Query - Evolution without Allelles

[madd0ct0r]

I assume evolution predates allelles, so I know it's possible, and the fact that life in general uses them so much suggests that they're a better way of doing things.

But are they? and why?

I'm looking at trying solve some problems with genetic crunching.
Things like optimum room dimensions and window location and opening for cooling, as an example.
(Basically, engineering problems where lots of factors interact in complex ways)

Simply taking two traits and averaging them in the offspring, that's easy to do and easy to understand.

How important is the ability of alleles to carry recessive genes down through the generations?

how important is the binary choice of this entire gene is activated or this one?


The former I can see being important for a species as a whole (it gives you a reservoir of potentially useful stuff for when the environment changes) but rather less useful for the fixed solution i'd want.

Do allelles contribute to the selection and adaption process in other meaningful ways?

[Ariphaos]

You seem to have a misunderstanding of what alleles are.

An allele is a specific formulation of a gene. If one gene controlled eye color (which isn't true but for simplicity's sake...), brown, green, blue, etc. eyes would be alleles.

The most basic definition of evolution is the change in allele frequency over time. This will apply to anything with genes, by definition, because genes represent the individual mutable components and alleles are the various possible expression of such. If you leave genes behind, then sure, you've left alleles behind and you can have evolution without them. But if you have genes, by definition you have alleles, and vise-versa.

[madd0ct0r]

ahh, well yeah. I'm misunderstanding the word then.

for genes, how important is the concept of dominant and reccessive genes to solution convergence?

I'd got as far as the different values of my variables being genes, but biological evolution would simply choose daddy value or mummy value for each variable instead of averaging them (which would be my instinct for problem solving).
I was thinking that the pair of genes (mummy and daddy) were called an allelle.

[Ziggy Stardust]

for genes, how important is the concept of dominant and reccessive genes to solution convergence?

Alyrium will be able to give you a better answer than I can, as genetics isn't really my field. That said, it is important to remember that alleles are not simply binary values (i.e. gene 'on', gene 'off'). The relationship between allelic frequency and an observable phenotype can be quite complex, with contributions from a large number of different alleles in different parts of the genome. Similarly, it is also possible for a large degree of allelic variation to result in no observable phenotype change. Many traits follow various patterns of polygenic inheritance, and can not be classified under the Mendelian dominant-recessive model.

In any case, the concept of dominance is, at its simplest level, for a gene with only two allelic forms, the expression of the heterozygote is indistinguishable from that of one of the homozygotes, such that one allele is said to be dominant to the other. That is, AB and AA are the same, and BB is different, so A is dominant to B. So the concept of genetic dominance is a classification of the way alleles interact to produce a phenotype. However, it is very rare for a gene to follow a complete dominance model such as this. It is possible for the heterozygote to be an intermediate characteristic, it is possible for alleles to be codominant (the heterozygote expresses both traits, as opposed to just one or a blend of the two), and then we get into sex-linked traits and epistasis and it gets a lot more complicated.

[madd0ct0r]

Running that through Google translate I get:

"There's lot's of ways to do it with binary dominance being unusual. Therefore you could probably just take the averages and still get a converging solution.

PS, you should have done biology beyond GCSE"

ncase you are really really bored, here's the origional problem:

trying to get an enviromentally effcient house. I'm worried that if the model has slight errors in it, we'll converge to the wrong solution. Making the house fit for a wider range of environments (as a site would experience over a few years) seemed safer.

Thinking it over, a robust approach might be to evolve the house in a naturally cycling environment. Depends what data you've got, but if you can get a couple of years worth of daily weather from a nearby weather station (hah!) here's what I'd try:

Each trait has it's own mutation rate, keyed to how difficult it is to change in real life. So the amount a window is open can mutate very easily, wall colour and floor finish quite easily while wall thickness and ceiling height change only rarely.

Each generation corresponds to a different time period. If the climate changes dramatically over the day then either assess the same design for morning, noon and night and produce an average fitness, or treat each part of the day as a seperate generation.

We're talking really heavy computing load here either way. You need a big population to avoid getting stuck at local optimals, espcially since, like animals, changing one thing in the house will probably affect a lot of other things so only small and gradual changes are likely to be beneficial. Say 100 houses in each generation with only the top 10% breeding. Or 400 with only the top 5% breeding and so forth

It'll converge quite slowly I suspect, but eventually you should get a reasonably stable house. Unless you somehow run a second fitness test for liveability you might have to 'fix' the design space so stairways and ceilings remain fit for humans.

The traits like stored heat would be inherited from generation to generation (half from each parent) and altered by the current generation before breeding. We're talking mendelavian evolution here It's the only way I can see to allow the species to develop stuff like storing heat for winter (a gap of dozens of generations). The other option is to have 1 year = 1 generation and test multiple times for an average fitness. Probably computationally more efficient, but rather less cool.