components and grains

[GoogleCodeExporter]

We are lab-scale beasts and yet the molecular-scale nature of the things we 
inhale and imbibe have lab-scale effects on us.

To describe this sort of thing we need, if only to describe samples in 
experiments (and I am being chased for this), relations of the sort:

MUSTS:

has_granular_part
granular_part_of

SHOULDS (for mixtures):

has_component_part
component_part_of

I am happy if they go into the BFO 2 release, which is hopefully imminent, not 
fully worked out.

Best wishes,
Colin.

Original issue reported on code.google.com by batchelorc@rsc.org on 1 Jun 2012 at 9:39

[GoogleCodeExporter]

(this is sufficiently general that I believe it merits going into BFO rather 
than OBI)

Original comment by batchelorc@rsc.org on 1 Jun 2012 at 9:41

[GoogleCodeExporter]

Bizarrely, OBI has both 'has grain', with a textual definition and an undefined 
'has granular part' with no textual definition BUT a BFO ID.  Happy for BFO to 
go with the OBI version.

Will stop now.

Original comment by batchelorc@rsc.org on 1 Jun 2012 at 9:46

[GoogleCodeExporter]

HI Colin,
The current reference has:
Relation of membership
This relation is defined not only for objects but for BFO categories in 
general, as follows:

Elucidation: a member_part_of b at t =Def. there is a mutually exhaustive and 
pairwise disjoint partition of b into entities of category X: x1, …,xn with a 
= xi for some natural number i. [026-002]

Domain: entity in category X

Range: aggregate of X

Theorem: if a member_part_of b at t then a continuant_part_of b at t. [104-001]

Examples: trees in a forest; pieces in a chess set.

with aggregate defined as:
Elucidation: a is an object aggregate means: a is a material entity consisting 
exactly of a plurality of objects as member_parts. [025-002]

More formally:
If a is an object aggregate, then if a exists at t, there are objects o1, 
…,on at t such that for all x (x continuant_part_of a at t iff x overlaps 
some oi at t)

An entity a is an object aggregate if and only if there is a mutually 
exhaustive and pairwise disjoint partition of a into objects [63]. 

Examples: a symphony orchestra, the aggregate of bearings in a constant 
velocity axle joint, the nitrogen atoms in the atmosphere, a collection of 
cells in a blood biobank.

Does that work for you for homogenous aggregates and mixtures?

Original comment by mcour...@gmail.com on 1 Jun 2012 at 5:15

[GoogleCodeExporter]

History:

OBI started to want to use granular parts so we included it to use. Then, my 
fast draft bfo.owl with newids incorporated it and gave it a BFO id because I 
thought, at the time, that we had some agreement to include it. Then, recently, 
 OBI, wanting to use newer BFO ids started to use that draft, but did not 
migrate that relation so there became a duplicate.

However, in the current reference draft it is not yet included, so it is once 
again not in BFO. But I support having it there. 

I'm less sure about has_component. I've yet to see a definition of 
has_component that makes good enough sense. Do you have one?

I've also had has_granular_process_part in my prototype, now also out, but 
which I think should be considered for inclusion. This is for bridging 
macroscopic processes that are aggregates of much smaller processes of 
homogenous type. Examples are bulk reactions which are composed of many 
individual reactions, bubbling in carbonated liquid, which are make small 
cavitation and condensation processes, rain formation, etc. 

Original comment by alanruttenberg@gmail.com on 1 Jun 2012 at 5:27

[GoogleCodeExporter]

Melanie, the aggregate definition misses out the homogeneity condition. All it 
says is that there is a bunch of objects (of any type or scale).

Original comment by alanruttenberg@gmail.com on 1 Jun 2012 at 5:29

[GoogleCodeExporter]

I understand that for aggregate there is currently no homogeneity condition. My 
question for Colin is is that absolutely required, and if yes, should that be 
part of the relation? Instead of creating a new relation, could we for example 
say has_member_part only nitrogen atoms, or would there be limitations I didn't 
consider?

Original comment by mcour...@gmail.com on 1 Jun 2012 at 5:42

[GoogleCodeExporter]

On aggregates and grains see also: 
http://www.jbiomedsem.com/content/2/S4/S2

Original comment by steschu@gmail.com on 15 Jun 2012 at 8:07

[GoogleCodeExporter]

Hello everybody

Sorry - posted that and ran away to the land of Tarski and Ingarden. Back now.

I think for BFO purposes then member parthood will be good.  A problem which 
occurs to me with homogeneity on the continuant side is that many systems above 
absolute zero will have ions in equilibrium with neutral molecules and indeed 
trivial impurities.

