

• RO:0002222 ! temporally related to
  • BFO:0000063 ! precedes
    • RO:0002090 ! immediately precedes
  • RO:0002086 ! ends after
    • BFO:0000062 ! preceded by
      • RO:0002087 ! immediately preceded by

The modal qualifier 'necessarily' is not clear - it may be built into OWL depending on what sense you mean.

Definition of α(x) and ω(x) are required for formal completeness (to see where I'm going with this, see the sequence analog here - http://biorxiv.org/content/early/2014/06/27/006650), but for most purposes existing relations should be fine.

Relations that link an occurrent to some kind of timepoint is a bit more difficult and depends on the use case - e.g. do you want to use xsd dates?

More on this later. In the interim you may be interested in the temporal modeling in https://github.com/obophenotype/developmental-stage-ontologies -- this ticket may be of interest: https://github.com/obophenotype/developmental-stage-ontologies/issues/19

[Public-Health-Bioinformatics]

Yes I was going to use xsd dates for conception/birth/death. And indeed "necessarily" isn't actually adding anything ('twas a floppy attempt to constrain states of a system, not just their order). So with RO:

"date of death" preceded by "date of birth" OR "date of death" is "date of birth" 

Which is the substance of a "is or precedes" relation.

Thanks for your details there, I will examine closely in the next week.

[nlharris]

What is the status of this?

[nlharris]

Can this be closed?

[nlharris]

Closing; reopen if needed.