Ben Kerr - Adaptive landscapes in changing environments
in Changing Environments
BEACON Congress 2013
Ben Kerr (UW)
“It may metaphorically be said that natural selection is daily and
hourly scrutinising, throughout the world, the slightest variations;
rejecting those that are bad, preserving and adding up all that are
good; silently and insensibly working, whenever and wherever
opportunity offers, at the improvement of each organic being in
relation to its organic and inorganic conditions of life.”
Darwin, The Origin of Species
aB ab AB Ab
• Darwin’s “opportunity”
depends on the variants
present in the
• The genotype to
phenotype (G→P) map
carves out the possible.
• The genotype to phenotype to fitness (G→P→F)
map gives information on evolutionary
accessibility and evolutionary constraints.
Mapping out Adaptation
beak depthaB ab AB Ab
Picking the Wright Metaphor
• In 1932, Sewall Wright was invited to give a non-
technical talk on his view of evolution at the sixth
International Congress of Genetics.
• Wright (1932) started with a simple idea: a map
from genotype to fitness, where “the entire field of
possible gene combinations [could] be graded with
respect to adaptive value.”
• Thus, a genotype-to-fitness (G→F) map and
specification of how genotypes are connected
defines an adaptive landscape.
Figure 2 from Wright (1932)
Evolution in the Balance
• Wright felt that the landscape was likely rugged; the
problem that occupied him was how a population
could move from a lower peak to a higher peak
• His shifting balance theory rests on two assumptions:
1. Epistasis leading to distinct “peaks” (rugged landscape)
2. The population is structured (as semi-isolated sparsely
• Wright’s shifting balance invokes several processes
(mutation, selection, drift, and migration):
Phase 1: Demes drift over the adaptive landscape
Phase 2: Selection drives demes to new peaks
Phase 3: Competition between demes where the most
fit pulls the metapopulation to its peak
• Despite the importance of Wright’s shifting balance
to evolution in natural populations, the
accompanying metaphor has been very popular.
Selectively Accessible Paths (to SSWiM Upstream)
• Maynard Smith imagined mutations as
steps on a walk through sequence space.
• Think of a game in which one word is
changed to another via single letter
changes and intermediates are words.
• A path (multiple mutational steps) is
selectively accessible if each step increases
fitness (takes you uphill).
• Epistasis occurs when the effect of a
mutation changes given different
background contexts in either:
• Simple walks along selectively accessible
paths occur when:
- Selection is strong
- Mutation is weak
• In “SSWM” conditions, the population can
be thought of as a point moving through a
directed network, where the topology is
dictated by the adaptive landscape.
WORD WORE GORE GONE GENE
(necessary for ruggedness)
WORDS FROM BEACON PROPOSALS
Burch Chao Weinreich
Changing Environments: Changing Metaphors?
• The standard incarnation of the metaphor does not emphasize the role of the
environment and the role of the evolving organism on the environment.
• Partly in response to Wright’s shifting
balance process, Fisher argued that
environmental change could lead to
movement from a former peak.
• Wright also acknowledged that
environmental change could move
populations in genotype space
1950’s NEW ZEALAND
Evolution in Changing and Changed Environments
• When the landscape is
rugged (e.g., due to
genetic epistasis), the
population can become
trapped on a sub-optimal
• A changing environment
can move a population to
a new peak (perhaps a
• An environment changed
by organisms can also
move a population to
• Coevolution or niche
construction can have a
Adapting the Adaptive Landscape
ab AbaB ABAB
• How might we visualize evolution under SSWM
conditions, but where the environment changes
• The landscapes
can be simplified
• The networks
can be stacked
in an ordered
• Evolution can be
down the wall.
• Also, epistasis
can be read off
Network Approach: Application to Multi-Drug Resistance
• Let’s consider microbial
• When drugs
accessible paths can
• Cycles are thus
- Cycles have been found
(Goulart et al. 2013).
• Possible “evolutionary flow” under
multiple drugs is given by connections
among the strongly connected clusters in
the union of the separate networks.
- In the b-lactam data, many (but not all)
“sink” clusters are single genotypes.
• Accessible paths between two genotypes
under multiple drugs can be enumerated.
- Certain combinations of drugs exposure may
make multi-drug resistance more likely.
Number of Accessible Paths
Distribution of Sink Sizes
Network Approach: Application to Coevolution
• One of the most important
components of “the
• How might we represent
“the landscape” for a
coevolving pair of species?
• One approach could
involve consideration of an
simply the concatenation
of genotypes into a single
• Under SSWM
involves movement in
different dimensions and
epistasis (intra- and inter-)
can be gauged. 00 10
ISSUES TO DISCUSS
1. RESEARCH DIRECTIONS: What are the most exciting experimental directions to pursue in
exploring evolution in changing environments? Are there obvious evolution experiments
(with microbes, Avidians, etc.) that would contribute to deeper understanding? What are
the most pressing theoretical issues to address? What computational/engineering
contributions would have the highest impact? Does the form of environmental change
2. VISUALIZATION: What are the best ways to visualize evolution in changing environments?
Can/should the landscape metaphor be salvaged? Are network approaches valuable (and
should we be focusing on static topology or dynamic movement on networks)? How should
environments be visually represented? Does it matter if the environment is abiotic/biotic or
exogenous/endogenous? Does the best way to visualize adaptation in changing worlds
depend on assumptions (e.g., SSWM) of the evolutionary process, and how might it change
with different assumptions?
3. BROADER APPLICATIONS: What are the most pressing applications of these topics? Climate
change? Antibiotic resistance? Are landscape metaphors (or network approaches) helpful
in addressing these issues? What is currently most necessary for understanding evolution in
changing environments in a way that has relevance for these broader applications?
4. WRITE-IN: Intersection between evolution and multi-modal optimization problems…