My presentation from 8th May 2012, at a workshop on Plant-Microbe Interactions, held at the Turin Botanical Gardens, University of Turin. The talk expands on concepts from this paper: Pritchard L, Birch P (2011) A systems biology perspective on plant-microbe interactions: Biochemical and structural targets of pathogen effectors. Plant Science 180: 584–603. doi:10.1016/j.plantsci.2010.12.008.
Organic Name Reactions for the students and aspirants of Chemistry12th.pptx
A Systems Biology Perspective on Plant-Pathogen Interactions 2012-05-08, Turin
1.
2.
3. A
Systems
Biology
Perspec2ve
on
Plant-‐Pathogen
Interac2ons
Leighton
Pritchard
4. A
Con2nuum
l Pathogenicity
is
a
loaded
term:
l o4en
reflects
human
interest
in
the
system
l disease
on
crop
plants
could
be
coincidental
to
‘wild
type’
interac<ons
l A
con<nuum
of
interac<on
modes,
including
symbiosis
and
pathogenicity
l The
loca<on
of
the
system
on
this
con<nuum
may
depend
on
context
l e.g.
Pectobacterium
atrosep/cum:potato
no
impact
host
death
5. A
basic
observa2on
Pathogen
Host
Biological
cells
(and
organisms)
can
be
represented
as
networks
6. Biological
networks
l Common
way
to
represent
structure
l Several
biological
subsystems
are
networks
l Universal
representa<on
l All
biological
systems
have
parts
that
can
be
represented
as
networks
l Networks
(a.k.a.
graphs)
are
mathema<cally
well-‐
understood:
Graph
Theory
l Many
tools
exist,
relevant
to
biology
7. Biological
networks
l Common
way
to
represent
structure
l Several
biological
subsystems
are
networks
l Universal
representa<on
l All
biological
systems
have
parts
that
can
be
represented
as
networks
l Networks
(a.k.a.
graphs)
are
mathema<cally
well-‐
understood:
Graph
Theory
l Many
tools
exist,
relevant
to
biology
8. Biological
networks
l Metabolic
networks
(e.g.
KEGG)
(generic)
Michal
(Ed.),
Biochemical
Pathways,
John
Wiley
and
Sons,
New
York,
1999.
11. Biological
networks
l Common
way
to
represent
structure
l Several
biological
subsystems
are
networks
l Universal
representa<on
l All
biological
systems
have
parts
that
can
be
represented
as
networks
l Networks
(a.k.a.
graphs)
are
mathema<cally
well-‐
understood:
Graph
Theory
l Many
tools
exist,
relevant
to
biology
12. What
is
a
network?
l Networks
have
nodes
(a.k.a.
ver<ces)
l Nodes
typically
represent
‘things’:
„ proteins,
chemical
compounds,
people,
towns,
junc<ons…
l Nodes
are
connected
by
edges
(a.k.a.
arcs)
l Edges
typically
indicate
some
rela<onship
between
nodes
„ physical
interac<on,
substrate:product,
friends
on
Facebook
l Edges
may
be
directed
(from
one
node
to
another)
or
undirected
(no
or
ambiguous
direc<on)
„ chemical
conversion:
directed;
interac<on:
undirected
n1
n2
13. What
is
a
network?
l Networks
have
nodes
(a.k.a.
ver<ces)
l Nodes
typically
represent
‘things’:
„ proteins,
chemical
compounds,
people,
towns,
junc<ons…
l Nodes
are
connected
by
edges
(a.k.a.
arcs)
l Edges
typically
indicate
some
rela<onship
between
nodes
„ physical
interac<on,
substrate:product,
friends
on
Facebook
l Edges
may
be
directed
(from
one
node
to
another)
or
undirected
(no
or
ambiguous
direc<on)
„ chemical
conversion:
directed;
interac<on:
undirected
n1
n2
14. What
is
a
network?
l Networks
have
nodes
(a.k.a.
ver<ces)
l Nodes
typically
represent
‘things’:
„ proteins,
chemical
compounds,
people,
towns,
junc<ons…
l Nodes
are
connected
by
edges
(a.k.a.
arcs)
l Edges
typically
indicate
some
rela<onship
between
nodes
„ physical
interac<on,
substrate:product,
friends
on
Facebook
l Edges
may
be
directed
(from
one
node
to
another)
or
undirected
(no
or
ambiguous
direc<on)
„ chemical
conversion:
directed;
interac<on:
undirected
n1
n2
n1
n2
n1
n2
n1
n2
15. Many
things
are
networks
l My
Facebook
friends
network:
l Nodes:
people
l Edges:
friendships
between
people
l Useful
concepts
for
biology:
l ‘friend
of
a
friend’;
‘six
degrees
of
separa<on’;
clusters
of
friends
Solange Mateo Montalcini
Maeve Price
Peter Cock
Catherine Tackley
Gavin Cowie
Steffi Keir
Yvonne McAvoy
Jennifer White
Rachel Clewes
Juan Morales
Karen Faulds
David Ian Ellis
Laura Banasiak
Andrea Semião
Daniel Tackley
Andrew Lipscombe
Bleddyn Hughes
Sue Stovell
Laura Didymus
Hywel Griffiths
Charles Twist
Christian Payne
Helen Johnson
Phil Parsonage
Colin McGill
Allan N. Gunn
Will Allwood
Katherine Hollywood
Judith Robertson
Andrew Murdoch
David Broadhurst
Lydia Castelli
Miles Armstrong
Paul Keir
Fiona White Gagg
Lizzie Wilberforce
Joanne Fitchet
Laura Baxter
Alison Gilhespie
Jorunn Bos
James Gagg
Andy Smith
Clare Baxter
Susan Somerville
Neil Bhaduri
Joanna Jones
Colleen Gagg
Susan Quinn McGhee
Al Macmillan
Norman StewartKevin Knox
Susan BreenMichael Barrow
Phil Dennison
Andrew McKenzie
Matthew Blackburn
Christelle Robert
Tim Arrowsmith
Emma Robertson
Jane Ballany
Chris Thorpe
Andrew Dalke
Sonia Humphris
Juan Morales
Eleanor Gilroy
Chris McDonald
Natalie Homer
Anna Åsman
Ruth Polwart
Tim Morley
Kenny Duncan
Iddo Friedberg
Remco Stam
Ramesh Vetukuri
Louise Matheson
Simon Easterman
Philip Law
Craig Shaddy Shadbolt
Simon Garrett
Agata Kaczmarek
Simon Pendlebury
Rays Jiang
Christiane AusJena
Pedro Mendes
Iris Stone
Ingo Hein
Adriana Ravagnani
Eduard Venter
Charles Gordon
David Cooke
Jonathan Gagg
Roger Jarvis
Ross McMahon
Stefan Engelhardt
Edgar Huitema
Thomas Pritchard
Tracy Canham
Sophien Kamoun
Florietta Jupe
Ambreen Owen
Hazel McLellan
16. Many
things
are
networks
l My
Facebook
friends
network:
l Nodes:
people
l Edges:
friendships
between
people
l Useful
concepts
for
biology:
l ‘friend
of
a
friend’;
‘six
degrees
of
separa<on’;
clusters
of
friends
Solange Mateo Montalcini
Maeve Price
Peter Cock
Catherine Tackley
Gavin Cowie
Steffi Keir
Yvonne McAvoy
Jennifer White
Rachel Clewes
Juan Morales
Karen Faulds
David Ian Ellis
Laura Banasiak
Andrea Semião
Daniel Tackley
Andrew Lipscombe
Bleddyn Hughes
Sue Stovell
Laura Didymus
Hywel Griffiths
Charles Twist
Christian Payne
Helen Johnson
Phil Parsonage
Colin McGill
Allan N. Gunn
Will Allwood
Katherine Hollywood
Judith Robertson
Andrew Murdoch
David Broadhurst
Lydia Castelli
Miles Armstrong
Paul Keir
Fiona White Gagg
Lizzie Wilberforce
Joanne Fitchet
Laura Baxter
Alison Gilhespie
Jorunn Bos
James Gagg
Andy Smith
Clare Baxter
Susan Somerville
Neil Bhaduri
Joanna Jones
Colleen Gagg
Susan Quinn McGhee
Al Macmillan
Norman StewartKevin Knox
Susan BreenMichael Barrow
Phil Dennison
Andrew McKenzie
Matthew Blackburn
Christelle Robert
Tim Arrowsmith
Emma Robertson
Jane Ballany
Chris Thorpe
Andrew Dalke
Sonia Humphris
Juan Morales
Eleanor Gilroy
Chris McDonald
Natalie Homer
Anna Åsman
Ruth Polwart
Tim Morley
Kenny Duncan
Iddo Friedberg
Remco Stam
Ramesh Vetukuri
Louise Matheson
Simon Easterman
Philip Law
Craig Shaddy Shadbolt
Simon Garrett
Agata Kaczmarek
Simon Pendlebury
Rays Jiang
Christiane AusJena
Pedro Mendes
Iris Stone
Ingo Hein
Adriana Ravagnani
Eduard Venter
Charles Gordon
David Cooke
Jonathan Gagg
Roger Jarvis
Ross McMahon
Stefan Engelhardt
Edgar Huitema
Thomas Pritchard
Tracy Canham
Sophien Kamoun
Florietta Jupe
Ambreen Owen
Hazel McLellan
17. Many
things
are
networks
l Google
Maps
l Nodes:
road
junc<ons
(and
end
points
in
culs
de
sacs)
l Edges:
roads
l Structure
view
l Flow/traffic
view
l Useful
concepts
for
biology:
l Network
‘flow’
or
‘flux’;
distance
on
a
network;
shortest
path
18. Many
things
are
networks
l Google
Maps
l Nodes:
road
junc<ons
(and
end
points
in
culs
de
sacs)
l Edges:
roads
l Structure
view
l Flow/traffic
view
l Useful
concepts
for
biology:
l Network
‘flow’
or
‘flux’;
distance
on
a
network;
shortest
path
19. Many
things
are
networks
l Google
Maps
l Nodes:
road
junc<ons
(and
end
points
in
culs
de
sacs)
l Edges:
roads
l Structure
view
l Flow/traffic
view
l Useful
concepts
for
biology:
l Network
‘flow’
or
‘flux’;
distance
on
a
network;
shortest
path
20. Networks
are
abstract
l Networks
are
collec<ons
of
nodes
and
edges
l Proper<es
of
the
network
are
the
proper<es
of
that
collec<on
l What
a
node
or
edge
represents
is
not
important
l If
a
network
describes
biology
well…
l …what
is
true
about
the
network
will
be
true
about
the
biology
l (some
networks
describe
biology
be`er
than
others)
l Abstract
truths
about
networks
can
be
true
about
biology
l If
a
network
of
type
X
is
robust
to
random
damage,
and
a
biological
network
is
of
type
X,
we
can
say
that
the
biological
network
is
robust
to
random
damage.
