Initial evolution (forwarded)

[Martin Taylor 950220 10:50]

The following is forwarded from the autopoesis mailing list. It seems
relevant to the now-and-again discussion we have about the origin of control,
though it doesn't mention control.


From: (A. Chu)

There are a few steps we know something about on this ladder of increasing

1. catalytic cycles, autocatalysis: a sequence of chemical reactions
  which have products which catalyze the next one.
2. Hypercycles: a cycle of catalytic cycles linked again by catalysts.
3. Cevolution, selection of these cycles.

<big unknown>

n. simple cells
and onward

The scenario you mention above is proposed by Eigen and colleagues, and
falls into the class of theories which can be called "RNA-first" theories
(the consensus seems to be that DNA were a rather late addition - serving
as templates for the RNA strands that really started the whole thing.)
There are also "proteins-first" theories which maintain that amino-acid
chains were the first - they have much more impressive catalytic properties.

The problem with both kinds of theories is in the explanation of how
RNA and proteins learned to work together - the question of coding. Eigen have put some work into this area, but their results are IMHO not
convincing. Kauffman's proposal seems much more plausible I think.

Briefly, he points out that in a random soup of catalysts some compounds
will catalyze the formation of others, with some very tiny probability of
any particular compound catalyzing some particular reaction (about on the
order of 10^-12, if I recall correctly). However, as polymer length increases,
the number of catalyzed reactions grows faster than the number of compounds.
In graph terms: the number of arcs grows faster than the number of nodes.
Thus, at some level of complexity (polymer length), closed catalytic loops
will inevitably form. These loops, or autocatalytic sets, will at first be
extremely bushy, but competition for resources will drive them toward higher
efficiency. Thus arcs may get pruned, favoring those that are more
One can imagine cross-catalytic relationships (like RNA) to be particularly
efficient and thus selected. Also, a mapping between RNA and proteins can
crystallize from a non-specific code to a specific one as a breaking of

So his theme is that a high degree of order emerges spontaneously in a complex
web of catalysis, and then gets pruned toward the RNA and protein system we
observe today. The genetic code falls out nicely out of this, as a frozen
accident. (Those who've read Kauffman - please correct me if I'm
mis-representing or leaving something out).

Anyway, the implication of this for autopoiesis is deep - autocatalytic sets
are autopoietic. This further suggests that autopoiesis is _older_ than
reproduction or even proper cells - since the autocatalytic system presumably
did not need to develop a cell wall or DNA or mitosis in order to function.
So some light falls on the question of reproduction and autopoiesis - they
need not come as a package. It also suggests that autopoiesis does
not require cell membranes or other physical delineations of organisms.


(Why did M & V insist on a physical boundary delineating an autopoietic
entity from its environment? Was it done for some functional reason
or merely because such a distinction makes it easier for us to think
about the concept?)

----- End Forwarded Message -----