Re: [CDAO-discuss] Obo-discuss Digest, Vol 40, Issue 1

[Arlin S. <sto...@um...>]

On Sep 11, 2009, at 11:34 AM, Alan Ruttenberg wrote:

> On Tue, Sep 8, 2009 at 3:51 PM, Arlin Stoltzfus  
> <sto...@um...> wrote:
>> On Sep 2, 2009, at 3:31 PM, Alan Ruttenberg wrote:
>>
>>> BTW, looked up and found CDAO. I think I've discussed this with  
>>> Hilmar
>>> in the past.
>>>
>>> There's overlap with interest that is pushing IAO development.  
>>> Perhaps
>>> you or one of your collaborators are interested in participating?
>>
>> Yes, many of the concepts in our ontology [CDAO] are information  
>> artefacts, though
>> this is not explicit. I consider "phyogenetic tree" to be an  
>> information
>> artefact (I can explain why if its not clear),
>
> I think it's clear, but it certainly couldn't hurt to say how you view
> this explicitly. I think that phyogenetic tree is more specifically a
> "model", (and I have an idea about how to define model) in a similar
> way that a gene regulatory network is a model.  See
> http://icbo.buffalo.edu/Presentations/Ruttenberg.pdf 91-93 . Not on
> slides, but talking point, was that the thing that distinguishes this
> sort of model from a data structure is that all the parts in the model
> are explicitly mapped to some sort of entity/statement about entities
> in "first order reality" (see Werner Ceuster's slide in that file #63
> - yellow is first order. Have to track down the defs of 1st,2nd,3rd he
> has)

Thanks for your response, Alan. I'm going to cc this to the cdao- 
discuss mailing list.  Others might be interested in trying to sort  
out the upper-level status of "tree" and other troublesome concepts.

The case with "tree" is a wee bit more complicated than "model" as  
defined.  Phylogeneticists often speak of an "unrooted tree" for the  
case where they have inferred the connectivity but not the direction  
of branches.  To the extent that branches map to the process of  
evolution, and evolution is in "first-order reality" (i'm not sure,  
because I don't what first-order reality is), the mapping must be less  
than a direct mapping to first-order reality, because evolution is  
always directional, whereas branches of phylogenetic trees may lack  
direction.   In this way, a tree is more like a data structure.

"Metabolic pathways" (which are, in some ways, similar to  
"phylogenetic trees") have the opposite problem-- their directionality  
is often over-specified relative to chemical reality, in which the  
flow of matter can go against the direction of a "pathway".

There might be other problems as well.  My perspective has been to  
avoid such issues on the grounds that I find upper-level ontologies  
are speculative and lack substantive implications for querying over  
domain knowledge.  Entire  books have been written on such things as  
what a "gene" is.  In earlier versions of the SO, a mRNA was "part_of"  
a gene, in spite of the fact that everyone learns in high school that  
mRNAs are made or RNA while genes are made of DNA (this error was  
removed in later versions of SO).

In other words, expert domain knowledge of genetics does not include  
knowing exactly what a gene is, and domain knowledge of genomics does  
not include knowing what a sequence is (or, at least, not in a frame- 
independent way-- and that might be part of the problem).  My view is  
that no one really knows what a tree is, and so we don't have to know  
this either.

Arlin