SlideShare a Scribd company logo
1 of 46
Download to read offline
Design principles in pattern formation: Robustness and
equivalences
Michael P.H. Stumpf
Theoretical Systems Biology, Melbourne Integrative Genomics
School of BioSciences &
School of Mathematics and Statistics
Engineering Turing Patterns
Development-
tal Biology
Systems
Biology
Synthetic
Biology
Mathematical
Modelling
James Briscoe
Michael Stumpf Heike Siebert
Mark Isalan
Statistical
Inference
Model
Reduction
Molecular
Biology
Control
Engineering
Multi-scale
Modelling
Genome
Editing
In Silico
Design
Stem Cell
Biology
Bio-
informatics
Natalie Scholes
David Schnoerr
Sean Vittadello
Turing Patterns
Alan Turing
A B
Local Activation & Lateral Inhibition
Turing Patterns are Robust
Gierer-Meinhardt Model
(A)
(B) (D)
(C)
Vittadello et al., Phil Trans Roy Soc B (2021); Leyshon et al., J Theo Biol (2021).
Positional Information ( aka French Flag Model)
n
or
bition
Concentration
T1
T2
Concentration
Space
Positional information creating
the French Flag pattern
B Wolpert – PI
i
Green & Sharpe, Development (2015)
Lewis Wolpert
Why the Acrimony?
Why the Acrimony?
Why the Acrimony?
Why the Acrimony?
Why the Acrimony?
Stumpf†, Nature (1966) Raspopovic et al., Science (2014)
†Hildegard F. Stumpf (1924-1967) – No relation.
Turing Patterns vs
Positional Information
The simple dichotomy is wrong. There is
room and need for both mechanisms.
We are interested in
The design principles of Turing Pat-
terns.
The design principles of Positional
Information Mechanisms.
The differences between the two
sets of design principles.
Concentration
Space
Concentration
Space
Concentration
Space
Concentration
Space
Concentration
Space
Auto-activation
Activation
of inhibitor
Lateral inhibition
of activator
Diffusion
of inhibitor
Consistent ‘wavelength’ controlled by global parameters
i
ii
iii
iv
v
Molecular
fluctuations
Concentration
T1
T2
Concentration
Space
Positional information creating
the French Flag pattern
T7
Concentration
Space
T6
T5
T4
T3
T2
T1
B Wolpert – PI
i
ii
Positional information creating
a periodic pattern
Green & Sharpe, Development (2015)
Turing Generating Mechanisms and Their Properties
We considered
> 5,000 network structures.
> 1011
parameter combina-
tions.
We sampled neighbouring
4–node motifs.
Scholes et al., Cell Systems (2019)
Turing Generating Mechanisms and Their Properties
We need two ingredients:
Non-spatial dynamics allow for a non-trivial stationary state.
Spatial dynamics (≡ spatial dynamics but with diffusion of molecules turned on) are unstable.
C
A
“Fast inhibitor”
“Slow activator”
Non-
homogeneous
initial state
Reaction &
Diffusion
Pattern
formation
time
time
1 A B 2 A B 3 A B 4 A B 5
8 A B 9 A B 10 A B 11 A B 12
15 A B 16 A B 17 A B 18 A B 19
E Non-Competitive
time
activating
inhibiting
D Topology
B
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB (2)
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA (3)
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB (4)
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5 (5)
@A
@t
= ↵ + ·
A2
B
µAA (6)
@B
@t
= · A2
µBB (7)
2
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB (2)
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA (3)
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB (4)
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5 (5)
@A
@t
= ↵ + ·
A2
B
µAA (6)
@B
@t
= · A2
µBB (7)
2
B
2-Node:
3-Node:
0
1
-1
Reduce according to symmetry
Only consider connectednetworks
…
Adjacency matrix
representation
Translating networks into ODE equations
1 1
−1 0
C Estimating Turing Pattern capability
Competitive
Non-Competitive
Graph
representation
Parameter Range
V 0.1-100
Specify parameter boundaries Stability analysis:
A B
A B
1
A B
2
A B
3
A B
4
A
5
A B
6
A B
7
A B
8
A B
9
A
10
A B
11
A B
12
A B
13
A B
14
A
15
A B
16
A B
17
A B
18
A B
19
A
20
A B
21
A
B C
composes the central node and mediates the signals between both HGF and NK4 (see Fig 7 C).
With this model, the network equals the core topology #28 with nodes B and C diffusing. Indeed,
this topology is amongst the top 10 % of the most robust topologies according to our results. Even
so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely to
diffuse at quite similar rates, making this network only implementable if HGF could be modified
to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain the
original single promoter design, the network atlas reveals that including a positive feedback loop
on top of the receptor would suffice to make equal diffusivity accessible for given network (# 63
BC). This, however, though improving robustness with respect to extracellular parameters does
not improve intracellular robustness significantly and thus in total, robustness is only marginally
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interaction
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this shows
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA (1)
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB (2)
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA (3)
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB (4)
2@fA
D q2 @fA
3
so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely t
diffuse at quite similar rates, making this network only implementable if HGF could be modifie
to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain th
original single promoter design, the network atlas reveals that including a positive feedback loo
on top of the receptor would suffice to make equal diffusivity accessible for given network (# 6
BC). This, however, though improving robustness with respect to extracellular parameters do
not improve intracellular robustness significantly and thus in total, robustness is only marginal
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interactio
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this show
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA (1
