Biology Reference
In-Depth Information
evolve was the miserable failure to produce functioning computer programs by the pro-
cesses of mutation and selection of code, which randomized their behavior rather than
improved them (for an overview of this work and its relationship to evolutionary biology,
see
Wagner and Altenberg, 1996
). Not surprisingly, research in evolutionary computing
turned to the question of what could enable programs to evolve. This question, in turn,
prompted questions about what enables organisms to generate selectively useful variation?
Modularity is now regarded as one of the key attributes of evolvable systems because it
makes it possible to improve one part (of both computer code and morphology) without
interfering with already optimized parts. The integration of adaptively interdependent
traits within modules, and the (quasi)-autonomy of individual modules, enables one func-
tional complex to evolve when others are under stabilizing selection. This theory is the
basis for one definition of “modularity”, one which incorporates the idea of “selectively
useful” variation into the definition of modularity itself. According to this definition, mod-
ules comprise traits that collectively serve a primary function, with different modules serv-
ing different primary functions (
Wagner, 1996
). Each complex is internally integrated due
to the same genes affecting multiple traits within the complex and the complexes are
genetically independent, or nearly so. This definition of modularity is represented by a
classic diagram (
Figure 12.11
; after
Wagner, 1996; Wagner and Altenberg, 1996
) which
shows two functions, Function 1 and Function 2, each served by multiple traits (T1
T7),
with each trait being affected by many genes (G1
G6), most of which affect more than
one trait. According to this diagram (and to the theory it represents), a gene typically
affects two or more traits within a single module, with few genes affecting traits within
different modules. This diagram presents a sharp contrast to one long-standing view of
genetic architecture
universal pleiotropy. The idea of universal pleiotropy raised ques-
tions about the causes of uncorrelated traits; the explanation for the lack of a correlation is
that positive and negative pleiotropic effects cancel out, i.e. “antagonistic pleiotropy”, and
it is one contrast to the theory portrayed by the diagram: independent traits are indepen-
dent because pleiotropic effects are restricted to subsets of traits. As emphasized by
Mezey
FIGURE 12.11
The classic depiction of modularity (after
Wagner, 1996; Wagner and Altenberg, 1996
) as the restric-
tion of pleiotropic effects to complexes of traits serving the
same primary function. Shown are two functional com-
plexes, Function 1 served by traits T1
T4, and Function 2,
served by traits T5
T7. The traits serving Function 1 are
affected by genes G1
G3, all of which affect multiple traits
within the complex and, with the exception of G3 that
affects T6 belonging to the complex serving Function 2, all
the affects of G1
G3 are restricted to the traits serving
Function 1. Similarly,
the traits serving Function 2 are
affected by genes G4
G6, all of which affect multiple traits
within the complex and, with the exception of G4 that
affects T4 belonging to the complex serving Function 1, all
the affects of G4
G6 are restricted to the traits serving
Function 2.
