Search Mailing List Archives


Limit search to: Subject & Body Subject Author
Sort by: Reverse Sort
Limit to: All This Week Last Week This Month Last Month
Select Date Range     through    

Note from an interloper (thanks)

Bill Andersen andersen at ontologyworks.com
Wed Apr 10 01:52:50 PDT 2002


On 4/9/02 15:04, "pdy951 at merle.it.northwestern.edu"
<pdy951 at merle.it.northwestern.edu> wrote:

> Bill,

I'm at a loss - I didn't get your name...

> Don't worry, you didn't hurt my feelings. Thanks for leaving my name out but
> people can get that by following the thread back.
> 
> Thanks for the critique, I'll keep it in mind. Are you aware that GO is not an
> ontology? I don't think that any of the computer scientists who is involved in
> GO think that it is a real ontology (subject of ISMB meeting last summer),
> it's a hierarchy of processes, components and functions (not gene names)
> ranging from general to quite specific. The name has already caught on so
> changing it to Gene Nontology or G!O is proabably not going to happen.

I don't want the GO crowd to give up on building an ontology.  I think that
it's a really important thing to do, and not just because my company is in
the ontology-based software business.

I not GO, then what?  GO already has a tremendous amount of momentum behind
it and a lot of smart people working on it.  The amount of data out there
isn't shrinking, and some kind of organizing framework is going to prove
essential - if it's built the right way.

One thing that could be done is to do a parallel effort (call it LoGO, for
logical GO, or something fun like that) to give as much logical and
philosophical underpinning to GO as makes sense, given the fact that much of
it represents a science that is in rapid flux.  Ontology geeks could, for
example, take the existing GO as a starting place, go off in a corner and
work on formalizing the low-hanging fruit, and then sanity check the result
(in plain English) with the group at large.  Anything that is uncomfortable
to the group could simply be shelved or rejected.

I said something like this at a bioinformatics workshop at NAS two years ago
in response to someone who told me "you can't put all those logical rules
into a database - the data are too dirty".  Maybe so, but my response to him
was "Are you telling me you don't know *anything* about your domain?  You
aren't sure, for example, if molecules exist?  Or that a molecule is not a
reaction?"  Well, that was an admittedly hyperbolic example, but it makes a
point.  There is always much more than we think that is stable and can be
formalized.  But a lot of this "bread-and-butter" stuff tends not to spring
to mind (understandably so) for subject experts, who cached a lot of it away
in the recesses of their minds before they entered grad school.

The need to formalize this background theory becomes critical whenever you
take humans even partially out the loop.  Computers have no idea what our
natural language terms mean, but they're pretty good at keeping us honest in
their use if we only tell them a little bit about how to constrain their
possible interpretations.  Peter Karp's work is a good example of this - you
can get a long way with a small amount of formalization, and even farther
with more.  It's kind of an black art, given the present state of ontology
and knowledge representation, to decide how much is enough, but it can be
done.

Correct me if I'm wrong, but from what I know about GO, it's an attempt to
organize, as you say, terminology relating to gene function.  And GO seems
to attack the problem directly at that level, without relying for example on
having a theory of chemistry or molecular structure, etc., to build on.  I
have a feeling that this lack of theoretical basis was at the core of Sydney
Brenner's beef.
 
> I was just as vague and opinionated as Sydney in my first post. I apologize.
> Sydney used the analogy of our names and how they don't really define who we
> are. We don't name the genes in GO we name the things they do based on
> scientific evidence. So for Sydney, I have evidence that he is a molecular
> biologist so I would call him that. Maybe if I had less evidence, I would call
> him a biologist or a scientist. Calling him Sydney Brenner doesn't say
> anything about who he is but trying to chraracterize what he does based on
> evidence has merit. From his analogy I didn't think he understood that we were
> not trying to name the genes.

That could be.  Without more details from him on what motivated his critique
it's hard to say.  There is a lot more to be said about this business of
"self-naming" objects but it gets quickly into hairy philosophical and
logical issues.  The upshot is that no such auto-naming scheme as Sydney
suggests can be guaranteed to work - we will always have to rely on our folk
use of plain old names, as GO does, to interpret what some of these concepts
mean.  So, while his heart is, IMHO, in the right place, I think he has the
details wrong.  If anyone is interested in why I think this is, please feel
free to write, but since it's not central to GO and will be a diversion to
anyone who doesn't like logic, perhaps it's best kept offline.

> Maybe we all need to work on our reading comprehension. Anyways, I have
> learned never to check my email on a weekend.

Agreed!

 .bill


--
This message is from the GOFriends moderated mailing list.  A list of public
announcements and discussion of the Gene Ontology (GO) project.
Problems with the list?           E-mail: owner-gofriends at geneontology.org
Subscribing   send   "subscribe"   to   gofriends-request at geneontology.org
Unsubscribing send   "unsubscribe"  to  gofriends-request at geneontology.org
Web:          http://www.geneontology.org/



More information about the go-friends mailing list