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB (2
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA (3
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB (4
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5 (5
so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely to
diffuse at quite similar rates, making this network only implementable if HGF could be modified
to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain the
original single promoter design, the network atlas reveals that including a positive feedback loop
on top of the receptor would suffice to make equal diffusivity accessible for given network (# 63
BC). This, however, though improving robustness with respect to extracellular parameters does
not improve intracellular robustness significantly and thus in total, robustness is only marginally
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interaction
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this shows
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA (1)
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB (2)
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA (3)
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB (4)
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5 (5)
on top of the receptor would suffice to make equal diffusivity accessible for given network (# 6
BC). This, however, though improving robustness with respect to extracellular parameters do
not improve intracellular robustness significantly and thus in total, robustness is only marginal
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interactio
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this show
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA (
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB (
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA (
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB (
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5 (
@A
@t
= ⇢ ·
A2
B
µAA (
so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely
diffuse at quite similar rates, making this network only implementable if HGF could be modifi
to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain
original single promoter design, the network atlas reveals that including a positive feedback lo
on top of the receptor would suffice to make equal diffusivity accessible for given network (#
BC). This, however, though improving robustness with respect to extracellular parameters d
not improve intracellular robustness significantly and thus in total, robustness is only margina
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interact
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this sho
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5
Initial conditions:
B
2-Node:
3-Node:
0
1
-1
Reduce according to symmetry
Only consider connectednetworks
…
Adjacency matrix
representation
Translating networks into ODE equations
1 1
−1 0
C Estimating Turing Pattern capability
Competitive
Non-Competitive
Graph
representation
Parameter Range
V 0.1-100
Specify parameter boundaries Stability analysis:
A B
A B
1
A B
2
A B
3
A B
4
A
5
A B
6
A B
7
A B
8
A B
9
A
10
A B
11
A B
12
A B
13
A B
14
A
15
A B
16
A B
17
A B
18
A B
19
A
20
A B
21
A
B C
composes the central node and mediates the signals between both HGF and NK4 (see Fig 7 C).
With this model, the network equals the core topology #28 with nodes B and C diffusing. Indeed,
this topology is amongst the top 10 % of the most robust topologies according to our results. Even
so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely to
diffuse at quite similar rates, making this network only implementable if HGF could be modified
to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain the
original single promoter design, the network atlas reveals that including a positive feedback loop
on top of the receptor would suffice to make equal diffusivity accessible for given network (# 63
BC). This, however, though improving robustness with respect to extracellular parameters does
not improve intracellular robustness significantly and thus in total, robustness is only marginally
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interaction
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this shows
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA (1)
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB (2)
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA (3)
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB (4)
2@fA
D q2 @fA
3
so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely t
diffuse at quite similar rates, making this network only implementable if HGF could be modifie
to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain th
original single promoter design, the network atlas reveals that including a positive feedback loo
on top of the receptor would suffice to make equal diffusivity accessible for given network (# 6
BC). This, however, though improving robustness with respect to extracellular parameters do
not improve intracellular robustness significantly and thus in total, robustness is only marginal
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interactio
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this show
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA (1
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB (2
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA (3
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB (4
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5 (5
so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely to
diffuse at quite similar rates, making this network only implementable if HGF could be modified
to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain the
original single promoter design, the network atlas reveals that including a positive feedback loop
