From: Jon Froehlich (jfroehli@cs.washington.edu)
Date: Wed Dec 01 2004 - 10:44:49 PST
1. Paper title/Author
The Evolutionary Origin of Complex Features by Richard Lenski, Charles
Ofria, Robert Pennock and Christoph Adami
2. One-line Summary
Darwin's theory of evolution is applied to the domain of "digital organisms"
in an attempt to show how complex functions can originate from random
mutation and natural selection.
3. Two Most Important Ideas and Why
(i) The authors show how experiments involving "digital organisms" may be
useful in understanding biological features that exist in the physical
world, particularly with regards to evolution. They demonstrate through
experimentation that in evolution simple functions serve as foundational
building blocks for larger, more complex functions. Their experiments show
that, in fact, complex features never evolve when simpler functions were not
rewarded (and therefore not preexisting). This is a valuable result as it is
consistent with Darwin's hypothesis that complex features generally evolve
by modifying existing structures and functions (i.e. mutation) rather than
through some arbitrary formation.
(ii) Building on point (i) above, the authors show that there are typically
many different evolutionary paths that result in producing the same complex
function. This implies that there is a relatively dense space of
satisfactory genomic solutions for any given function in their evolutionary
model. The authors pursued this idea by comparing trajectories and solutions
of many populations that independently evolved the same complex function
(the EQU function). Interestingly, though different populations always
evolved simpler functions before EQU, no particular simple function was
required for EQU's development. Again, I think the importance here relates
to its value to biology and evolutionary theory. Is it the case that highly
rewarded evolutionary traits may have developed independently of each other
- that perhaps similar complex features in organisms provide evidence of
different evolutionary strains?
4. Flaws in Paper
Given that the strength of the authors' thesis rests strongly on their
analogy between the evolution of digital organisms and the evolution of
biological organisms, I would have liked to see more background on their
methodology (e.g. how did they come up with their evolutionary model? What
choices influenced the specific parameters chosen for evolution? What are
some of the consequences of the simplifications in their model, for example,
asexual vs. sexual reproduction).
The authors did a good job of highlighting the consistencies between
Darwin's hypothesis (that complex features generally evolve by modifying
existing structure) and their results in deriving the EQU function;
however, I thought that their other primary finding (that complex functions
evolve with a high probability) deserved further explication. Particularly,
is there evidence from the biological sciences that this sort of
evolutionary behavior exists in biological organisms as well?
5. Three Important Open Research Questions on Topic
(i) The authors invoked a quote by Daniel Dennett that establishes a
tractable system of evolution: ".evolution will occur whenever and wherever
three conditions are met: replication, variation (mutation), and
differential fitness (competition)." What other non-critical variables
influence evolution? Could other systems of evolution be applied? If so, how
does this impact the results of their experiment?
(ii) How does one measure the precision of the "digital organism"
evolutionary model to its biological counterpart? What sort of parameters
are necessary to build an analogous computational system? Answers to these
questions will allow us to build better virtual models of evolution and,
perhaps, better understand evolution itself.
(iii) It is clear that digital organisms have the potential to provide
biologists with an extremely rich and efficient method of experimentation
particularly in those areas which, ordinarily, would require large amounts
of resources to investigate in the real world. What other fields of biology
could benefit from the application of digital organisms?
6. Is this problem and solution just another instance of search? What, if
anything, makes this particular search problem different from the usual
search problem?
It is an interesting exercise to apply the perspective of search to
evolution - could evolution simply be described as "searching" for the most
rewarding mutation (albeit over an extremely long timeframe), a kind of
greedy local search. In this sense, the search space is all possible
mutations from a given genetic makeup. Each mutation has a reward assignment
which could be positive or negative. A positive reward would result from
acquiring a new feature while a negative reward would result from losing or
debilitating a current feature. The challenge in this search space is that
we don't necessarily know the reward outright; the reward is a reflection of
how much the mutation allows the organism to thrive in the world.
This archive was generated by hypermail 2.1.6 : Wed Dec 01 2004 - 10:44:55 PST