Biology Reference
In-Depth Information
suggest that this probability increases nonlinearly as well. Selection should
tend to entrench robust or generic features, though one would also expect
a steady accumulation of contingent, arbitrary, improbable features that
become sufficiently entrenched to persist and become phylogenetically dis-
tinguishing features.
A concept used several places above, which should become very useful to NSB
and is elaborated to great advantage by Wagner, is that of a 'neutral space'.
15
This is a further abstraction and generalization of the idea of 'neutral percolation
surface' introduced by Huynen, Stadler, and Fontana in their important 1996
paper on evolution in RNA configuration space. Huynen et al. looked at the
major forms of folded configurations of RNA molecules of length 100 nucleic
acid bases and found that a few major forms dominated. They also found that
the regions in which specific major forms appeared tended to be connected
in mutation space, such that one could often move, one mutation at a time,
throughout a connected neighborhood without changing the folded configuration,
and thus, to a first approximation, remaining 'neutral', preserving the function or
the fitness of the molecule. This meant that a population could 'percolate', one
mutation at a time, to distant parts of the space. They also found that most major
forms were reachable within a small number of mutations from one another.
These 'neutral spaces' thus provided 'don't care' conditions for the composition
and behavior of systems using these molecules, but the diversity of 'neutral'
positions they could occupy could lead to rapid divergence if external conditions
changed and selection for different configurations became advantageous. This
idea itself is clearly related to the concepts of a fitness topography and to
an energy surface, but is usefully exploited here as applied to discrete state
systems.
Wagner's extensive discussions show that this situation or others analogous
to it (as he points out, one cannot always formulate problems such that a well-
organized discrete space can be defined for them) is characteristic not only of
RNA space, but also at other levels of organization as well, in particular, to
protein space and to spaces characterizing the dynamics of metabolic systems.
This generally means that organic systems are designed so that they have many
'don't care' conditions - that behavior may often not need to be fully specified
to be able to predict the dynamics with reasonable confidence. This is not
15
Wagner's topic is also distinguished by the robustness of his review of the evidence. Most claims are
addressed using two or more distinct modeling strategies, often several, with the limitations of the different
strategies compared. Some of the modeling concepts, for example, the idea of a 'lattice model' of a protein,
in which all amino acids are of the same size and separation and their bonds can only take up angles at 90
intervals (0, 180, 270 in two dimensions, and the analogous lattice angles in three dimensions), are lovely and
revealing. This one, for example, explores the consequences of topology (one-dimensional connectivity) and
the distribution of hydrophobic and hydrophilic sites for the folding configurations.