on top of the receptor would suffice to make equal diffusivity accessible for given network (# 63
BC). This, however, though improving robustness with respect to extracellular parameters does
not improve intracellular robustness significantly and thus in total, robustness is only marginally
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interaction
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this shows
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA (1)
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB (2)
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA (3)
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB (4)
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5 (5)
on top of the receptor would suffice to make equal diffusivity accessible for given network (# 6
BC). This, however, though improving robustness with respect to extracellular parameters do
not improve intracellular robustness significantly and thus in total, robustness is only marginal
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interactio
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this show
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA (
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB (
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA (
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB (
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5 (
@A
@t
= ⇢ ·
A2
B
µAA (
so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely
diffuse at quite similar rates, making this network only implementable if HGF could be modifi
to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain
original single promoter design, the network atlas reveals that including a positive feedback lo
on top of the receptor would suffice to make equal diffusivity accessible for given network (#
BC). This, however, though improving robustness with respect to extracellular parameters d
not improve intracellular robustness significantly and thus in total, robustness is only margina
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interact
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this sho
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5
Initial conditions:
C
qmax
2π /qmax
A
“Fast inhibitor”
“Slow activator”
Non-
homogeneous
initial state
Reaction &
Diffusion
Pattern
formation
time
time
1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7
8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14
15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21
E
Competitive
Non-Competitive
time
activating
inhibiting
D Topology
Regulatory function
Function parameters
Diffusion constants
B
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB (4)
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5 (5)
@A
@t
= ↵ + ·
A2
B
µAA (6)
@B
@t
= · A2
µBB (7)
2
@B
@t
= VB ·
(kAB
)
1 + ( A
kAB
)n
+ bB µBB (4)
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5 (5)
@A
@t
= ↵ + ·
A2
B
µAA (6)
@B
@t
= · A2
µBB (7)
2
B
2-Node:
3-Node:
0
1
-1
Reduce according to symmetry
Only consider connectednetworks
…
Adjacency matrix
representation
Translating networks into ODE equations
1 1
−1 0
C Estimating Turing Pattern capability
Competitive
Non-Competitive
Graph
representation
Parameter Range
V 0.1-100
Specify parameter boundaries Stability analysis:
A B
A B
1
A B
2
A B
3
A B
4
A
5
A B
6
A B
7
A B
8
A B
9
A
10
A B
11
A B
12
A B
13
A B
14
A
15
A B
16
A B
17
A B
18
A B
19
A
20
A B
21
A
B C
composes the central node and mediates the signals between both HGF and NK4 (see Fig 7 C).
With this model, the network equals the core topology #28 with nodes B and C diffusing. Indeed,
this topology is amongst the top 10 % of the most robust topologies according to our results. Even
so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely to
diffuse at quite similar rates, making this network only implementable if HGF could be modified
to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain the
original single promoter design, the network atlas reveals that including a positive feedback loop
on top of the receptor would suffice to make equal diffusivity accessible for given network (# 63
BC). This, however, though improving robustness with respect to extracellular parameters does
not improve intracellular robustness significantly and thus in total, robustness is only marginally
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interaction
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this shows
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA (1)
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB (2)
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA (3)
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB (4)
2@fA
D q2 @fA
3
so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely t
diffuse at quite similar rates, making this network only implementable if HGF could be modifie
to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain th
original single promoter design, the network atlas reveals that including a positive feedback loo
on top of the receptor would suffice to make equal diffusivity accessible for given network (# 6
BC). This, however, though improving robustness with respect to extracellular parameters do
not improve intracellular robustness significantly and thus in total, robustness is only marginal
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interactio
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this show
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA (1
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB (2
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA (3
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB (4
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5 (5
so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely to
diffuse at quite similar rates, making this network only implementable if HGF could be modified
to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain the