21. Networks
are
abstract
l Networks
are
collec<ons
of
nodes
and
edges
l Proper<es
of
the
network
are
the
proper<es
of
that
collec<on
l What
a
node
or
edge
represents
is
not
important
l If
a
network
describes
biology
well…
l …what
is
true
about
the
network
will
be
true
about
the
biology
l (some
networks
describe
biology
be`er
than
others)
l Abstract
truths
about
networks
can
be
true
about
biology
l If
a
network
of
type
X
is
robust
to
random
damage,
and
a
biological
network
is
of
type
X,
we
can
say
that
the
biological
network
is
robust
to
random
damage.
24. Networks
are
abstract
l Networks
are
collec<ons
of
nodes
and
edges
l Proper<es
of
the
network
are
the
proper<es
of
that
collec<on
l What
a
node
or
edge
represents
is
not
important
l If
a
network
describes
biology
well…
l …what
is
true
about
the
network
will
be
true
about
the
biology
l (some
networks
describe
biology
be`er
than
others)
l Abstract
truths
about
networks
can
be
true
about
biology
l If
a
network
of
type
X
is
robust
to
random
damage,
and
a
biological
network
is
of
type
X,
we
can
say
that
the
biological
network
is
robust
to
random
damage.
25. Networks
are
abstract
l Networks
are
collec<ons
of
nodes
and
edges
l Proper<es
of
the
network
are
the
proper<es
of
that
collec<on
l What
a
node
or
edge
represents
is
not
important
l If
a
network
describes
biology
well…
l …what
is
true
about
the
network
will
be
true
about
the
biology
l (some
networks
describe
biology
be`er
than
others)
26. Networks
are
abstract
l Networks
are
collec<ons
of
nodes
and
edges
l Proper<es
of
the
network
are
the
proper<es
of
that
collec<on
l What
a
node
or
edge
represents
is
not
important
l If
a
network
describes
biology
well…
l …what
is
true
about
the
network
will
be
true
about
the
biology
l (some
networks
describe
biology
be`er
than
others)
n1
n2
n3
n4
n5
27. Networks
are
abstract
l Networks
are
collec<ons
of
nodes
and
edges
l Proper<es
of
the
network
are
the
proper<es
of
that
collec<on
l What
a
node
or
edge
represents
is
not
important
l If
a
network
describes
biology
well…
l …what
is
true
about
the
network
will
be
true
about
the
biology
l (some
networks
describe
biology
be`er
than
others)
l Any
network
with
this
structure
has
the
same
behaviour
l Behaviour
of
specific
regulatory
network
is
dictated
by
its
structure:
l Behaviour
dependent
on
structure
of
system
as
a
whole:
need
to
understand
this
at
a
systems
level
MacLean
and
Studholme.
A
Boolean
model
of
the
Pseudomonas
syringae
hrp
regulon
predicts
a
<ghtly
regulated
system.
PLoS
ONE
(2010)
vol.
5
(2)
pp.
e9101
doi:10.1371/journal.pone.0009101
28. Networks
are
abstract
l Networks
are
collec<ons
of
nodes
and
edges
l Proper<es
of
the
network
are
the
proper<es
of
that
collec<on
l What
a
node
or
edge
represents
is
not
important
l If
a
network
describes
biology
well…
l …what
is
true
about
the
network
will
be
true
about
the
biology
l (some
networks
describe
biology
be`er
than
others)
l Abstract
truths
about
networks
can
be
true
about
the
biology
they
represent
l If
a
network
of
type
X
is
robust
to
random
damage,
and
a
biological
network
is
of
type
X,
we
can
say
that
the
biological
network
is
robust
to
random
damage.
29. Choosing
a
representa2on
l Network
should
be
an
adequate
representa<on
of
biology
l Choice
of
representa<on
should
suit
biological
ques<on
l e.g.
do
we
represent
chemical
compounds,
or
moie<es?
30. Choosing
a
representa2on
l Network
should
be
an
adequate
representa<on
of
biology
l Choice
of
representa<on
should
suit
biological
ques<on
l e.g.
do
we
represent
chemical
compounds,
or
moie<es?
31. Choosing
a
representa2on
l Network
should
be
an
adequate
representa<on
of
biology
l Choice
of
representa<on
should
suit
biological
ques<on
l e.g.
do
we
represent
chemical
compounds,
or
moie<es?
32. Choosing
a
representa2on
l What
does
this
diagram
mean?
l Are
all
enzymes
expressed
at
same
<me?
l Are
all
enzymes
expressed
in
all
<ssues?
l Are
all
metabolites
always
available?
l 30-‐40%
of
metabolic
ac<vity
has
no
known
gene
associated
with
it
(Chen
and
Vitkup.
Distribu<on
of
orphan
metabolic
ac<vi<es.
Trends
Biotechnol
(2007)
vol.
25
(8)
pp.
343-‐348
doi:
10.1016/j.<btech.2007.06.001)
Michal
(Ed.),
Biochemical
Pathways,
John
Wiley
and
Sons,
New
York,
1999.
33. Choosing
a
representa2on
l What
does
this
diagram
mean?
l Are
all
enzymes
expressed
at
same
<me?
l Are
all
enzymes
expressed
in
all
<ssues?
l Are
all
metabolites
always
available?
l 30-‐40%
of
metabolic
ac<vity
has
no
known
gene
associated
with
it
(Chen
and
Vitkup.
Distribu<on
of
orphan
metabolic
ac<vi<es.
Trends
Biotechnol
(2007)
vol.
25
(8)
pp.
343-‐348
doi:
10.1016/j.<btech.2007.06.001)
Michal
(Ed.),
Biochemical
Pathways,
John
Wiley
and
Sons,
New
York,
1999.
34. Choosing
a
representa2on
l Biological
networks
are
dynamic
l There
may
be
homeostasis,
but
it’s
dynamic
homeostasis
l “The
only
steady-‐state
is
death”
l What
kind
of
dynamics?
l Kine<c
equa<ons
l ODE/Stochas<c
representa<on
of
processes
„ e.g.
enzyme
kine<cs
E + S ⌦ ES ⌦ EP ! E + P
v =
[S]Vmax
[S] + [Km]
35. Choosing
a
representa2on
l Biological
networks
are
dynamic
l There
may
be
homeostasis,
but
it’s
dynamic
homeostasis
l “The
only
steady-‐state
is
death”
l What
kind
of
dynamics?
l Kine<c
equa<ons
l ODE/Stochas<c
representa<on
of
processes
„ e.g.
enzyme
kine<cs
E + S ⌦ ES ⌦ EP ! E + P
v =
[S]Vmax
[S] + [Km]
36. Choosing
a
representa2on
l Biological
networks
are
dynamic
l There
may
be
homeostasis,
but
it’s
dynamic
homeostasis
l “The
only
steady-‐state
is
death”
l What
kind
of
dynamics?
l Boolean
(on/off,
0/1)
„ e.g.
regula<on/signalling
nodes
<me
43. Host-‐pathogen
interac2on
Pathogen
Host
Host-‐pathogen
interac2on
is
the
coming
together
of
two
networks
into
a
single
network:
different
proper2es
than
either
network
alone
44. Host-‐pathogen
interac2on
Pathogen
Host
How
does
this
affect
culturability?
Tight
connec2on
correlates
with
obligate
biotrophy,
hence
difficult
to
culture?
46. Host-‐pathogen
interac2on
Pathogen
Host
l How
does
host/pathogen
network
respond
to
interac<on?
l What
is
best
way
to
a`ack
a
network?
l What
is
best
way
to
defend
against
mul<ple
a`ack
strategies?
l Are
some
parts
of
a
network
predictably
more
influen<al
than
others?
