Saturday, June 18, 2005

Identical, same, similar, different

Well, it's been a busy week preparing to move to Philadelphia for a while, but now I have to write--it's bugging me too much not to! My graduate research is in the informatics of comparative anatomy, specifically in the validity and implications of comparing model organisms, and I feel like writing about that tonight.

First of all, a meta-observation: because my work is at the intersection of biology (anatomy), computer science, and mathematics, I have a pesky homonymy problem--a lot of the terms I use have one meaning in biology and a different one in mathematics or knowledge representation or computer science or whatever. For example, "model", "homology", "topology"--all of these are terms whose meanings vary depending on context. So if I seem to spend an awful lot of time up front defining terms, it's only partly because of my inherent pedantry--it's also very important to make sure that we're all on the same page about what things mean as we proceed. My PhD committee has 3 computer scientists, a vertebrate embryologist, and a comparative genomicist on it--so we definitely use words differently sometimes.

For instance, in biology a "model" is an organism--in this case an animal, such as a mouse or a zebrafish. In knowledge representation, on the other hand, a "model" is a simplified representation of a real-world object (or referent). My work is based on the Foundational Model of Anatomy, a model (in the second sense) of human anatomy for the computer. I used the previously-developed human anatomy model as a template for representing mouse anatomy. But because the mouse is used in medicine as a model (in the first sense) for human disease, that makes my work a model of a model. That kind of homonymy happens a lot in my work, so we'll spend time up front establishing some definitions, so that later we can make faster progress.

In order to talk about comparing anatomy across species, we'll need to have some kind of metric for comparison. Obviously (or maybe not so obviously, so we'll stipulate it), it's kind of silly to talk about anatomical similarity in quantitative terms (what would it mean to say that the human prostate is 20% similar to the mouse prostate?). If we are describing anatomical similarity, we use qualitative terms such as "identical", "same", "similar", or "different". Given that, how do we establish a metric for comparison?

We can talk about distance along a number line, even without assigning exact quantities to each property. It would look something like this:

(ok, this ASCII art has to go, because those dots to force it to do space right are just confusing. consider this a placeholder for the moment, till I can whip out something in PhotoShop. UPDATE: done)

So if we can set up such a line, we can compare similarity, even without assigning quantities to the anatomical structures under comparison. The more similar two things are, the closer they are on the number line, and the more different they are, the more distant they are. The example above is a representation that the mouse heart is more similar to the human heart than the mouse prostate is to the human prostate--the hearts are close to each other on the line, while the prostates are relatively far apart.

That will be our basic modus operandi; now we have to establish what we compare, why, and how.


Post a Comment

<< Home