original single promoter design, the network atlas reveals that including a positive feedback loop
on top of the receptor would suffice to make equal diffusivity accessible for given network (# 63
BC). This, however, though improving robustness with respect to extracellular parameters does
not improve intracellular robustness significantly and thus in total, robustness is only marginally
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interaction
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this shows
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA (1)
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB (2)
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA (3)
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB (4)
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5 (5)
on top of the receptor would suffice to make equal diffusivity accessible for given network (# 6
BC). This, however, though improving robustness with respect to extracellular parameters do
not improve intracellular robustness significantly and thus in total, robustness is only marginal
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interactio
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this show
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA (
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB (
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA (
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB (
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5 (
@A
@t
= ⇢ ·
A2
B
µAA (
so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely
diffuse at quite similar rates, making this network only implementable if HGF could be modifi
to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain
original single promoter design, the network atlas reveals that including a positive feedback lo
on top of the receptor would suffice to make equal diffusivity accessible for given network (#
BC). This, however, though improving robustness with respect to extracellular parameters d
not improve intracellular robustness significantly and thus in total, robustness is only margina
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interact
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this sho
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5
Initial conditions:
C
qmax
2π /qmax
A
“Fast inhibitor”
“Slow activator”
Non-
homogeneous
initial state
Reaction &
Diffusion
Pattern
formation
time
time
1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7
8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14
15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21
E
Competitive
Non-Competitive
time
activating
inhibiting
D Topology
Regulatory function
Function parameters
Diffusion constants
B
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB (4)
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5 (5)
@A
@t
= ↵ + ·
A2
B
µAA (6)
@B
@t
= · A2
µBB (7)
2
@B
@t
= VB ·
(kAB
)
1 + ( A
kAB
)n
+ bB µBB (4)
J =
2
4
@fA
@A D1q2 @fA
@B
@fB
@A
@fB
@B D2q2
3
5 (5)
@A
@t
= ↵ + ·
A2
B
µAA (6)
@B
@t
= · A2
µBB (7)
2
B
A Defining network structure
2-Node:
3-Node:
0
1
-1
Reduce according to symmetry
Only consider connectednetworks
…
Adjacency matrix
representation
Translating networks into ODE equations
1 1
−1 0
Competitive
Non-Competitive
Graph
representation
A B
A B
1
A B
2
A B
3
A B
4
A
A B
6
A B
7
A B
8
A B
9
A
A B
11
A B
12
A B
13
A B
14
A
A B
16
A B
17
A B
18
A B
19
A
A B
21
A
B C
MMP-1 promoter construct. In a previous study, this system was analyzed as a classical 2-node
system and the authors suggested that a single promoter construct driving the expression of both
activating and inhibiting species could suffice to create Turing Patterns [?]. This system would,
however, require differential diffusion (10 fold) and a co-operativity factor of >= 2 for feasible
Turing pattern generation. In reality, the network can be seen as a 3-node network in which c-Met
composes the central node and mediates the signals between both HGF and NK4 (see Fig 7 C).
With this model, the network equals the core topology #28 with nodes B and C diffusing. Indeed,
this topology is amongst the top 10 % of the most robust topologies according to our results. Even
so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely to
diffuse at quite similar rates, making this network only implementable if HGF could be modified
to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain the
original single promoter design, the network atlas reveals that including a positive feedback loop
on top of the receptor would suffice to make equal diffusivity accessible for given network (# 63
BC). This, however, though improving robustness with respect to extracellular parameters does
not improve intracellular robustness significantly and thus in total, robustness is only marginally
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interaction
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this shows
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA (1)
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB (2)
@A ( A
k )n
Turing pattern generation. In reality, the network can be seen as a 3-node network in which
composes the central node and mediates the signals between both HGF and NK4 (see Fig
With this model, the network equals the core topology #28 with nodes B and C diffusing. In
this topology is amongst the top 10 % of the most robust topologies according to our results.
so, the classical requirements found for the 2-node system persist HGF and NK4 are more lik
diffuse at quite similar rates, making this network only implementable if HGF could be mo
to diffuse much more slowly. To decrease the necessity of differential diffusion, and mainta
original single promoter design, the network atlas reveals that including a positive feedback
on top of the receptor would suffice to make equal diffusivity accessible for given network