47. Host-‐pathogen
interac2on
Pathogen
Host
l How
does
host/pathogen
network
respond
to
interac<on?
l What
is
best
way
to
a`ack
a
network?
l What
is
best
way
to
defend
against
mul<ple
a`ack
strategies?
l Are
some
parts
of
a
network
predictably
more
influen<al
than
others?
48. Host-‐pathogen
interac2on
Pathogen
Host
l How
does
host/pathogen
network
respond
to
interac<on?
l What
is
best
way
to
a`ack
a
network?
l What
is
best
way
to
defend
against
mul<ple
a`ack
strategies?
l Are
some
parts
of
a
network
predictably
more
influen<al
than
others?
49. Host-‐pathogen
interac2on
Pathogen
Host
l How
does
host/pathogen
network
respond
to
interac<on?
l What
is
best
way
to
a`ack
a
network?
l What
is
best
way
to
defend
against
mul<ple
a`ack
strategies?
l Are
some
parts
of
a
network
predictably
more
influen<al
than
others?
50. Influence
in
networks
l Efficient
a`ackers:
l cause
greatest
favourable
host
disrup<on
for
least
effort
l should
target
influen<al
points
in
host
network
l Efficient
defenders:
l protect
against
greatest
amount
of
poten<al
change
for
least
effort
l protect
against
most
commonly-‐targeted
points
in
network
l should
target
influen<al
points
in
host
network
l Greatest
benefit
for
least
cost
l What
are
the
most
influen<al
points
in
a
network?
51. Influence
in
networks
l Efficient
a`ackers:
l cause
greatest
favourable
host
disrup<on
for
least
effort
l should
target
influen<al
points
in
host
network
l Efficient
defenders:
l protect
against
greatest
amount
of
poten<al
change
for
least
effort
l protect
against
most
commonly-‐targeted
points
in
network
l should
target
influen<al
points
in
host
network
l Greatest
benefit
for
least
cost
l What
are
the
most
influen<al
points
in
a
network?
52. Influence
in
networks
l Efficient
a`ackers:
l cause
greatest
favourable
host
disrup<on
for
least
effort
l should
target
influen<al
points
in
host
network
l Efficient
defenders:
l protect
against
greatest
amount
of
poten<al
change
for
least
effort
l protect
against
most
commonly-‐targeted
points
in
network
l should
target
influen<al
points
in
host
network
l Greatest
benefit
for
least
cost
l What
are
the
most
influen<al
points
in
a
network?
53. Influence
in
networks
l Efficient
a`ackers:
l cause
greatest
favourable
host
disrup<on
for
least
effort
l should
target
influen<al
points
in
host
network
l Efficient
defenders:
l protect
against
greatest
amount
of
poten<al
change
for
least
effort
l protect
against
most
commonly-‐targeted
points
in
network
l should
target
influen<al
points
in
host
network
l Greatest
benefit
for
least
cost
l What
are
the
most
influen2al
points
in
a
network?
l can
we
predict/iden<fy
them?
54. Robustness
in
biological
networks
l Biological
networks
are
typically
robust
and
error-‐tolerant
l (necessary
for
descent
with
modifica<on)
l e.g.
only
17%
of
yeast
genes
essen<al
to
cell
viability
in
rich
media
Winzeler
et
al.
Func<onal
characteriza<on
of
the
S.
cerevisiae
genome
by
gene
dele<on
and
parallel
analysis.
Science
(1999)
vol.
285
(5429)
pp.
901-‐906
55. Robustness
in
biological
networks
l Biological
networks
are
typically
robust
and
error-‐tolerant
l (necessary
for
descent
with
modifica<on)
l e.g.
only
17%
of
yeast
genes
essen<al
to
cell
viability
in
rich
media
Winzeler
et
al.
Func<onal
characteriza<on
of
the
S.
cerevisiae
genome
by
gene
dele<on
and
parallel
analysis.
Science
(1999)
vol.
285
(5429)
pp.
901-‐906
56. Structural
robustness
in
biological
networks
l Some
network
structures
enhance
robustness
l Many
biological
networks
have
converged
to
same
network
structures
Barabási
and
Oltvai.
Network
biology:
understanding
the
cell's
func<onal
organiza<on.
Nat
Rev
Genet
(2004)
vol.
5
(2)
pp.
101-‐13
doi:
10.1038/nrg1272
Kitano.
Biological
robustness.
Nat
Rev
Genet
(2004)
vol.
5
(11)
pp.
826-‐37
doi:10.1038/nrg1471
• Aa:
random
Erdös-‐Renyi
graph:
not
robust
to
random
a`ack
(not
common
in
biology)
• Ba:
random
‘scale-‐free’
network:
robust
to
random
a`ack
(most
biological
networks)
• Ca:
hierarchical
network:
robust
to
random
a`ack
(many
signalling
networks)
57. l Some
network
structures
enhance
robustness
l Many
biological
networks
have
converged
to
same
network
structures
Barabási
and
Oltvai.
Network
biology:
understanding
the
cell's
func<onal
organiza<on.
Nat
Rev
Genet
(2004)
vol.
5
(2)
pp.
101-‐13
doi:
10.1038/nrg1272
Kitano.
Biological
robustness.
Nat
Rev
Genet
(2004)
vol.
5
(11)
pp.
826-‐37
doi:10.1038/nrg1471
• Aa:
random
Erdös-‐Renyi
graph:
not
robust
to
random
a`ack
(not
common
in
biology)
• Ba:
random
‘scale-‐free’
network:
robust
to
random
a`ack
(most
biological
networks)
• Ca:
hierarchical
network:
robust
to
random
a`ack
(many
signalling
networks)
Structural
robustness
in
biological
networks
58. l Network
bridges/bo`lenecks
l essen<al
intermediate
nodes
in
a
network
l dele<on
or
disrup<on
dissociates
(breaks)
the
network
Structural
robustness
in
biological
networks
• Pathways
from
detec<on
(e.g.
immune
recep<on)
to
host
response
• Signalling
pathways
• E.g.
Cladosporum
fulvum
Avr4
suppresses
produc<on
of
chi<n,
a
‘bridge’
59. l Network
bridges/bo`lenecks
l essen<al
intermediate
nodes
in
a
network
l dele<on
or
disrup<on
dissociates
(breaks)
the
network
Structural
robustness
in
biological
networks
MAMP
detec2on
• Pathways
from
detec<on
(e.g.
immune
recep<on)
to
host
response
• Signalling
pathways
• e.g.
Cladosporum
fulvum
Avr4
suppresses
produc<on
of
chi<n,
a
‘bridge’
60. l Network
bridges/bo`lenecks
l essen<al
intermediate
nodes
in
a
network
l dele<on
or
disrup<on
dissociates
(breaks)
the
network
Structural
robustness
in
biological
networks
MAMP
detec2on
• Pathways
from
detec<on
(e.g.
immune
recep<on)
to
host
response
• Signalling
pathways
• E.g.
Cladosporum
fulvum
Avr4
suppresses
produc<on
of
chi<n,
a
‘bridge’
chi<n
chi<nase
61. l Network
bridges/bo`lenecks
l essen<al
intermediate
nodes
in
a
network
l dele<on
or
disrup<on
dissociates
(breaks)
the
network
Structural
robustness
in
biological
networks
MAMP
detec2on
• Pathways
from
detec<on
(e.g.
immune
recep<on)
to
host
response
• Signalling
pathways
• E.g.
Cladosporum
fulvum
Avr4
suppresses
produc<on
of
chi<n,
a
‘bridge’
chi<n
chi<nase
Avr4
62. l Network
bridges/bo`lenecks
l essen<al
intermediate
nodes
in
a
network
l dele<on
or
disrup<on
dissociates
(breaks)
the
network
Structural
robustness
in
biological
networks
MAMP
detec2on
• Redundancy
and
cross-‐talk
in
signalling
pathways
protects
against
this
fragility
• e.g.
PTI/ETI
cross-‐talk
63. l Network
hubs
l highly-‐connected
nodes
l characteris<c
of
‘scale-‐free’
(and
similar)
networks
l dele<on
or
disrup<on
dissociates
(breaks)
the
network
Structural
robustness
in
biological
networks
• Why
do
hubs
occur?
• How
many
hubs
do
we
expect?
• How
are
they
related
to
biology?
64. l Network
hubs
l highly-‐connected
nodes
l characteris<c
of
‘scale-‐free’
(and
similar)
networks
l dele<on
or
disrup<on
dissociates
(breaks)
the
network
Structural
robustness
in
biological
networks
• Why
do
hubs
occur?
• How
many
hubs
do
we
expect?
• How
are
they
related
to
biology?
65. l Power-‐law
(a.k.a.
‘scale-‐free’)
networks
l Robust
because
of
node
degree
distribu<on
l Very
few
‘hubs’;
most
nodes
make
few
connec<ons
l Random
dele<on
more
likely
to
remove
node
with
few
connec<ons
Structural
robustness
in
biological
networks
Albert
et
al.
Error
and
a`ack
tolerance
of
complex
networks.
Nature
(2000)
vol.
406
(6794)
pp.
378-‐82
doi:
10.1038/35019019
66. l Power-‐law
(a.k.a.
