Gavrilets, S., and J.Gravner. 1997. ``Percolation on the fitness hypercube and the evolution of reproductive isolation.''
*Journal of Theoretical Biology* **184**: 51-64.
** ABSTRACT **

We study the structure and properties of adaptive landscapes arising from
the assumption that genotype fitness can only be 0 (inviable genotype) or
1 (viable genotype). An appropriate image of resulting (``holey'')
fitness landscapes
is a (multidimensional) flat surface with many holes. We have demonstrated that
in the genotype space there are clusters of viable
genotypes whose members can evolve from any member by single mutations
and drift and that there are ``species'' defined according to the biological
species concept. Assuming that the number of genes is very large
while the proportion of ``good'' combinations of genes is very small, we
have deduced many important qualitative and quantitative properties of
holey adaptive landscapes which may be related to the patterns of speciation.
Relationship between the proportion of viable genotypes
among all possible genotypes and the number of loci, $n$, determines
two qualitatively different regimes: subcritical and supercritical.
The subcritical regime takes place if the proportion of viable genotypes
is extremely small. In this case, the largest clusters of viable genotypes
in the genotype space have size of order
$n$ and there are many of such size; typical members of
a cluster are connected by a single (``evolutionary'') path; the number
of different (biological)
species in the cluster has order $n$; the expected number
of different species in the cluster within $k$ viable substitutions
from any its member is of order $k$.
The supercritical regime takes place if the proportion of viable genotypes
is small but not extremely small. In this case, there exists
a cluster of viable genotypes (a ``giant'' component) that has size of order $2^n/n$; the giant component comes ``near'' every point of the genotype space;
typical members of the giant component are connected by many evolutionary
paths; the number of different (biological) species on the ``giant''
component has at least order $n^2$; the expected number
of different species on the ``giant'' component within $k$ viable
substitution from any its member is at least of order $k n$.
At the boundary of two regimes all properties of adaptive landscapes
undergo dramatic changes, a physical analogy of which is a phase transition.
We have considered the most probable (within the present framework)
scenario of biological evolution on holey landscapes assuming that
it starts on a genotype from the largest connected component and proceeds
along it by mutation and genetic drift. In this scenario, there
is no need to cross any ``adaptive valleys''; reproductive
isolation between populations evolves as a side effect of accumulating
different mutations. We have shown that the rate of divergence is very
fast: a few substitutions are sufficient to result in a new biological
species. Our results also provide a ``genetic'' explanation for
hump-shaped patters in relationships between species diversity and
quality of the environment, a formal justification of the idea of
``extra-dimensional bypass'' on adaptive landscapes and some additional
information about uncorrelated ``rugged'' landscapes.