BC). This, however, though improving robustness with respect to extracellular parameter
not improve intracellular robustness significantly and thus in total, robustness is only marg
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct inter
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA
A n
Turing pattern generation. In reality, the network can be seen as a 3-node network in which c-Met
composes the central node and mediates the signals between both HGF and NK4 (see Fig 7 C).
With this model, the network equals the core topology #28 with nodes B and C diffusing. Indeed,
this topology is amongst the top 10 % of the most robust topologies according to our results. Even
so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely to
diffuse at quite similar rates, making this network only implementable if HGF could be modified
to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain the
original single promoter design, the network atlas reveals that including a positive feedback loop
on top of the receptor would suffice to make equal diffusivity accessible for given network (# 63
BC). This, however, though improving robustness with respect to extracellular parameters does
not improve intracellular robustness significantly and thus in total, robustness is only marginally
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interaction
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this shows
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA (1)
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB (2)
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA (3)
so, the classical requirements found for the 2-node system persist HGF and NK4 are more lik
diffuse at quite similar rates, making this network only implementable if HGF could be mo
to diffuse much more slowly. To decrease the necessity of differential diffusion, and mainta
original single promoter design, the network atlas reveals that including a positive feedbac
on top of the receptor would suffice to make equal diffusivity accessible for given network
BC). This, however, though improving robustness with respect to extracellular parameter
not improve intracellular robustness significantly and thus in total, robustness is only marg
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct inter
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA
@B
@t
= VB ·
( A
kAB
)n
1 + ( A
kAB
)n
+ bB µBB
2 3
Turing pattern generation. In reality, the network can be seen as a 3-node network in which
composes the central node and mediates the signals between both HGF and NK4 (see Fi
With this model, the network equals the core topology #28 with nodes B and C diffusing.
this topology is amongst the top 10 % of the most robust topologies according to our result
so, the classical requirements found for the 2-node system persist HGF and NK4 are more l
diffuse at quite similar rates, making this network only implementable if HGF could be m
to diffuse much more slowly. To decrease the necessity of differential diffusion, and maint
original single promoter design, the network atlas reveals that including a positive feedba
on top of the receptor would suffice to make equal diffusivity accessible for given network
BC). This, however, though improving robustness with respect to extracellular paramete
not improve intracellular robustness significantly and thus in total, robustness is only mar
improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct inte
between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, thi
how the network atlas can help decipher engineering options.
2 formulas
@A
@t
= VA ·
1
1 + (kAA
A )n
·
1
1 + ( B
kBA
)n
+ bA µAA
@B
@t
= VB ·
1
1 + (kAB
A )n
+ bB µBB
@A
@t
= VA ·
( A
kAA
)n
1 + ( A
kAA
)n + ( B
kBA
)n
+ bA µAA
A n
Turing Pattern are CommonA
B
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
complexity
activating inhibiting
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
C D Non-Competitive Competitive
ID
8
Level 1 1 2 3
61% of 3-node networks can generate
Turing Patterns.
But only 5 out of 21 2-node networks
can generate Turing Patterns
Turing Pattern are CommonA
B
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
complexity
activating inhibiting
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
C D Non-Competitive Competitive
ID
8
Level 1 1 2 3
61% of 3-node networks can generate
Turing Patterns.
But only 5 out of 21 2-node networks
can generate Turing Patterns
B
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
complexity
activating inhibiting
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
2-Node Turing Pattern Generating
Mechanisms
B
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
complexity
activating inhibiting
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
C D
complexity
A B A B
Non-Competitive Competitive
ID
8
9
15
16
20
Level 1
Level 2
Level 3
1 2
4 5
3
6
7 8 9 10 11
12 13 17 14 16
15 20 18 19 21
No pattern
No pattern
C D
complexity
A B A B
Non-Competitive Competitive
ID
8
9
15
16
20
Level 1
Level 2
Level 3
1 2
4 5
3
6
7 8 9 10 11
12 13 17 14 16
15 20 18 19 21
No pattern
No pattern
2-Node Turing Pattern Generating
Mechanisms
B
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
complexity
activating inhibiting
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
C D
complexity
A B A B
Non-Competitive Competitive
ID
8
9
15
16
20
Level 1
Level 2
Level 3
1 2
4 5
3
6
7 8 9 10 11
12 13 17 14 16
15 20 18 19 21
No pattern
No pattern
C D
complexity
A B A B
Non-Competitive Competitive
ID
8
9
15
16
20
Level 1
Level 2
Level 3
1 2
4 5
3
6
7 8 9 10 11
12 13 17 14 16
15 20 18 19 21
No pattern
No pattern
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
complexity
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
C D
complexity
A B A B
Non-Competitive Competitive
ID
8
9
15
16
20
Level 1
Level 2
Level 3
1 2
4 5
3
6
7 8 9 10 11
12 13 17 14 16