‘scale-‐free’)
networks
l Robust
because
of
node
degree
distribu<on
l Very
few
‘hubs’;
most
nodes
make
few
connec<ons
l Random
dele<on
more
likely
to
remove
node
with
few
connec<ons
Structural
robustness
in
biological
networks
Albert
et
al.
Error
and
a`ack
tolerance
of
complex
networks.
Nature
(2000)
vol.
406
(6794)
pp.
378-‐82
doi:
10.1038/35019019
67. l Power-‐law
(a.k.a.
‘scale-‐free’)
networks
l Diagnos<c
‘degree
distribu<on’
(count
of
connec<ons
to
each
node)
l Yeast
protein
interac<on
network
has
power-‐law
distribu<on
l Essen<al
17%
of
genes
correlated
with
highly-‐connected
nodes
(hubs)
Structural
robustness
in
biological
networks
68. l Power-‐law
(a.k.a.
‘scale-‐free’)
networks
l Diagnos<c
‘degree
distribu<on’
(count
of
connec<ons
to
each
node)
l Yeast
protein
interac<on
network
has
power-‐law
distribu<on
l Essen<al
17%
of
genes
correlated
with
highly-‐connected
nodes
(hubs)
Structural
robustness
in
biological
networks
69. l Power-‐law
(a.k.a.
‘scale-‐free’)
networks
l Most
studied
biological
networks
are
‘scale-‐free’
l ‘Scale-‐free’
property
proposed
to
arise
from
network
evolu<on
l ‘older’
nodes
more
likely
to
be
hubs
l ‘older’
nodes
more
likely
to
be
func<onally-‐conserved,
sequence
constrained?
l Hubs
are
good
targets
for
network
disrup<on:
what
role
do
they
play
in
pathogen/host
evolu<on?
Structural
robustness
in
biological
networks
70. l Power-‐law
(a.k.a.
‘scale-‐free’)
networks
l Most
studied
biological
networks
are
‘scale-‐free’
l ‘Scale-‐free’
property
proposed
to
arise
from
network
evolu<on
l ‘older’
nodes
more
likely
to
be
hubs
l ‘older’
nodes
more
likely
to
be
func<onally-‐conserved,
sequence
constrained?
l Hubs
are
good
targets
for
network
disrup<on:
what
role
do
they
play
in
pathogen/host
evolu<on?
Structural
robustness
in
biological
networks
71. l Bacterial
Type
III
effectors
engage
a
limited
set
of
host
processes
across
host
kingdoms
e.g.:
l turnover
by
modula<on
of
ubiqui<na<on
l altera<on
of
transcrip<on
l altera<on
of
phosphoryla<on
l Strategies
such
as
the
targe<ng
of
ubiqui<na<on
are
used
by
bacterial
fungal
and
oomycete
pathogens
across
a
range
of
hosts
Structural
robustness
in
biological
networks
72. l Bacterial
Type
III
effectors
engage
a
limited
set
of
host
processes
across
host
kingdoms
e.g.:
l turnover
by
modula<on
of
ubiqui<na<on
l altera<on
of
transcrip<on
l altera<on
of
phosphoryla<on
l Strategies
such
as
the
targe<ng
of
ubiqui<na<on
are
used
by
bacterial
fungal
and
oomycete
pathogens
across
a
range
of
hosts
Structural
robustness
in
biological
networks
73. The
Guard
Hypothesis
l The
Guard
Hypothesis
describes
indirect
R
gene:effector
interac<on
l Direct
R
gene:effector
interac<on
could
lead
to
overwhelming
R
gene
load
l A.
thaliana
has
≈200
R
genes
(1%
of
gene
complement)
l If
‘hubs’
are
common
targets
for
pathogens…
l …guarding
the
hub
with
one
R
gene
is
more
efficient
than
gene-‐for-‐gene
interac<ons
l …network
topology
implies
the
Guard
Hypothesis
l If
‘hubs’
are
universal
targets…
l …network
topology
determines
which
nodes
are
likely
to
be
involved
in
host-‐pathogen
interac<on
Dangl
and
Jones.
Plant
pathogens
and
integrated
defence
responses
to
infec<on.
Nature
(2001)
vol.
411
(6839)
pp.
826-‐33
doi:10.1038/35081161
74. The
Guard
Hypothesis
l The
Guard
Hypothesis
describes
indirect
R
gene:effector
interac<on
l Direct
R
gene:effector
interac<on
could
lead
to
overwhelming
R
gene
load
l A.
thaliana
has
≈200
R
genes
(1%
of
gene
complement)
l If
‘hubs’
are
common
targets
for
pathogens…
l …guarding
the
hub
with
one
R
gene
is
more
efficient
than
gene-‐for-‐gene
interac<ons
l …network
topology
implies
the
Guard
Hypothesis
l If
‘hubs’
are
universal
targets…
l …network
topology
determines
which
nodes
are
likely
to
be
involved
in
host-‐pathogen
interac<on
Dangl
and
Jones.
Plant
pathogens
and
integrated
defence
responses
to
infec<on.
Nature
(2001)
vol.
411
(6839)
pp.
826-‐33
doi:10.1038/35081161
75. Dangl
and
Jones.
Plant
pathogens
and
integrated
defence
responses
to
infec<on.
Nature
(2001)
vol.
411
(6839)
pp.
826-‐33
doi:10.1038/35081161
The
Guard
Hypothesis
l The
Guard
Hypothesis
describes
indirect
R
gene:effector
interac<on
l Direct
R
gene:effector
interac<on
could
lead
to
overwhelming
R
gene
load
l A.
thaliana
has
≈200
R
genes
(1%
of
gene
complement)
l If
‘hubs’
are
common
targets
for
pathogens…
l …guarding
the
hub
with
one
R
gene
is
more
efficient
than
gene-‐for-‐gene
interac<ons
l …network
topology
implies
the
Guard
Hypothesis
l If
‘hubs’
are
universal
targets…
l …network
topology
determines
which
nodes
are
likely
to
be
involved
in
host-‐pathogen
interac<on
76. Interac2ons
with
hubs
l Host:
Arabidopsis
thaliana
l Pathogens:
Pseudomonas
syringae,
Hyaloperonospora
arabidopsidis
l Independent
effector
evolu<on
l Matrix-‐2-‐hybrid
(yeast-‐2-‐hybrid)
l Pathogen
effectors
share
more
common
targets
than
expected
(if
random)
l Common
targets
more
highly
connected
(i.e.
are
‘hubs’)
than
expected
(if
random)
Mukhtar
MS,
et
al.
(2011)
Independently
evolved
virulence
effectors
converge
onto
hubs
in
a
plant
immune
system
network.
Science
333:
596–601.
doi:10.1126/science.1203659.
77. Interac2ons
with
hubs
l Host:
Arabidopsis
thaliana
l Pathogens:
Pseudomonas
syringae,
Hyaloperonospora
arabidopsidis
l Independent
effector
evolu<on
l Matrix-‐2-‐hybrid
(yeast-‐2-‐hybrid)
l Pathogen
effectors
share
more
common
targets
than
expected
(if
random)
l Common
targets
more
highly
connected
(i.e.
are
‘hubs’)
than
expected
(if
random)
Mukhtar
MS,
et
al.
(2011)
Independently
evolved
virulence
effectors
converge
onto
hubs
in
a
plant
immune
system
network.
Science
333:
596–601.
doi:10.1126/science.1203659.
78. Interac2ons
with
hubs
l Host:
Arabidopsis
thaliana
l Pathogens:
Pseudomonas
syringae,
Hyaloperonospora
arabidopsidis
l Independent
effector
evolu<on
l Matrix-‐2-‐hybrid
(yeast-‐2-‐hybrid)
l Pathogen
effectors
share
more
common
targets
than
expected
(if
random)
l Common
targets
more
highly
connected
(i.e.
are
‘hubs’)
than
expected
(if
random)
Mukhtar
MS,
et
al.
(2011)
Independently
evolved
virulence
effectors
converge
onto
hubs
in
a
plant
immune
system
network.
Science
333:
596–601.
doi:10.1126/science.1203659.
79. Modules
in
networks
l Mo<fs
are
small
subnetworks
l Many
have
specific
dynamic
and
logic
behaviour:
„ Accelerate/slow
response
„ Enforce
sequen<al
responses
„ Lock
signal
on
or
off
„ Filter
out
noise
in
signals
„ Generate
pulse
in
response
to
signal
„ Generate
oscilla<ons
„ Integrate
and
process
mul<ple
signals
Shoval
and
Alon.
SnapShot:
network
mo<fs.
Cell
(2010)
vol.
143
(2)
pp.
326-‐e1
doi:10.1016/j.cell.2010.09.050
80. Modules
in
networks
l Mo<fs
are
small
subnetworks
l Many
have
specific
dynamic
and
logic
behaviour:
„ Accelerate/slow
response
„ Enforce
sequen<al
responses
„ Lock
signal
on
or
off
„ Filter
out
noise
in
signals
„ Generate
pulse
in
response
to
signal
„ Generate
oscilla<ons
„ Integrate
and
process
mul<ple
signals
Shoval
and
Alon.
SnapShot:
network
mo<fs.
Cell
(2010)
vol.
143
(2)
pp.
326-‐e1
doi:10.1016/j.cell.2010.09.050
81. Modules
in
networks
l Mo<fs
are
small
subnetworks
l Many
have
specific
dynamic
and
logic
behaviour:
„ Generate
pulse
in
response
to
signal
„ Generate
oscilla<ons
Shoval
and
Alon.
