Monday, November 11, 2002
      ( 9:17 AM ) Matt  


Last weekend was pretty busy. I went to my first football game in a real stadium. My band, Tater and the Botany Boys, practiced for an upcoming gig. Fun stuff.

But back to okonomigo! Friday night I was out at the pub with some of my buddies, and we were talking about sexual reproduction. No, not like that. These are my botany buddies. We were talking about the production of gametes and the recombination to form a diploid cell.

You might be asking: how does this relate to okonomigo? Well, I plan to use a genetic algorithm to design an optimal neural network for learning the game of go. This has been attempted before, but what I hope will make my approach unique is that I will base my genetics directly on the way DNA is used in the real world. I figure this system seems to have worked pretty well for us, so maybe it will work for my computer too. So, my ears always perk up when people talk about the way DNA recombines.

At the pub, Pete proposed the question:

When DNA crosses over during meiosis, are the crossover points always lined up?

Well, that's a really interesting question. If they weren't lined up, then you might get DNA which is slightly longer or shorter after the crossover. Also, if they were mismatched by some amount that's not a multiple of 3 (the length of a codon) then there would be a frame shift, completely changing the meaning of the DNA past that point.

The crazy thing is, this does happen. How often it happens is not known. It is known that there are proteins which aid this process and detect errors like this, and try to correct for them. However, these error's aren't always caught. Sometimes these errors happen in a 'junk DNA' region. That is, a region after a stop codon, and before the next start codon. There are 3 stop codons, and only 1 start codon, so if you had completely random DNA, a majority of it would be junk. If some of the DNA doesn't crossover correctly, the gamete may be aborted. Even if the gamete were to make it all the way to recombine with a gamete of the opposite sex, the resulting diploid cell may not be viable, and the result would probably go unnoticed. On the very rare chance that the resulting cell survives, and the altered region is not junk DNA, this may introduce a mutant into the population. More likely though, the protein which that segment encodes for would be broken, and the gene on the corresponding chromosome would still work, effectively producing a heterozygous individual instead of a homozygous dominant individual.

But, on a very rare occasion, a successful mutant may be introduced into the population. This sort of mutation is different than the type introduced in most genetic algorithms because it alters the length of the DNA. Most mutations in genetic algorithms only alter values within a sequence instead of altering the length of the sequence. It's this sort of fascinating stuff I want to add to the okonomi project.

# -

Comments: Post a Comment

Dreams I have...

Powered by Blogger
Feel free to e-mail me.

free hit counter