15 20 18 19 21
No pattern
No pattern
Turing Pattern are Common
We consider all possible 2– and
3–node networks and identify
shared core topologies that are
capable of generating Turing pat-
terns.
3
activating
inhibiting
-
D
A
B C
3774
A
B C
3772
B C
4564
B C
4147
B C
1291
B C
733
B C
67
4030 1372 475 126
A
B C
4147
A
B C
1291
A
B C
733
A
B C
67
high
low
concentration
A
B
C
Scholes et al., Cell Systems (2019)
A
B C
49
A
B C
50
A
B C
52
A
B C
53
A
B C
65
A
B C
86
A
B C
87
A
B C
93
A
B C
131
A
B C
132
A
B C
16
A
B C
17
A
B C
64
A
B C
67
A
B C
68
A
B C
92
A
B C
119
A
B C
120
A
B C
49
A
B C
50
A
B C
52
A
B C
53
A
B C
65
A
B C
86
A
B C
87
A
B C
93
A
B C
131
A
B C
132
A
B C
16
A
B C
17
A
B C
64
A
B C
67
A
B C
68
A
B C
92
A
B C
119
A
B C
120
A
B C
46
A
B C
47
A
B C
83
A
B C
84
A
B C
125
A
B C
126
A
B
Competitive Turing topologies
Non-Competitive Turing topologies
activating inhibiting
In-Phase Core Topologies
Out-of-Phase Core Topologies
Turing Pattern are Common, but Robust They are Not
A
B
Turing Pattern are Common, but Robust They are Not
A
B
Ann Babtie Paul Kirk
Turing Pattern are Common, but Robust They are Not
A
B
Evolutionary Robustness of Turing Pattern Generators
B
C
3
4
Top 10 Non-Competitive Networks:
activating
inhibiting
-
D
A
B C
4030
A
B C
1372
A
B C
1528
A
B C
4491
A
B C
3774
A
B C
3775
A
B C
4300
A
B C
3881
A
B C
1331
A
B C
3772
A
B C
5651
A
B C
4564
A
B C
4147
A
B C
1291
A
B C
733
A
B C
67
B C
5703
B C
4491
B C
4030
B C
1372
B C
475
B C
126
A
B C
5651
A
B C
4564
A
B C
4147
A
B C
1291
A
B C
733
A
B C
67
0
high
low
concentration
A
B
C
A
B C
A
B
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
complexity
activating inhibiting
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
C D
complexity
A B A B
Non-Competitive Competitive
ID
8
9
15
16
20
Level 1
Level 2
Level 3
1 2
4 5
3
6
7 8 9 10 11
12 13 17 14 16
15 20 18 19 21
No pattern
No pattern
B
C D
Top 10 Competitive Networks:
A
B C
1291
A
B C
733
A
B C
725
A
B C
1807
A
B C
4147
A
B C
1261
A
B C
1423
A
B C
726
A
B C
1474
A
B C
2924
B C
5651
B C
4564
B C
4147
0
A
B
C
A
B C
Frequency vs Robustness
There are more non-competitive than competitive networks that can generate Turing patterns.
However, competitive networks tend to be more robust.
Evolutionary Robustness of Turing Pattern Generators
B
C
3
4
Top 10 Non-Competitive Networks:
activating
inhibiting
-
D
A
B C
4030
A
B C
1372
A
B C
1528
A
B C
4491
A
B C
3774
A
B C
3775
A
B C
4300
A
B C
3881
A
B C
1331
A
B C
3772
A
B C
5651
A
B C
4564
A
B C
4147
A
B C
1291
A
B C
733
A
B C
67
B C
5703
B C
4491
B C
4030
B C
1372
B C
475
B C
126
A
B C
5651
A
B C
4564
A
B C
4147
A
B C
1291
A
B C
733
A
B C
67
0
high
low
concentration
A
B
C
A
B C
A
B
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
complexity
activating inhibiting
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
A B
1
A B
2
A B
3
A B
4
A B
5
A B
6
A B
7
A B
8
A B
9
A B
10
A B
11
A B
12
A B
13
A B
14
A B
15
A B
16
A B
17
A B
18
A B
19
A B
20
A B
21
C D
complexity
A B A B
Non-Competitive Competitive
ID
8
9
15
16
20
Level 1
Level 2
Level 3
1 2
4 5
3
6
7 8 9 10 11
12 13 17 14 16
15 20 18 19 21
No pattern
No pattern
B
C D
Top 10 Competitive Networks:
A
B C
1291
A
B C
733
A
B C
725
A
B C
1807
A
B C
4147
A
B C
1261
A
B C
1423
A
B C
726
A
B C
1474
A
B C
2924
B C
5651
B C
4564
B C
4147
0
A
B
C
A
B C
Frequency vs Robustness
There are more non-competitive than competitive networks that can generate Turing patterns.
However, competitive networks tend to be more robust.
Evolutionary Change
loss of interactions 3
gain of interaction 3
change of interaction 7
change of regulation 3/7
Simple prototypical Turing pattern generating mecha-
nisms are probably frequently encountered, but easily lost
as well.
Core-motifs can be stabilized by additional interactions
with external nodes.
Positional Information in Mathematical Terms
Schaerli et al., Nature Communications (2014)
There is a notable lack of rigorous mathemat-
ical definitions, especially in the early litera-
ture, related to positional information.
Turing Pattern vs Stripe Formation Mechanisms
10 most robust non-competitive networks
B
C
3
4
Top 10 Non-Competitive Networks:
activating
inhibiting
-
D
A
B C
4030
A
B C
1372
A
B C
1528
A
B C
4491
A
B C
3774
A
B C
3775
A
B C
4300
A
B C
3881
A
B C
1331
A
B C
3772
5651 4564 4147 1291 733 67
A
B C
5651
A
B C
4564
A
B C
4147
A
B C
1291
A
B C
733
A
B C
67
A
B C
5703
A
B C
4491
A
B C
4030
A
B C
1372
A
B C
475
A
B C
126
A
B C
5651
A
B C
4564
A
B C
4147
A
B C
1291
A
B C
733
A
B C
67
0
A
B
C
A
B C
10 most robust competitive networks
activating
inhibiting
D
3
4
5
6
-
op 10 Competitive Networks:
A
B C
1291
A
B C
733
A
B C
725
A
B C
1807
A
B C
4147
A
B C
1261
A
B C
1423
A
B C
726
A
B C
1474
A
B C
2924
B C
5651
B C
4564
B C
4147
B C
1291
B C
733
B C
67
A
B C
5703
A
B C
4491
A
B C
4030
A
B C
1372
A
B C
475
A
B C
126
A
B C
5651
A
B C
4564
A
B C
4147
A
B C
1291
A
B C
733
A
B C
67
A
B C
5703
A
B C
4491
A
B C
4030
A
B C
1372
A
B C
475
A
B C
126
A
B C
5651
A
B C
4564
A
B C
4147
A
B C
1291
A
B C
733
A
B C
67
A
B C
5703
A
B C
4491
A
B C
4030
A
B C
1372
A
B C
475
A
B C
126
A
B C
5651
A
B C
4564
A
B C
4147
A
B C
1291
A
B C
733
A
B C
67
0
Hierarchy of stripe forming networks
How Different are Models?
Positional Information Model
The morphogen concentrations r(x, t) and s(x, t) can be modelled as
𝜕r
𝜕t
= Dr∇2
r − krr − krs
𝜕s
𝜕t
= Ds∇2
s − kss − krs
Turing–Gierer–Meinhardt Model
This model can be represented mathematically as
𝜕r
𝜕t
= Dr∇2
r +
𝜌r2
s(1 + 𝜇rr2)
− krr + 𝜌r
𝜕s
𝜕t
= Ds∇2
s + 𝜌r2
− kss + 𝜌s
How Different are Models?
Positional Information Model
The morphogen concentrations r(x, t) and s(x, t) can be modelled as
𝜕r
𝜕t
= Dr∇2
r − krr − krs
𝜕s
𝜕t
= Ds∇2
s − kss − krs
with Dr, Ds diffusivities, and kr, ks degradation rates (with Neumann boundary conditions).
Turing–Gierer–Meinhardt Model
This model can be represented mathematically as
𝜕r
𝜕t
= Dr∇2
r +
𝜌r2
s(1 + 𝜇rr2)
− krr + 𝜌r
𝜕s