SnapShot:
network
mo<fs.
Cell
(2010)
vol.
143
(2)
pp.
326-‐e1
doi:10.1016/j.cell.2010.09.050
82. Modules
in
networks
l Bow-‐<e
structure
l Many
inputs
→
restricted
set
of
intermediates
→
many
outputs
83. Modules
in
networks
l Bow-‐<e
structure
l Many
inputs
→
restricted
set
of
intermediates
→
many
outputs
l e.g.
complex
nutrients
→
metabolic
intermediates
→
complex
compounds
84. Modules
in
networks
l Open
ques<ons:
l Do
a`ackers
preferen<ally
target
(or
introduce)
par<cular
mo<fs?
l Do
a`ackers
preferen<ally
target
the
‘knots’
of
bow-‐<e
structures?
85. Influence
in
networks
l Network
structure
(topology)
is
not
everything
l Network
topology
is
determined
by
dynamic
processes
n1
n2
n3
n4
n5
n1
n2
n3
n4
n5
n1
n2
n3
n4
n5
Idealised
topology
Expression
pa`ern
1
Expression
pa`ern
2
86. Influence
in
networks
l Network
structure
(topology)
is
not
everything
l Dynamic
processes
are
overlaid
on
topology
n1
n2
n3
n4
n5
n1
n2
n3
n4
n5
Idealised
topology
Reac<on
kine<cs
dictate
rela<ve
flux
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
87. Metabolic
Control
Analysis
(MCA)
l Metabolic
Control
Analysis
(MCA)
l Some
processes
more
influen<al
because
of
dynamic
(kine<c)
considera<ons
l ODE
representa<on
of
biochemical
network
l Used
to
understand
biochemical
pathways
l Used
in
ra<onal
drug
design:
target/priori<se
elements
with
large
control
coefficients
n1
n2
n3
n4
n5
Reac<on
kine<cs
dictate
rela<ve
flux
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
Kacser
and
Burns.
The
molecular
basis
of
dominance.
Gene/cs
(1981)
vol.
97
(3-‐4)
pp.
639-‐66
Kacser
and
Burns.
The
control
of
flux.
Biochem
Soc
Trans
(1995)
vol.
23
(2)
pp.
341-‐66
Westerhoff
and
Kell.
What
biotechnologists
knew
all
along
...?.
J
Theor
Biol
(1996)
vol.
182
(3)
pp.
411-‐420
Sato
et
al.
Network
Modeling
Reveals
Prevalent
Nega<ve
Regulatory
Rela<onships
between
Signaling
Sectors
in
Arabidopsis
Immune
Signaling.
PLoS
Pathog
(2010)
vol.
6
(7)
pp.
E1001011
doi:10.1371/journal.ppat.1001011
88. Metabolic
Control
Analysis
(MCA)
l Metabolic
Control
Analysis
(MCA)
l Some
processes
more
influen<al
because
of
dynamic
(kine<c)
considera<ons
l ODE
representa<on
of
biochemical
network
l Used
to
understand
biochemical
pathways
l Used
in
ra<onal
drug
design:
target/priori<se
elements
with
large
control
coefficients
n1
n2
n3
n4
n5
Reac<on
kine<cs
dictate
rela<ve
flux
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
Kacser
and
Burns.
The
molecular
basis
of
dominance.
Gene/cs
(1981)
vol.
97
(3-‐4)
pp.
639-‐66
Kacser
and
Burns.
The
control
of
flux.
Biochem
Soc
Trans
(1995)
vol.
23
(2)
pp.
341-‐66
Westerhoff
and
Kell.
What
biotechnologists
knew
all
along
...?.
J
Theor
Biol
(1996)
vol.
182
(3)
pp.
411-‐420
Sato
et
al.
Network
Modeling
Reveals
Prevalent
Nega<ve
Regulatory
Rela<onships
between
Signaling
Sectors
in
Arabidopsis
Immune
Signaling.
PLoS
Pathog
(2010)
vol.
6
(7)
pp.
E1001011
doi:10.1371/journal.ppat.1001011
89. Metabolic
Control
Analysis
(MCA)
l Metabolic
Control
Analysis
(MCA)
l Key
points:
l Rela<ve
change
in
pathway
flux
in
response
to
a
change
in
[enzyme]
is
the
flux
control
coefficient
l Rela<ve
change
in
[metabolite]
in
response
to
a
change
in
[enzyme]
is
the
concentra2on
control
coefficient
l Control
coefficient
=
0
⇒
no
influence
l Control
coefficient
=
1
⇒
strong
posi<ve
influence
l Control
coefficient
=
-‐1
⇒
strong
nega<ve
influence
n1
n2
n3
n4
n5
Reac<on
kine<cs
dictate
rela<ve
flux
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
90. Metabolic
Control
Analysis
(MCA)
l Metabolic
Control
Analysis
(MCA)
l Key
points:
l Rela<ve
change
in
pathway
flux
in
response
to
a
change
in
[enzyme]
is
the
flux
control
coefficient
l Rela<ve
change
in
[metabolite]
in
response
to
a
change
in
[enzyme]
is
the
concentra2on
control
coefficient
l Control
coefficient
=
0
⇒
no
influence
l Control
coefficient
=
1
⇒
strong
posi<ve
influence
l Control
coefficient
=
-‐1
⇒
strong
nega<ve
influence
l We
might
expect
aWackers
to
target
network
elements
with
large
control
coefficients
n1
n2
n3
n4
n5
Reac<on
kine<cs
dictate
rela<ve
flux
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
91. Metabolic
Control
Analysis
(MCA)
l Metabolic
Control
Analysis
(MCA)
l Key
points:
l Control
coefficients
dependent
on
rest
of
network:
calculated
at
same
<me
l Control
coefficients
are
a
system-‐level
property
(can’t
be
determined
in
isola<on)
l It
is
unusual
for
any
single
element
to
have
complete
control
over
any
part
of
the
network
l (Nearly)
no
rate-‐limi<ng
steps
l Any
part
of
the
network
is
typically
under
control
of
mul<ple
other
network
elements
l Distributed/democra<c
control
is
the
norm
n1
n2
n3
n4
n5
Reac<on
kine<cs
dictate
rela<ve
flux
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
Pritchard
and
Kell.
Schemes
of
flux
control
in
a
model
of
Saccharomyces
cerevisiae
glycolysis.
Eur
J
Biochem
(2002)
vol.
269
(16)
pp.
3894-‐904
D.
Fell,
Understanding
the
Control
of
Metabolism,
first
ed.,
Portland
Press,
1997.
92. Metabolic
Control
Analysis
(MCA)
l Metabolic
Control
Analysis
(MCA)
l Key
points:
l Control
coefficients
dependent
on
rest
of
network:
calculated
at
same
<me
l Control
coefficients
are
a
system-‐level
property
(can’t
be
determined
in
isola<on)
l It
is
unusual
for
any
single
element
to
have
complete
control
over
any
part
of
the
network
l (Nearly)
no
rate-‐limi<ng
steps
l Any
part
of
the
network
is
typically
under
control
of
mul<ple
other
network
elements
l Distributed/democra<c
control
is
the
norm
n1
n2
n3
n4
n5
Reac<on
kine<cs
dictate
rela<ve
flux
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
Pritchard
and
Kell.
Schemes
of
flux
control
in
a
model
of
Saccharomyces
cerevisiae
glycolysis.
Eur
J
Biochem
(2002)
vol.
269
(16)
pp.
3894-‐904
D.
Fell,
Understanding
the
Control
of
Metabolism,
first
ed.,
Portland
Press,
1997.
93. Metabolic
Control
Analysis
(MCA)
l Metabolic
Control
Analysis
(MCA)
l Key
points:
l Control
coefficients
dependent
on
rest
of
network:
calculated
at
same
<me
l Control
coefficients
are
a
system-‐level
property
(can’t
be
determined
in
isola<on)
l It
is
unusual
for
any
single
element
to
have
complete
control
over
any
part
of
the
network
l (Nearly)
no
rate-‐limi<ng
steps
l Any
part
of
the
network
is
typically
under
control
of
mul<ple
other
network
elements
l Distributed/democra2c
control
is
the
norm
n1
n2
n3
n4
n5
Reac<on
kine<cs
dictate
rela<ve
flux
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
v =
[S]Vmax
[S] + [Km]
Pritchard
and
Kell.
Schemes
of
flux
control
in
a
model
of
Saccharomyces
cerevisiae
glycolysis.
Eur
J
Biochem
(2002)
vol.
269
(16)
pp.
3894-‐904
D.
Fell,
Understanding
the
Control
of
Metabolism,
first
ed.,
Portland
Press,
1997.
94. Metabolic
Control
Analysis
(MCA)
l Yeast
glycolysis
l Most
enzyme
kine<c
parameters
known
l Fit
to
known
fluxes,
then
parameter-‐scan
(>8000
dis<nct
simula<ons)
l Three
regimes
of
control
found:
l Main
regime:
only
significant
control
by
hexose
transport
(HXT)
and
hexokinase
(HK)
l Minor
regime:
HXT,
HK
and
alcohol
dehydrogenase
(ADH)
l Biologically
inaccessible
regime:
[GLCi]
≈
300mM
phosphofructokinase
(PFK)
control
Pritchard
and
Kell.
