From: Seth Cooper (scooper@cs.washington.edu)
Date: Wed Dec 01 2004 - 09:14:49 PST
The paper read is titled "The Evolutionary Origin of Complex Features"
written by Lenski et al. It describes how complex features, such as
eyes, can come about through Evolution by simulating the evolution of
analogous features in virtual organisms.
An argument often made against evolution is that it does not explain
the occurrence of complex features in organisms, such as eyes. These
features are not likely to suddenly evolve by themselves. Evolution
tries to explain these features as the result of smaller, incremental
changes that build over time, finally producing a more complex feature.
An important result presented in the paper is that it is possible for
such complex features to come about by evolution. The fact that the
most complex of the functions, EQU, was evolved in a reasonably large
percentage of the trials in an impressive indication that incremental
changes can, given time, add up to becoming something greater than just
the sum of their parts.
Another important result of the paper is that the experiments have
recorded the entire history of the organisms that evolved the complex
feature. A problem with looking at incremental evolution in biology is
that many of the intermediate forms have been lost. Because these
experiments were conducted on a computer, all of the intermediate forms
are remembered and can be analyzed. Some interesting observations are
made. For instance, the organisms that evolved the complex features did
not necessarily follow evolutionary paths that were always positive.
Sometimes deleterious steps were made that ended up being beneficial in
the long run, due to some other mutation that occurred later. Another
interesting observation is that receiving rewards for the simpler blocks
that made up EQU was necessary to eventually reach an organism that
performed EQU.
The largest flaw in the paper is the question of whether or not the
results seen in virtual organisms actually apply to biological
evolution. The virtual organisms in the paper can be seen as performing
a sort of hill-climbing search for the EQU function, where each genome
is a state and moves are made somewhat randomly throughout the state
space by mutating. The search is guided by a heuristic, which is
receiving lesser rewards for lesser subgoals. Indeed, without that
heuristic, the EQU function is never found. When viewed in this light,
it is not at all surprising that some organisms eventually reached the
goal. It is possible that biological evolution is a similar type of
search; however, the paper does not tie the two together.
One important research question that is mentioned at the end of the
paper is to see what happens when these ideas are attempted with
organisms using sexual reproduction. All the organisms in the
experiments presented in the paper reproduce asexually. It would be
interesting to see what kind if impact sexual reproduction would have on
the results, for instance, if it would speed up the development a
complex feature. This would matter because it is the method that humans
and many of the higher level organisms we are interested in use for
reproduction.
It might also be interesting to compare the results in this paper with
a search algorithm that is simply searching the space of genomes,
looking for ones that perform the EQU function. Then the relative
efficiency of the evolutionary approach could be determined. This would
matter because it would give some insight into how efficient
evolutionary biology is. Also, one could look at it the opposite way
and try applying this evolutionary algorithm to other kinds of searches.
This archive was generated by hypermail 2.1.6 : Wed Dec 01 2004 - 09:14:51 PST