[Obo-anatomy] Re: anatomy -> GO development questions

[chris m. <cj...@fr...>]

Hi Doug & Melissa

I think this is a good question to kick off the obo-anatomy mail  
list; even though this is development, it's impossible to tease  
anatomy and dev apart, especially if we want to come up with a  
reference anatomy that accounts for developmental stages.

On Apr 28, 2006, at 1:39 PM, Doug howe wrote:

> There has been some discussion recently about temporal aspects of  
> development.  It seems like the following situation exists (Here is  
> one example of many):
>
> In an anatomy ontology, a structure X develops from a structure Y.   
> for example, the inner ear develops from the otic vesicle.
> So, the otic vesicle would not all be an is_a (its not a subtype of  
> inner ear) or a part_of (they do not coexist) child of inner ear.   
> Really, they have a develops_from relation.

Is this a transformation relation (ie identity is preserved in  
changing from otic vesicle to inner ear) or a derivation one?

I believe the relation here is transformation_of

http://obo.sourceforge.net/relationship/#OBO_REL:transformation_of

currently you (and everyone else) is using develops_from, not in RO,  
but which could be defined as being inclusive of both derives_from  
and transformation_of. I know if may be a lot of work, but I think it  
could be beneficial to work towards distinguishing these in the  
zebrafish anatomy ontology as a long term goal - read on for my  
arguments why..

> An anatomy ontology would have:
> otic vesicle
> ---[dev. from]inner ear
>
> In the GO, we might have:
> ear development
> ---[p]inner ear development
> ------[p]otic vesicle development
>
> In GO, X development terms for these structures are considered to  
> have a part_of relationship to each other, while the anatomical  
> terms they are derived from of have a develops from relationship to  
> each other.

just a minor correction - I think you mean "the anatomical entity  
they are derived from"!

> I think this means that instances of 'inner ear' and 'otic  
> vesicle' (according to the anatomy ontology) may overlap  
> incompletely in time or the last moment of existence of otic  
> vesicle may be the first moment of existence of inner  
> ear..depending on what your anatomy ontology dictates....But the  
> structures do not necessarily have to co-exist.

if it is a transformation relation, then it is just one anatomical  
entity, so there is no question of co-existence. If it is  
derives_from, then there could be co-existence.

Let's carry on with the assumption that this is a transformation:

I imagine the dividing line between what constitutes an otic vesicle  
and what constitutes an inner ear is fiat - and in fact it may be  
valid to talk of intermediate types. This is fine. We may even have n  
types, U0, U1, ..., Un, each with arbitrary fiat demarcations (eg  
Thelier stages if this was a mouse). We would like (I think) to  
divide it in such a way that the types are disjoint - i.e. it is  
impossible for a single anatomical entity instance to instantiate  
both U2 and U3 *at the same time*. Of course, this is arbitrary, as  
developmental transitions are in general continuous rather than  
leaping from one state to another. Reasonable people may disagree as  
to where to draw the line, but disjoint fiat types are still useful.

So yes, it is indeed the case that both otic vesicle and inner ear  
may not co-exist (at least on the same side of the head in any one  
fish) in any snapshot in time

>
> However in GO, for 'otic vesicle development' to be part_of 'inner  
> ear development', wouldn't 'otic vesicle' and 'inner ear' need to  
> co-exist in time [and possibly space as well?].?   For example, how  
> can otic vesicle development be part of inner ear development if  
> the otic vesicle development is occurring before the inner ear exists?

Using standard GO templates, the definition of inner ear development  
would be:
"The process whose specific outcome is the progression of the inner  
ear over time, from its formation to the mature structure"

So the spatio-temporal extent of any instance of this process could  
include include the otic vesicle stage, if we consider the formation  
of the inner ear to really be the formation of the otic vesicle (or  
even otic placode? this is interesting, as we have a material entity  
forming from an immaterial one, a cavity. not unusual in early  
development I'm sure but I wonder if this causes problems for our  
definitions...).

David has long been an advocate of making the temporal extent of  
developmental processes very explicit in the definition.  
(Coincidentally I just asked him a similar question about heart  
development). So the definition above could give both "from" and "to"  
with preferably both coming from a reference ontology with  
developmental stages. Alternatively the "from" could be left implicit  
as it is not if we have a formal definition of "has_outcome" such  
that we can infer unambiguously the temporal extent (eg using  
transormation_of)

I'm still musing over the current definition template for  
development. According to the definition above, any part of X  
development would also stand in an is_a relation to X development.  
This is because the smaller subprocesses result in a progression.  
Perhaps the "specific" qualifier excludes this, I'm not sure. Perhaps  
we're really talking about the aggregate of processes that each have  
some progression towards an inner ear as their outcome?

> I think biologically, such arrangements of GO and anatomy  
> ontologies in isolation are correct, but does it create some sort  
> of conflict for the derivation of GO terms that reference anatomy  
> terms?

this is a valid concern - I'm against having ad-hoc rules for  
inference of the DAG structure of GO development ontology from the  
DAG structure of the anatomical ontology - we should put this on a  
firm footing

>
> Are there new issues with more complex arrangements?  For example  
> the anatomy says that anatomical structures X, Y and Z develop from  
> structure Y?  Is GO: 'Y development' part_of 'X development', 'Y  
> development' and 'Z development'?

this would be derives_from:

X derives_from Y
Z derives_from Y
(but we wouldn't state that Y derives_from Y)

In the case of derivation, I think you can only infer that:

X development has_part Y development
Z development has_part Y development

if you were to say:

Y development part_of X development

this may be wrong, as it could be the case that Y may decide to  
derive_into an X or a Z, an either-or situation. This is allowed by  
the anatomy, but the process ontology state that all Y-devs are  
part_of some X-dev.

(and the has_part inference only works if X, Y and Z development have  
start and end points defined in such a way as to span the fiat or  
bona-fide temporal demarcations between these types)

To summarise: distinguishing derives_from and transformation_of is  
useful for inference, and stronger definitions may be needed for the  
development sub-ontology of GO to precisely state the temporal extent.

Cheers
Chris

> Thanks for your comments,
> Doug and Melissa
> Designated Troublemakers at ZFIN