Schemes
of
flux
control
in
a
model
of
Saccharomyces
cerevisiae
glycolysis.
Eur
J
Biochem
(2002)
vol.
269
(16)
pp.
3894-‐904
95. Metabolic
Control
Analysis
(MCA)
l Yeast
glycolysis
l Most
enzyme
kine<c
parameters
known
l Fit
to
known
fluxes,
then
parameter-‐scan
(>8000
dis<nct
simula<ons)
l Three
regimes
of
control
found:
l Main
regime:
only
significant
control
by
hexose
transport
(HXT)
and
hexokinase
(HK)
l Minor
regime:
HXT,
HK
and
alcohol
dehydrogenase
(ADH)
l Biologically
inaccessible
regime:
[GLCi]
≈
300mM
phosphofructokinase
(PFK)
control
Pritchard
and
Kell.
Schemes
of
flux
control
in
a
model
of
Saccharomyces
cerevisiae
glycolysis.
Eur
J
Biochem
(2002)
vol.
269
(16)
pp.
3894-‐904
96. Metabolic
Control
Analysis
(MCA)
l Yeast
glycolysis
l Most
enzyme
kine<c
parameters
known
l Fit
to
known
fluxes,
then
parameter-‐scan
(>8000
dis<nct
simula<ons)
l Three
regimes
of
control
found:
l Main
regime:
only
significant
control
by
hexose
transport
(HXT)
and
hexokinase
(HK)
l Minor
regime:
HXT,
HK
and
alcohol
dehydrogenase
(ADH)
l Biologically
inaccessible
regime:
[GLCi]
≈
300mM
under
phosphofructokinase
(PFK)
control
Pritchard
and
Kell.
Schemes
of
flux
control
in
a
model
of
Saccharomyces
cerevisiae
glycolysis.
Eur
J
Biochem
(2002)
vol.
269
(16)
pp.
3894-‐904
97. Metabolic
Control
Analysis
(MCA)
l Yeast
glycolysis
l HXT
dominates
pathway
control
l External
[hexose]
is
a
signal,
as
HXT
is
sensi<ve
to
it.
Pritchard
and
Kell.
Schemes
of
flux
control
in
a
model
of
Saccharomyces
cerevisiae
glycolysis.
Eur
J
Biochem
(2002)
vol.
269
(16)
pp.
3894-‐904
98. Distributed
Control
l MCA
implies
distributed
control
of
networks
l Network
topology
also
implies
distributed
control
(minimal
interven<on
sets:
MIS)
l What
does
this
imply
for
host-‐pathogen
interac<ons?
l Several
points
in
network
are
influen<al
„ Can
be
predicted
with
sufficient
informa<on
about
system
l A
pathway/network
element
may
be
under
distributed
control
„ May
need
to
hit
several
parts
of
the
network
to
produce
change
„ Single
effectors
unlikely
to
be
sufficient
99. Distributed
Control
l MCA
implies
distributed
control
of
networks
l Network
topology
also
implies
distributed
control
(minimal
interven<on
sets:
MIS)
l What
does
this
imply
for
host-‐pathogen
interac<ons?
l Several
points
in
network
are
influen<al
„ Can
be
predicted
with
sufficient
informa<on
about
system
l A
pathway/network
element
may
be
under
distributed
control
„ May
need
to
hit
several
parts
of
the
network
to
produce
change
„ Single
effectors
unlikely
to
be
sufficient
100. Distributed
Control
l MCA
implies
distributed
control
of
networks
l Network
topology
also
implies
distributed
control
(minimal
interven<on
sets:
MIS)
l What
does
this
imply
for
host-‐pathogen
interac<ons?
l Several
points
in
network
are
influen<al
„ Can
be
predicted
with
sufficient
informa<on
about
system
l A
pathway/network
element
may
be
under
distributed
control
„ May
need
to
hit
several
parts
of
the
network
to
produce
change
„ Single
effectors
unlikely
to
be
sufficient
101. Distributed
Control
l MCA
implies
distributed
control
of
networks
l Network
topology
also
implies
distributed
control
l What
does
this
imply
for
host-‐pathogen
interac<ons?
l Several
points
in
network
are
influen<al
l A
pathway/network
element
may
be
under
distributed
control
„ Pathogens
may
require
‘sets’
of
effectors
„ Implies
‘Redundant
Effector
Groups’
and
func2onal
redundancy?
Kvitko
et
al.
Dele<ons
in
the
repertoire
of
Pseudomonas
syringae
pv.
tomato
DC3000
type
III
secre<on
effector
genes
reveal
func<onal
overlap
among
effectors.
PLoS
Pathog
(2009)
vol.
5
(4)
pp.
E1000388
doi:10.1371/journal.ppat.1000388
102. Distributed
Control
l MCA
implies
distributed
control
of
networks
l Network
topology
also
implies
distributed
control
l What
does
this
imply
for
host-‐pathogen
interac<ons?
l Context-‐dependence
of
effector
func<on:
„ H.arabidopsidis
ATR13
suppresses
callose
deposi<on
„ P.
syringae
HopM1
suppresses
callose
deposi<on
„ ATR13
complements
callose
deposi<on,
but
does
not
fully
restore
virulence
in
HopM1
mutant
(EDV)
K.H.
Sohn,
R.
Lei,
A.
Nemri,
J.D.G.
Jones,
The
downy
mildew
effector
proteins
ATR1
and
ATR13
promote
disease
suscep<bility
in
Arabidopsis
thaliana,
Plant
Cell
19
(2007)
4077–4090.
103. Distributed
Control
l We
can
consider
‘system’
as
defining
a
landscape,
permi~ng
types
of
control
l Autocra<c
control:
l Flat
landscape
l Can
move
any
network
element
to
any
‘state’
l Democra<c
control:
l Rugged
landscape
(constrained
by
rest
of
network)
l Network
elements
restricted
to
‘valleys’
in
the
landscape
Bar-‐Yam
et
al.
Systems
biology.
A`ractors
and
democra<c
dynamics.
Science
(2009)
vol.
323
(5917)
pp.
1016-‐7
doi:10.1126/science.1163225
104. Distributed
Control
l We
can
consider
‘system’
as
defining
a
landscape,
permi~ng
types
of
control
l Autocra<c
control:
l Flat
landscape
l Can
move
any
network
element
to
any
‘state’
l Democra<c
control:
l Rugged
landscape
(constrained
by
rest
of
network)
l Network
elements
restricted
to
‘valleys’
in
the
landscape
Bar-‐Yam
et
al.
Systems
biology.
A`ractors
and
democra<c
dynamics.
Science
(2009)
vol.
323
(5917)
pp.
1016-‐7
doi:10.1126/science.1163225
105. Distributed
Control
l We
can
consider
‘system’
as
defining
a
landscape,
permi~ng
types
of
control
l Autocra<c
control:
l Flat
landscape
l Can
move
any
network
element
to
any
‘state’
l Democra<c
control:
l Rugged
landscape
(constrained
by
rest
of
network)
l Network
elements
restricted
to
‘valleys’
in
the
landscape
Bar-‐Yam
et
al.
Systems
biology.
A`ractors
and
democra<c
dynamics.
Science
(2009)
vol.
323
(5917)
pp.
1016-‐7
doi:10.1126/science.1163225
106. Distributed
Control
l We
can
consider
‘system’
as
defining
a
landscape,
permi~ng
types
of
control
l Autocra<c
control:
l Flat
landscape
l Can
move
any
network
element
to
any
‘state’
l Democra<c
control:
l Rugged
landscape
(constrained
by
rest
of
network)
l Network
elements
restricted
to
‘valleys’
in
the
landscape
l Pathogens
introduce
new
elements
that
change
the
landscape:
effectors
Bar-‐Yam
et
al.
Systems
biology.
A`ractors
and
democra<c
dynamics.
Science
(2009)
vol.
323
(5917)
pp.
1016-‐7
doi:10.1126/science.1163225
107. A
state-‐based
model
of
interac2on
l Prevailing
model:
zig-‐zag(-‐zig…)
Hein
et
al.
The
zig-‐zag-‐zig
in
oomycete-‐plant
interac<ons.
Mol
Plant
Pathol
(2009)
vol.
10
(4)
pp.
547-‐62
doi:10.1111/j.
1364-‐3703.2009.00547.x
Jones
and
Dangl.
The
plant
immune
system.
Nature
(2006)
vol.
444
(7117)
pp.
323-‐9
doi:10.1038/nature05286
108. A
state-‐based
model
of
interac2on
l Prevailing
model:
zig-‐zag(-‐zig…)
l Has
some
problems:
l scope
(only
host
immune
system,
not
rest
of
interac<on
with
pathogen)
l ordering
of
events
(are
PTI/ETI
etc.
dis<nct
and
well-‐ordered?)
l <mescale
(evolu<onary,
or
during
interac<on?)
l size
scale
(organism
or
cell
level)
l Quan<ta<ve
or
qualita<ve
(what
is
the
‘amplitude’
of
defence?)
