nodalpoint.org - The biological ontology backlash - Comments

I've just finished reading

Greg — Tue, 11 Oct 2005 12:38:34 -0400

I've just finished reading your extended rant, and I agree these kinds of debates have existed for ages in various software development circle. I'll leave my extended comment here, rather than reply to your post as last time I left a comment on your site it went missing ?

The reusable code debate is only just starting to flare up in a bioinformatics research context, see here (or any pub within close proximity to a bioinformatics research department).

I would argue that design standards are more akin to best practices rather than coding standards which are more syntactic. Thus best practices are in fact equal to the design standards and the guidelines for refactoring you mention in your post. Ontology refactoring rules I would see as more generalized versions of the numbered points in the nature article. Or at least they might provide a useful guide.

The semantic web best practices working group is a good start, along with the semantic web for life sciences mailing list for developing biologically specific best practices. At the end of the day very little formal work on these kinds of problems has been done, in fact these recent articles are the first time I've heard of these kinds of issues being discussed in the main stream biological journals.

One issue that strikes me is whether or not these kinds of best practices can be arrived at via common sense. I've had it put to me that understanding ontological engineering principles is necessary to author an ontology. I wonder if a community based approach might feasible ? For example a wiki-ontology development environment were everyone in the biological community (many experts) might be able to participate. This may also go some way towards solving the problem of multiple incompatible ontologies. In a sense you don't really solve the problem, you rely on the community to come to a consensus and simply distinguish anything else as an "uncommon" term. A consensus may change at some point, however the semantic web provides mechanisms to write back that type of information.

refactoring bio-ontologies

mmhohman — Fri, 07 Oct 2005 17:46:09 -0400

I think there's a useful middle position to be taken here, between the immediately practical and the lofty ivory tower purist, a position taken from a similar debate that has been ongoing in the software development community for quite some time. Basically, I think we ought to think about developing the concepts of "ontology refactoring" and ontology "code smells" (perhaps better called poor design symptoms or something more appropriate to the vernacular of biomedical informatics). I have described my thoughts at probably greater length than necessary in a post on my own blog: http://blog.moseshohman.com/?p=19

Has anyone proposed developing a catalog of ontology refactorings? I couldn't find anything with a very brief Google search.

Standards will have to be enforced

Spitshine — Tue, 04 Oct 2005 06:57:41 -0400

The problem with standards is that there are costs for following them. I don't see that the community of biologists and bioinformaticians will come up with enough peer pressure but the journals (or publishing bodies) might enforce standards as they did with the deposition of nucleotide sequences a long while ago.
Personally, the strongest lever would be with the funding bodies but they don't seem to have been very instrumental in influencing how science is done (in contrast to what is done).

Usually, these standards want you to publish as much experimental information as possible, still many people want to keep the nitty-gritty details to themselves.

The response from the MGED

Greg — Mon, 03 Oct 2005 19:20:37 -0400

The response from the MGED people is interesting. The following quote seems highlight the main point of contention:

Controlled terminology as provided through an ontology was urgently required for the description of microarray experiments; biologists did not expect emerging ontologies to be ideally-engineered, and we would be the first to state that the MO contains compromises. In the absence of such controlled terminology, much of the data being generated today would not be available in a form readily amenable to comparative analysis, either now or in the future.

I have seen this argument played out many times: Getting the job done vs. doing it right. I believe the way that the MO people see this (and biologists too) is that it is more important to do biology than to worry about the minutia of ontology construction and data format standardization. There are many reasons why this is the case, least of which is that doing biology is arguably more interesting than the latter. However I think this is a dangerous position to take.

The problem is one of short term vs. long term gain. As has been discussed on nodalpoint many times, doing biological data integration is annoying and time consuming. The only possible way to progress in this area is to take biological data standards seriously. The only parallel that I can thing of to bring this point home is web standards. It wasn't until people (the users, not the companies) started to take web standards seriously that the their usage increased.

Until biologists and biologists turned bioinformaticians (yes, that means you) take standards development seriously, the situation is *not* likely to improve.

Hmm, the second time I heard

Spitshine — Thu, 29 Sep 2005 03:38:11 -0400

Hmm, the second time I heard about the Gene Ontologies initiative(Georgia Tech Conf., 1999?), it was already critized the highly pragmatic way it was constructed.
Michael Ashburner was quoted as not too much being interested in satisfying the pundits. Technically, I am with him, it's better to have a complete ontology than a precise but useless one. And if there is critic, we get good ones eventually. The points that Soldatova and King make can be said about most databases in biology really, I am afraid.

The biological ontology backlash

Greg — Wed, 28 Sep 2005 13:32:19 -0400

It seems that the emperor has no clothes:

The utility of ontologies has been clearly demonstrated in several biological domains (e.g., Gene Ontology). However, within biology, the enthusiasm for ontologies has been accompanied by a general lack of awareness of what exactly ontologies are and how to use them.

This is from another Semantic Web related Nature article: Are the current ontologies in biology good ontologies?, via Julio Bonis. The point the authors are trying to make is that many of the current bio-ontologies are poorly designed and are therefor unsuitable for use in knowledge sharing applications and automated reasoning (i.e. Semantic Web). They put the MGED microarray ontology under the microscope, finding many inconsistencies.