Biology Reference
In-Depth Information
At multiple levels, Wagner (2005) found recurrent patterns, whether it be for
the conformation of an RNA or a protein, the generation of a crucial product,
or the production or maintenance of a required morphology confirming this
assumption of robustness: The natural system was more robust under neighboring
perturbations, whether genetic, structural, or dynamic, than for values of that
system drawn at random from possible systems like it. This could be for different
reasons, and Wagner investigates their plausibility and scope extensively at
various levels of organization.
12
One of his conclusions was that robustness to
environmental fluctuations probably was the source of selection that conferred
robustness to the effect of mutations as a secondary effect.
13
This is interesting:
The kind of results systems biologists can get directly is relevant to evolutionary
questions. More generally, one must consider that:
(1) A system property might be robust because state-space neighborhoods where
a property is robust are easier to find in an evolutionary search.
(2) Once found, if the property is selectively advantageous, it is easier to main-
tain in the face of mutation and environmental perturbation if small induced
changes in state leave the property relatively unchanged.
14
(3) Structural changes in the system may change the character of neighboring
state-spaces and, in particular, may act to increase the size of what Wagner
calls the 'neutral space' in which the relevant property remains unchanged.
Something like this must be going on in what Waddington (1957) called
'genetic assimilation', in which selection changes the expression of a prop-
erty manifested only in the presence of an environmental shock so that it is
manifested under a much wider range of conditions.
(4) The probability that something will be entrenched, that other parts of an
accumulating adaptive structure should come to depend upon it, should be
a monotonically increasing function of its stability and persistence. The
nonlinear amplifications of selection found by Wimsatt and Schank (2004)
12
In his discussion, Wagner focuses on the first two, though I believe that all of them come up in passing
elsewhere in his discussion.
13
My one reservation about this claim is that it turns on the fact that environmental perturbations are much
more common than mutations. But if one counts recombination in a system with lots of epistatic effects
as producing new mutations (as they likely will at the phenotypic level), the number of mutations goes up
enormously - by orders of magnitude.
14
Kauffman (1969, 1985, 1993) commonly assumed that the given circuits would be realized in only one
specific state - that any deviations led to reduced fitness. This made them highly sensitive to mutation and was
the major reason why his results seemed to establish that selection could not maintain large complex systems.
The larger is the 'neutral neighborhood' for a property, the more easily it is maintained by selection. We called
this 'degree of genericity' when we argued that selection and self-organization would work most effectively
in concert, when the selected state was multiply realizable, and thus not radically improbable (Wimsatt, 1986,
Schank and Wimsatt, 1988). Given the sizeable neutral neighborhoods for features found at many levels of
organization in evolved systems (Wagner, 2005), Kauffman's claims were too pessimistic.