109. A
state-‐based
model
of
interac2on
l Prevailing
model:
zig-‐zag(-‐zig…)
l Has
some
problems:
l scope
(only
host
immune
system,
not
rest
of
interac<on
with
pathogen)
l ordering
of
events
(are
PTI/ETI
etc.
dis<nct
and
well-‐ordered?)
l <mescale
(evolu<onary,
or
during
interac<on?)
l size
scale
(organism
or
cell
level)
l Quan<ta<ve
or
qualita<ve
(what
is
the
‘amplitude’
of
defence?)
110. A
state-‐based
model
of
interac2on
l Prevailing
model:
zig-‐zag(-‐zig…)
l Has
some
problems:
l scope
(only
host
immune
system,
not
rest
of
interac<on
with
pathogen)
l ordering
of
events
(are
PTI/ETI
etc.
dis<nct
and
well-‐ordered?)
l <mescale
(evolu<onary,
or
during
interac<on?)
l size
scale
(organism
or
cell
level)
l Quan<ta<ve
or
qualita<ve
(what
is
the
‘amplitude’
of
defence?)
111. A
state-‐based
model
of
interac2on
l Prevailing
model:
zig-‐zag(-‐zig…)
l Has
some
problems:
l scope
(only
host
immune
system,
not
rest
of
interac<on
with
pathogen)
l ordering
of
events
(are
PTI/ETI
etc.
dis<nct
and
well-‐ordered?)
l <mescale
(evolu<onary,
or
during
interac<on?)
l size
scale
(organism
or
cell
level)
l Quan<ta<ve
or
qualita<ve
(what
is
the
‘amplitude’
of
defence?)
112. A
state-‐based
model
of
interac2on
l Prevailing
model:
zig-‐zag(-‐zig…)
l Has
some
problems:
l scope
(only
host
immune
system,
not
rest
of
interac<on
with
pathogen)
l ordering
of
events
(are
PTI/ETI
etc.
dis<nct
and
well-‐ordered?)
l <mescale
(evolu<onary,
or
during
interac<on?)
l size
scale
(organism
or
cell
level)
l Quan<ta<ve
or
qualita<ve
(what
is
the
‘amplitude’
of
defence?)
113. A
state-‐based
model
of
interac2on
l Prevailing
model:
zig-‐zag(-‐zig…)
l Has
some
problems:
l scope
(only
host
immune
system,
not
rest
of
interac<on
with
pathogen)
l ordering
of
events
(are
PTI/ETI
etc.
dis<nct
and
well-‐ordered?)
l <mescale
(evolu<onary,
or
during
interac<on?)
l size
scale
(organism
or
cell
level)
l Quan<ta<ve
or
qualita<ve
(what
is
the
‘amplitude’
of
defence?)
l Is
there
a
more
general
framework
for
host-‐pathogen
interac2ons?
Pritchard
L,
Birch
P
(2011)
A
systems
biology
perspec<ve
on
plant-‐microbe
interac<ons:
Biochemical
and
structural
targets
of
pathogen
effectors.
Plant
Science
180:
584–603.
doi:10.1016/j.plantsci.2010.12.008.
114. A
state-‐based
model
of
interac2on
l Biological
cells
can
be
represented
as
networks
l Each
element
in
the
network
can
be
quan<fied:
l enzyme
concentra<on
(or
expression
level)
l metabolite
concentra<on
l phosphoryla<on/ubiqui<na<on/charge
states
as
dis<nct
en<<es
l etc.
l We
represent
lists
of
values
as
vectors
[v1, v2, v3, . . . , vk]
115. A
state-‐based
model
of
interac2on
l Biological
cells
can
be
represented
as
networks
l Each
element
in
the
network
can
be
quan<fied:
l enzyme
concentra<on
(or
expression
level)
l metabolite
concentra<on
l phosphoryla<on/ubiqui<na<on/charge
states
as
dis<nct
en<<es
l etc.
l We
represent
ordered
lists
of
values
as
vectors
[v1, v2, v3, . . . , vk]
116. A
state-‐based
model
of
interac2on
l Biological
cells
can
be
represented
as
networks
l Each
element
in
the
network
can
be
quan<fied:
l enzyme
concentra<on
(or
expression
level)
l metabolite
concentra<on
l phosphoryla<on/ubiqui<na<on/charge
states
as
dis<nct
en<<es
l etc.
l We
represent
ordered
lists
of
values
as
vectors
[v1, v2, v3, . . . , vk]
117. A
state-‐based
model
of
interac2on
l Vectors
are
co-‐ordinates
in
space
l vectors
of
length
two:
points
on
a
surface
(2D
space)
l vectors
of
length
three:
points
in
3D
space
l vectors
of
length
k:
points
in
k-‐dimensional
space
l Points
that
are
close
together
are
‘similar’
118. A
state-‐based
model
of
interac2on
l Vectors
are
co-‐ordinates
in
space
l vectors
of
length
two:
points
on
a
surface
(2D
space)
l vectors
of
length
three:
points
in
3D
space
l vectors
of
length
k:
points
in
k-‐dimensional
space
l Points
that
are
close
together
are
‘similar’
119. A
state-‐based
model
of
interac2on
l Let
our
vector
represent
the
measured
state
of
the
cell
(e.g.
host-‐pathogen)
system
l enzyme/metabolite
concentra<ons,
etc.
l Each
point
in
k-‐space
represents
a
different
state
of
the
system
l similar
states
are
close
together
in
k-‐space
[v1, v2, v3, . . . , vk]
120. A
state-‐based
model
of
interac2on
l Let
our
vector
represent
the
measured
state
of
the
cell
(e.g.
host-‐pathogen)
system
l enzyme/metabolite
concentra<ons,
etc.
l Each
point
in
k-‐space
represents
a
different
state
of
the
system
l similar
states
are
close
together
in
k-‐space
[v1, v2, v3, . . . , vk]
121. A
state-‐based
model
of
interac2on
l States
that
lead
to
similar
phenotypes
can
be
grouped
in
phases:
l regions
of
space
where
cell
state
corresponds
to
named
behaviour
l Temporal
evolu<on
of
a
cell
can
be
viewed
as
a
transi<on
through
states
v1
v2
apoptosis
ROS
produc<on
seed
leaf
root
HR
122. A
state-‐based
model
of
interac2on
l States
that
lead
to
similar
phenotypes
can
be
grouped
in
phases:
l regions
of
space
where
cell
state
corresponds
to
named
behaviour
l Temporal
evolu<on
of
a
cell
can
be
viewed
as
a
transi<on
through
states
v1
v2
apoptosis
ROS
produc<on
seed
leaf
root
HR
123. A
state-‐based
model
of
interac2on
l Complex
systems
can
behave
in
complex
ways
l A
common
feature
of
complex
systems
is
aJractors
l A`ractors
are
‘endpoints’:
states,
or
sets
of
states,
to
which
the
system
is
‘a`racted’
l Analogous
to
stable
equilibria:
when
the
system
is
perturbed,
it
returns
to
its
a`ractor.
l Do
cell
phenotypes
correspond
to
a`ractors?
124. A
state-‐based
model
of
interac2on
l Complex
systems
can
behave
in
complex
ways
l A
common
feature
of
complex
systems
is
aJractors
l A`ractors
are
‘endpoints’:
states,
or
sets
of
states,
to
which
the
system
is
‘a`racted’
l Analogous
to
stable
equilibria:
when
the
system
is
perturbed,
it
returns
to
its
a`ractor.
l Do
cell
phenotypes
correspond
to
a`ractors?
125. A
state-‐based
model
of
interac2on
l Complex
systems
can
behave
in
complex
ways
l A
common
feature
of
complex
systems
is
aJractors
l A`ractors
are
‘endpoints’:
states,
or
sets
of
states,
to
which
the
system
is
‘a`racted’
l Analogous
to
stable
equilibria:
when
the
system
is
perturbed,
it
returns
to
its
a`ractor.
l Do
cell
phenotypes
correspond
to
a`ractors?
126. A
state-‐based
model
of
interac2on
l Complex
systems
can
behave
in
complex
ways
l A
common
feature
of
complex
systems
is
aJractors
l A`ractors
are
‘endpoints’:
states,
or
sets
of
states,
to
which
the
system
is
‘a`racted’
l Analogous
to
stable
equilibria:
when
the
system
is
perturbed,
it
returns
to
its
a`ractor.
l Do
cell
phenotypes
correspond
to
a`ractors?
apoptosis
ROS
produc<on
seed
leaf
root
HR
127. A
state-‐based
model
of
interac2on
l A`ractors
are
associated
with
the
regions
of
space
that
lead
to
them:
‘basins’
l A`ractors
can
be:
l Single
points
l Cycles
l Complex
‘regions’
128. A
state-‐based
model
of
interac2on
l A`ractors
are
associated
with
the
regions
of
space
that
lead
to
them:
‘basins’
l A`ractors
can
be:
l Single
points
l Cycles
l Complex
‘regions’
129. A
state-‐based
model
of
interac2on
l Interac<on
of
a
pathogen
with
the
host
can
push
the
system
from
one
basin
of
aJrac/on
to
another
l There
may
be
mul<ple
routes
between
basins
of
a`rac<on,
depending
on
the
direc<on
or
<ming
of
perturba<on
l There
may
be
more
than
one
way
to
provoke
a
specific
outcome
from
the
host
(or
from
the
pathogen)
130. A
state-‐based
model
of
interac2on
l Interac<on
of
a
pathogen
with
the
host
can
push
the
system
from
one
basin
of
aJrac/on
to
another
l There
may
be
mul<ple
routes
between
basins
of
a`rac<on,
depending
on
the
direc<on
or
<ming
of
perturba<on
l There
may
be
more
than
one
way
to
provoke
a
specific
outcome
from
the
host
(or
from
the
pathogen)
131. A
state-‐based
model
of
interac2on
l Effectors
may
divert
the
expected
WT
system
trajectory:
l ‘Pushing’
the
host
cell
state
towards
a
different
aJractor/state
l ‘State’
may
be
a
developmental
checkpoint
l Diversion
of
the
trajectory
may
also
be
beneficial
to
the
host
l The
pathogen
may
detect
the
host
state
and
respond
accordingly
(e.g.
