Slashdot Mirror
Gould Op-Ed: Genes' Emergent Properties Matters
A reader writes "The New York Times has an
op-ed piece in Monday's paper about the smaller-than-expected number of genes in the human genome (around 30,000 genes, versus 19,000 for a simple roundworm and the 100,000+ that were expected).
With so few genes, it may be the case that the emergent properties of the combinations of genes, as much as the genes themselves, are contributing to our complexity. I suppose the honchos at Santa Fe Institute are rewriting their grant proposals already."
printf("%d\n",i); and yet this code: printf("1\n");
printf("2\n"); can only produce 2. So why should I be astonished when one genome can do more with 30,000 genes than another genome with 100,000? Are biologists really no smarter than those managerial types that compute productivity by counting lines of code?
And am I to start using fancy schmancy language like 'emergent property' when I talk about ooh...so sophisticated coding techniques like looping and reusable subroutines?
---- SIGFPE
I've been listening to this stuff about the lower-than-expected number of genes for a while now, and it is surprising to me that no one has mentioned the less of chaos theory. Chaos theory is the study of systems which have very simple equations of motion, but which have an extremely compilcated behaviour.
So... even quite a simple creature, such as an insect, can have very complex behaviour, even with simple equations of motion. So it's really the system structure that matters, not the number of parameters in the system specification.
And remember that computers are very simple indeed. They're just interconnected switches and things. But the program loaded into the hardware makes it complex. So the complexity of human beings comes from the ability to load programs and execute them.
One reason why mammals have so few genes compared to, say, amphibians, is because we're better designed. Being warm-blooded, there's a controlled environment inside the body; so you only need enzymes that work for that environment.
Amphibians need enzymes that work in all temperatures from about ten centigrade up. And because each enzyme only works in a limited range of temperatures, they need lots of different enzymes to do the same task; and these all need encoding. Hence the very large sequences.
You write: "Isn't this a lot like having a lot of small programs that, when scripted together, can outperform a large, monolithic one?"
:o)
:o)
Alas, I fear it is the exact opposite
In the Unix philosphy, each little script is totally self-contained, its operation can be analyzed independently of the context, and combining several scripts will just yield the combination of their results.
Our genes, on the other hand, are not independant from each other : the presence of one given gene can have significant influence on the expression of another gene. A genotype cannot be analyzed gene-by-gene: the result of a genotype cannot be predicted from the result of each gene taken separately.
Which means that we are much more a MSWindows-like machinery, where in order to get any little thing working you must have tons of other programs / services / libraries installed and running in a very precise way - otherwise you're on your way to catastrophe. Every component is totally dependent on the context, and the context is the total sum of *all* other components.
Sad, but true : from a software engineering perspective, we are a perfect example of brain-damaged design
Thomas Miconi
It's not a troll just because you don't get the joke.
--
--
You are a fucking moron.
It seems that it's a joke based on a bad Japanse to English translation. This should help you out a bit. It's damn funny.
http://rmitz.org/AYB3.swf
If tits were wings it'd be flying around.
But with each additional gene, the number of interactions between its expression and the expressions of other genes rises exponentially, doesn't it? If I'm surprised by anything, it's that people actually thought there might be a "gene for foo".
The "cue the foo posts in 3, 2, 1..." posts will commence with no subsequent foo posts in 3, 2, 1...
"Emergent properties of the combination of genes" have been known for decades to be the dominant factor in genotype-to-phenotype translation. AI computer scientists working on genetic algorithms have called this epistasis, borrowing the word from biology (see here), and giving it a slightly broader meaning:
"You have epistasis when the expression of a given gene has a significant effect on the expression of other genes, thereby inducing the fact that a genotype of N genes cannot be analyzed by observing the effect of each gene separately". The unwritten corollary being: "which is quite a pain in the ass".
Genetic algorithms work best (in comparison to other methods) when the problem space is highly-yet-not-too-highly epistatic. See this page for extensive information, or just try a Google search.