quote:
In the Dawkins model (Fig. 1, Below right), the target sequence and parameters are set as per Dawkins’ original exercise. Running the model will show convergence on the target usually in 30 to 60 generations (iterations). Since this is a probabilistic exercise involving a random starting sequence and random mutations, the result will vary with each run.
This model is illustrative of the progressive nature of mutation and selection, but it is not a true evolutionary algorithm as its selection function selects for a predetermined pattern.
quote:
Error catastrophe occurs when genetic information is destroyed by mutations at such a rate that all progeny are less fit than the parent/s so that selection cannot maintain the integrity of the genome and, in a Dawkinsian-type model, a target sequence cannot be achieved.
This is what would happen to all of us if we were to suffer several tens of thousands of germline mutations per generation. To put it bluntly, this model is not biologically accurate.
quote:
In effect, the mutation rate cannot be much greater than one per genome per generation. This then severely limits the rate of progress from a chimp-like species to human, if this were possible, even with perfect selection and all the other assumptions.
This does not follow. The measured mutation rate is 2-4 nucleotides per generation (as high as 5 in one study). Given a reasonable population size, this means that there have been several billion mutations in the human germline since our last common ancestor.
And this is still far below the threshold for the "error catastrophe" scenario in this simulation.
quote:
DNA Model
Any standard biochemistry text would describe how proteins are made from the information contained in the base sequences on DNA. We have provided a brief tutorial provided with the program (under ). An important difference between the DNA model and Dawkins’ Model, or any alphabet model, is that the DNA of an organism is not compared directly with the target as it is in alphabetical model. Another important factor is redundancy, some of the amino acids can be coded by different codons. With some codons, only the first two base pairs are needed to determine which amino acid is produced. This gives the genetic code some resistance to change. In some cases you would require more than one mutation to convert the code of one expressed amino acid into the code for another.
Since the evolution of proteins is not a pattern matching proceedure, we do not expect every set of proteins to be a mere permutation search from a similar sized protein.
As with the error catastrophe model, this model also neglects to model the way real mutations occur in actual organisms. It ignores the process of inversion, duplication, and recombination in producing novel proteins.
In other words, you can't knowingly make an inaccurate model of protein evolution and pretend that it says anything about real protein evolution. It make work on the uneducated, but it won't work on anyone else.
I would call this piece of software the "Grand Canyon Project" of Evolutionary Algorithms. Programmed specifically to mislead.
[This message has been edited by Rationalist, 09-03-2002]
™ Version 4.2