𝜕t
= Ds∇2
s + 𝜌r2
− kss + 𝜌s
with Dr, Ds diffusivities, 𝜌r, 𝜌s basal production rates, kr, ks degradation rates, and 𝜇r a saturation con-
stant; 𝜌 is the source density (zero flux boundary conditions).
Setting Out the Problem
Model Space Dimension 1
Model Space Dimension 2
M1
M2
M3
M4
M5
M6
M7
M8
M9
M10
M13
M11
M12
M17
M14
M15
M16
M18
M19
M20
Define a metric space, (ℳ, Δ), via a
distance function, Δ(Mi, Mj), over the
set of models, ℳ={M1, … , Mn}.
Setting Out the Problem
Model Space Dimension 1
Model Space Dimension 2
M1
M13
M3
M14
M15
M6
M7
M8
M9
M20
M2
M11
M12
M17
M4
M5
M16
M18
M19
M10
Define a metric space, (ℳ, Δ), via a
distance function, Δ(Mi, Mj), over the
set of models, ℳ={M1, … , Mn}.
Distances between Mathematical Models
To our surprise there was not much useful in the sense of distances between models. There were some
candidates:
We are looking for a distance not based on model outputs.
Candidate a priori Distances
Graph distances: Network representations of mathematical models
fail to capture fundamental aspects of the dynamics.
Semantic distances: This is promising route if we can describe differ-
ent models adequately and consistently.
… : There are a host of other candidates; most of these cannot
treat different models to the same degree.
Simplicial Complexes
dx
dt
= x(𝛼 − 𝛽y)
dy
dt
= y(𝛾x − 𝛿)
Prey Prey
growth
Predation Oscillatory
population
Predator Predator
growth
Predator
death
Prey Y Y Y
Prey growth Y Y Y
Predation Y Y Y
Oscillatory popu-
lation
Y Y Y Y Y Y
Predator Y Y Y Y
Predator growth Y Y Y Y
Predator death Y Y Y
Vittadello & Stumpf, Roy Soc Open Sci (2021).
Model Structures
𝜕r
𝜕t
= Dr∇2
r − krr − krs,
𝜕s
𝜕t
= Ds∇2
s − kss − krs,
v1 ⟷ Morphogen 1 (r)
v2 ⟷ Diffusion 1
v3 ⟷ Degradation 1
v6 ⟷ Influx 1
v9 ⟷ Morphogen 2 (s)
v10 ⟷ Diffusion 2
v11 ⟷ Degradation 2
v13 ⟷ Influx 2
v26 ⟷ Annihilation between morphogens 1 and 2
v40 ⟷ Monotonic gradient
v43 ⟷ Global scale-invariance
Model Structures and Topological Data Analysis
Model Distances
There is no clear way to separate
Turing Pattern models
from
Positional Information models.
Are PI and TP models fundamentally different?
Apparently not.
Can we get PI behaviour from a TP model or
vice versa?
Mathematically: Yes
Biologically: ?
Model Distances
There is no clear way to separate
Turing Pattern models
from
Positional Information models.
Are PI and TP models fundamentally different?
Apparently not.
Can we get PI behaviour from a TP model or
vice versa?
Mathematically: Yes
Biologically: ?
Equivalences ``≡"
A more useful concept to study differences is based on the concept of equivalence.
The key here is to identify which aspects that we may regard as essentially identical.
Then we can see if these equivalences correspond to biologically meaningful concepts or
interventions.
Equivalences ``≡"
A more useful concept to study differences is based on the concept of equivalence.
The key here is to identify which aspects that we may regard as essentially identical.
Then we can see if these equivalences correspond to biologically meaningful concepts or
interventions.
Example equivalences between PI and TPG Models
Positional information Turing Pattern
Influx 1 ⟺ {Basal production 1, Self-activation of morphogen 1}
Influx 2 ⟺ Basal production 2
Monotonic gradient ⟺ Periodic gradient field
Annihilation between
morphogens 1 and 2
⟺
{Activation of morphogen 2 by morphogen 1,
Inhibition of morphogen 1 by morphogen 2}
Model Equivalences
Patterning Mechanisms are Not that Different
It is possible to define a distance between models using simplicial complexes. This distance takes
into account the structure of the network, but also properties of the nodes.
Based on this distance there is no clear way to separate Turing pattern mechanisms and positional
information mechanisms.
Equivalences allow us to determine systematically where differences between models are only
superficial.
Vittadello &Stumpf, arXiv (2021); Scholes & Isalan, Curr Opin Chem Biol (2017).
Patterning Mechanisms are Not that Different
Scholes & Isalan, Curr Opin Chem Biol (2017)
Scholes et al., Cell Systems (2019) 9:243–257; Vittadello & Stumpf, Roy Soc Open Sci (2021)
8:211361; Leyshon et al., J Theo Biol (2021) 531:110901; Vittadello et al., Phil Trans Roy Soc
B (2021) 379:20200272.
Natalie Scholes
David Schnoerr
Sean Vittadello
Patterning Mechanisms are Not that Different
Scholes & Isalan, Curr Opin Chem Biol (2017)
Boundary conditions, not motif structures, are the biggest difference
between PI and TP systems.
Scholes et al., Cell Systems (2019) 9:243–257; Vittadello & Stumpf, Roy Soc Open Sci (2021)
8:211361; Leyshon et al., J Theo Biol (2021) 531:110901; Vittadello et al., Phil Trans Roy Soc
B (2021) 379:20200272.
Natalie Scholes
David Schnoerr
Sean Vittadello
Patterning Mechanisms are Not that Different
Scholes & Isalan, Curr Opin Chem Biol (2017)
Boundary conditions, not motif structures, are the biggest difference
between PI and TP systems.
Mathematical equivalences identify routes for interconversion between
PI systems and TP systems.
Scholes et al., Cell Systems (2019) 9:243–257; Vittadello & Stumpf, Roy Soc Open Sci (2021)
8:211361; Leyshon et al., J Theo Biol (2021) 531:110901; Vittadello et al., Phil Trans Roy Soc
B (2021) 379:20200272.
Natalie Scholes
David Schnoerr
Sean Vittadello
Patterning Mechanisms are Not that Different
Scholes & Isalan, Curr Opin Chem Biol (2017)
Boundary conditions, not motif structures, are the biggest difference
between PI and TP systems.
Mathematical equivalences identify routes for interconversion between
PI systems and TP systems.
Patterning systems are robust. The mechanism may change, however.
Scholes et al., Cell Systems (2019) 9:243–257; Vittadello & Stumpf, Roy Soc Open Sci (2021)
8:211361; Leyshon et al., J Theo Biol (2021) 531:110901; Vittadello et al., Phil Trans Roy Soc
B (2021) 379:20200272.
Natalie Scholes
David Schnoerr
Sean Vittadello