I kind of want to say

'portion of gold' has_member_part 
mostly_other_parts_present_to_the_extent_that_they_don't_result_in_causal_powers
_at_variance_with_the_notional_pure_substance 'gold atom'

for something as simple as a lump of gold. (I don't think that quantifier is in 
OWL.)

Generally it's the atypical bits in a substance like only those atoms that are 
on the surface, or the ions, or the vacancies that are responsible for many of 
the substance's causal powers.  I don't know of a good treatment for this in 
general.

On components, having thought further, my initial wish for a has_component 
relation actually differed from the sort of components in the J Biomed Sem 
paper Stefan talks about.  I was thinking of components as being like the water 
and ethanol in vodka, but really these are just member parts that are scattered 
aggregates.

So, I am happy!

Best wishes,
Colin.

Original comment by batchelorc@rsc.org on 20 Jun 2012 at 10:47

[GoogleCodeExporter]

This is still unresolved, and many ontologies like PRO need to represent 
cardinality using parthood requiring a has_component like hack.

There was a suggestion to use member_part_of, but I think the existing 
definition is problematic:

DEFINITION: b member_part_of c at t =Def. b is an object 
& there is at t a mutually exhaustive and pairwise disjoint partition of c into 
objects x1,  …, xn (for some n > 1) with b = xi  for some 1 ≤ i ≤ n. 
[026-004]

Note there is no definition in the CLIF (often a sign that the definition is 
masquerading as formal but is actually unclear).

Note that transitivity obviously follows from this definition. I have no idea 
whether this corresponds to peoples' intuitions about member parts, but it 
should be explicitly stated (or the definition changed). It should be noted 
that this would bar member_part_of as working for the OWL cardinality hack.

As an example, consider a protein complex. There are multiple JEPD partitions 
of a complex. You could break it into proteins, or amino acids, or particles. 
This is recursive, with each protein having a JEPD partition consisting of AAs.

As an aside, the range declared in the OWL for at-some-times (continuant - 
axiom 509) does not match the reference (object aggregate).

This means there are a variety of ways to violate the reference without it 
being noticed in the OWL. For example, if I were to say that protein complex 
has proteins as member parts, and I were to also that the protein complex had a 
space as part (where the space is not part of an existing member), this would 
be in violation. This wouldn't be caught at the OWL level. We can't expect the 
OWL to catch everything but in this case I think it's possible. Should at least 
be noted as documentation for ordinary humans.

Anyway the main issue here is: is member parthood transitive in the reference? 
If not, the definition needs fixed.

Original comment by cmung...@gmail.com on 2 Apr 2013 at 6:33

[GoogleCodeExporter]

I raised this---Chris, do you have a list of use cases like for example in PRO 
for this?

I seem to have survived without this in BFO since last June, so I'm happy for a 
BFO 2 release to go ahead without it.

Original comment by batchelorc@rsc.org on 9 May 2013 at 3:32

[GoogleCodeExporter]

Barry has said that bfo2 will have has granular part and has granular process. 
Regarding transitivity of member part, I concur that the definition suggests 
this, which could occur for objects that are also object aggregates. Barry 
should address this - I will bring it to his attention. 

However, the member part of is not intended as a 'hack' for owl transitivity 
issues - that would be put of scope, IMO. I'm also no personally in favor of 
using a hack at all. 

I'm on the PRO project and have proposed a specific relation from protein 
complexes to constituents to avoid a hack. My recommendation is that others do 
the same - since the parent is part-of, interoperability at that level is 
achieved, and we avoid confusion over what the ontological criterion for using 
the relation is, when member of isn't the right choice. 

Original comment by alanruttenberg@gmail.com on 9 May 2013 at 4:09

• Changed state: Accepted
• Added labels: Milestone-bfo2-release

[GoogleCodeExporter]

I have spoken to Barry. He has withdrawn his support for granular part or 
granular process parts. So these will have to live in RO or successor, unless 
we want to push back.

Regarding member-part-of transitivity although he did not respond with a 
plausible example, he did not want to make it not transitive. However he 
proposes adding a non-transitive subproperty which has an additional axiom that 
prevents transitivity.

b direct_member_part_of c at t =Def. b member_part_of c and for all d if (b 
member_part_of d and d member_part_of c) then b = d or d ­= c.

Please comment. 

Original comment by alanruttenberg@gmail.com on 10 May 2013 at 7:20

[GoogleCodeExporter]

Hello,

I think granularity-crossing relations are too ill-developed at this stage to 
go into BFO2, which ought to be minimal.

I don't very much care what the position is on transitivity of member parthood 
as long as the documentation is clear.

Best wishes,
Colin.

Original comment by batchelorc@rsc.org on 13 May 2013 at 10:05