<ssue-‐specific
effector
produc<on
in
Us/lago
maydis;
stage-‐
and
<ssue-‐specific
oomycete
effectors)
v1
v2
nutrient
produc<on
PTI
seed
Epidermal
cell
root
HR
132. A
state-‐based
model
of
interac2on
l The
Jones-‐Dangl
Zig-‐Zag(-‐Zig)
model
is
encapsulated
within
a
state-‐based
model
PTI
No
challenge
ETI
ETS
133. A
state-‐based
model
of
interac2on
l The
Jones-‐Dangl
Zig-‐Zag(-‐Zig)
model
is
encapsulated
within
a
state-‐based
model
PTI
No
challenge
ETI
ETS
134. A
state-‐based
model
of
interac2on
l The
Jones-‐Dangl
Zig-‐Zag(-‐Zig)
model
is
encapsulated
within
a
state-‐based
model
PTI
No
challenge
ETI
ETS
135. A
state-‐based
model
of
interac2on
l The
Jones-‐Dangl
Zig-‐Zag(-‐Zig)
model
is
encapsulated
within
a
state-‐based
model
PTI
No
challenge
ETI
ETS
136. A
state-‐based
model
of
interac2on
l The
Jones-‐Dangl
Zig-‐Zag(-‐Zig)
model
is
encapsulated
within
a
state-‐based
model
PTI
No
challenge
ETI
ETS
137. A
state-‐based
model
of
interac2on
l The
Jones-‐Dangl
Zig-‐Zag(-‐Zig)
model
is
encapsulated
within
a
state-‐based
model
PTI
No
challenge
ETI
ETS
138. A
state-‐based
model
of
interac2on
l The
state-‐based
model
has
advantages:
l Scope:
can
include
host
and
pathogen,
and
extend
beyond
host
immunity
l Ordering:
explicit
ordering
of
events
represented
by
paths
in
the
model
(determined
by
model)
l Timescale:
explicit
(determined
by
model)
l Size
scale:
can
include
mul<cellular
systems
l Quan2ta2ve
or
qualita2ve:
explicit
(dependent
on
model)
PTI
No
challenge
ETI
ETS
139. A
state-‐based
model
of
interac2on
l The
state-‐based
model
has
advantages:
l Scope:
can
include
host
and
pathogen,
and
extend
beyond
host
immunity
l Ordering:
explicit
ordering
of
events
represented
by
paths
in
the
model
(determined
by
model)
l Timescale:
explicit
(determined
by
model)
l Size
scale:
can
include
mul<cellular
systems
l Quan2ta2ve
or
qualita2ve:
explicit
(dependent
on
model)
PTI
No
challenge
ETI
ETS
140. A
state-‐based
model
of
interac2on
l The
state-‐based
model
has
advantages:
l Scope:
can
include
host
and
pathogen,
and
extend
beyond
host
immunity
l Ordering:
explicit
ordering
of
events
represented
by
paths
in
the
model
(determined
by
model)
l Timescale:
explicit
(determined
by
model)
l Size
scale:
can
include
mul<cellular
systems
l Quan2ta2ve
or
qualita2ve:
explicit
(dependent
on
model)
PTI
No
challenge
ETI
ETS
141. A
state-‐based
model
of
interac2on
l The
state-‐based
model
has
advantages:
l Scope:
can
include
host
and
pathogen,
and
extend
beyond
host
immunity
l Ordering:
explicit
ordering
of
events
represented
by
paths
in
the
model
(determined
by
model)
l Timescale:
explicit
(determined
by
model)
l Size
scale:
can
include
mul<cellular
systems
l Quan2ta2ve
or
qualita2ve:
explicit
(dependent
on
model)
PTI
No
challenge
ETI
ETS
142. A
state-‐based
model
of
interac2on
l The
state-‐based
model
has
advantages:
l Scope:
can
include
host
and
pathogen,
and
extend
beyond
host
immunity
l Ordering:
explicit
ordering
of
events
represented
by
paths
in
the
model
(determined
by
model)
l Timescale:
explicit
(determined
by
model)
l Size
scale:
can
include
mul<cellular
systems
l Quan2ta2ve
or
qualita2ve:
explicit
(dependent
on
model)
PTI
No
challenge
ETI
ETS
143. Summary
l Biological
systems
have
natural
network
representa<ons
l But
representa<on
must
be
reasonable
and
suit
the
ques<on
being
asked
l Interac<on
of
host
and
pathogen
makes
a
new
single
network
from
two
ini<al
networks
l Network
topology
affects
l Network
behaviour
l Suscep<bility
to
a`ack
(hubs,
bridges)
l Network
dynamics
affect
l Network
behaviour
l Suscep<bility
to
a`ack
(distributed
control)
l A
state-‐based
framework
may
be
useful
for
understanding
host-‐
pathogen
interac<ons
144. Acknowledgements
l Systems
Biology
at
Aberystwyth/Manchester
l Doug
Kell,
David
Broadhurst,
Pedro
Mendes,
Roy
Goodacre,
Andy
Woodward,
Simon
Garre`,
l Computa<onal
biology
at
JHI
l Peter
Cock
l Phytophthora
research
at
JHI
l Paul
Birch,
Steve
Whisson,
Miles
Armstrong
l Bacteriology
research
at
JHI
l Ian
Toth,
Sonia
Humphris,
Nicola
Holden
l Many,
many
discussions
with
colleagues
145. Danger
Theory
l Proposed
by
computer
scien<sts
in
machine
learning:
avoids
detec<on
‘bloat’
of
one
‘recogni<on
gene’
per
threat.
l Popular
in
(animal)
immunology;
Analogous
to
Guard
Hypothesis
and
Dense
Overlapping
Regions
(DORs)
l Integra<on
of
mul<ple
signals
and
contextual
cues
Aickelin
et
al.
Danger
theory:
The
link
between
AIS
and
IDS?.
Lect
Notes
Comput
Sc
(2003)
vol.
2787
pp.
147-‐155
146. Danger
Theory
l Some
signals
‘cri<cal’
and
require
immediate
response
(e.g.
avirulence
gene
products?)
l Other
signals
contextual
–
require
‘processing’
(e.g.
MAMPs)
Boller
and
Felix.
A
renaissance
of
elicitors:
percep<on
of
microbe-‐associated
molecular
pa`erns
and
danger
signals
by
pa`ern-‐recogni<on
receptors.
Annu.
Rev.
Plant.
Biol.
(2009)
vol.
60
pp.
379-‐406
doi:10.1146/annurev.arplant.
57.032905.105346
147. Danger
Theory
l Context
dependence
and
non-‐linear
signal
may
lead
to
problems
of
interpreta<on
in
experiments.
l Danger
R
when
signal
≥
5
l a+b+c+d
=
6
⇒
R
l a+b+c
=
4
⇒
no
R
l a+b+d
=
4
⇒
no
R
l a+c+d
=
5
⇒
R
l b+c+d
=
5
⇒
R
l a+b
=
3
⇒
no
R
l {c
and
d}
required
for
R?
148. Danger
Theory
l Context
dependence
and
non-‐linear
signal
may
lead
to
problems
of
interpreta<on
in
experiments.
l Danger
R
when
signal
≥
5
l a+b+c+d
=
6
⇒
R
l a+b+c
=
4
⇒
no
R
l a+b+d
=
4
⇒
no
R
l a+c+d
=
5
⇒
R
l b+c+d
=
5
⇒
R
l a+b
=
3
⇒
no
R
l {c
and
d}
required
for
R?
149. Danger
Theory
l Context
dependence
and
non-‐linear
signal
may
lead
to
problems
of
interpreta<on
in
experiments.
l Danger
R
when
signal
≥
5
l a+b+c+d
=
6
⇒
R
l a+b+c
=
4
⇒
no
R
l a+b+d
=
4
⇒
no
R
l a+c+d
=
5
⇒
R
l b+c+d
=
5
⇒
R
l a+b
=
3
⇒
no
R
l {c
and
d}
required
for
R?
150. Danger
Theory
l Context
dependence
and
non-‐linear
signal
may
lead
to
problems
of
interpreta<on
in
experiments.
l Danger
R
when
signal
≥
5
l a+b+c+d
=
6
⇒
R
l a+b+c
=
4
⇒
no
R
l a+b+d
=
4
⇒
no
R
l a+c+d
=
5
⇒
R
l b+c+d
=
5
⇒
R
l a+b
=
3
⇒
no
R
l {c
and
d}
required
for
R?
l No:
a+b+c+e,
a+b+d+e
⇒
R