More Related Content

Similar to Design principles in pattern formation: Robustness and equivalences

Synthesis of a Sparse 2D-Scanning Array using Particle Swarm Optimization for...
Synthesis of a Sparse 2D-Scanning Array using Particle Swarm Optimization for...Synthesis of a Sparse 2D-Scanning Array using Particle Swarm Optimization for...
Synthesis of a Sparse 2D-Scanning Array using Particle Swarm Optimization for...Sivaranjan Goswami
 
Methods available in WIEN2k for the treatment of exchange and correlation ef...
Methods available in WIEN2k for the treatment  of exchange and correlation ef...Methods available in WIEN2k for the treatment  of exchange and correlation ef...
Methods available in WIEN2k for the treatment of exchange and correlation ef...ABDERRAHMANE REGGAD
 
Haplotype resolved structural variation assembly with long reads
Haplotype resolved structural variation assembly with long readsHaplotype resolved structural variation assembly with long reads
Haplotype resolved structural variation assembly with long readsGenome Reference Consortium
 
System overflow blocking-transients-for-queues-with-batch-arrivals-using-a-fa...
System overflow blocking-transients-for-queues-with-batch-arrivals-using-a-fa...System overflow blocking-transients-for-queues-with-batch-arrivals-using-a-fa...
System overflow blocking-transients-for-queues-with-batch-arrivals-using-a-fa...Cemal Ardil
 
Wideband Branch Line Coupler with Open Circuit Coupled Lines
Wideband Branch Line Coupler with Open Circuit  Coupled Lines Wideband Branch Line Coupler with Open Circuit  Coupled Lines
Wideband Branch Line Coupler with Open Circuit Coupled Lines IJECEIAES
 
Coupling Neural Networks to GCMs
Coupling Neural Networks to GCMsCoupling Neural Networks to GCMs
Coupling Neural Networks to GCMsNoah Brenowitz
 
Yangetal Efficient Letkf
Yangetal Efficient LetkfYangetal Efficient Letkf
Yangetal Efficient LetkfShuChih.Yang
 
Pham,Nhat_ResearchPoster
Pham,Nhat_ResearchPosterPham,Nhat_ResearchPoster
Pham,Nhat_ResearchPosterNhat Pham
 
Concurrent Ternary Galois-based Computation using Nano-apex Multiplexing Nibs...
Concurrent Ternary Galois-based Computation using Nano-apex Multiplexing Nibs...Concurrent Ternary Galois-based Computation using Nano-apex Multiplexing Nibs...
Concurrent Ternary Galois-based Computation using Nano-apex Multiplexing Nibs...VLSICS Design
 
Practical Active Filter Design
Practical Active Filter Design Practical Active Filter Design
Practical Active Filter Design Sachin Mehta
 
CONCURRENT TERNARY GALOIS-BASED COMPUTATION USING NANO-APEX MULTIPLEXING NIBS...
CONCURRENT TERNARY GALOIS-BASED COMPUTATION USING NANO-APEX MULTIPLEXING NIBS...CONCURRENT TERNARY GALOIS-BASED COMPUTATION USING NANO-APEX MULTIPLEXING NIBS...
CONCURRENT TERNARY GALOIS-BASED COMPUTATION USING NANO-APEX MULTIPLEXING NIBS...VLSICS Design
 
CONCURRENT TERNARY GALOIS-BASED COMPUTATION USING NANO-APEX MULTIPLEXING NIBS...
CONCURRENT TERNARY GALOIS-BASED COMPUTATION USING NANO-APEX MULTIPLEXING NIBS...CONCURRENT TERNARY GALOIS-BASED COMPUTATION USING NANO-APEX MULTIPLEXING NIBS...
CONCURRENT TERNARY GALOIS-BASED COMPUTATION USING NANO-APEX MULTIPLEXING NIBS...VLSICS Design
 
Bharat 5th DRC ppt_ nuclear fusion theoretical calculations
Bharat 5th DRC ppt_ nuclear fusion theoretical calculationsBharat 5th DRC ppt_ nuclear fusion theoretical calculations
Bharat 5th DRC ppt_ nuclear fusion theoretical calculationsBharatKashyap30
 
PCA-CompChem_seminar
PCA-CompChem_seminarPCA-CompChem_seminar
PCA-CompChem_seminarAnne D'cruz
 
Formulating Evolutionary Dynamics of Organism-Environment Couplings Using Gra...
Formulating Evolutionary Dynamics of Organism-Environment Couplings Using Gra...Formulating Evolutionary Dynamics of Organism-Environment Couplings Using Gra...
Formulating Evolutionary Dynamics of Organism-Environment Couplings Using Gra...Hiroki Sayama
 

Similar to Design principles in pattern formation: Robustness and equivalences (20)

Synthesis of a Sparse 2D-Scanning Array using Particle Swarm Optimization for...
Synthesis of a Sparse 2D-Scanning Array using Particle Swarm Optimization for...Synthesis of a Sparse 2D-Scanning Array using Particle Swarm Optimization for...
Synthesis of a Sparse 2D-Scanning Array using Particle Swarm Optimization for...
 
Lgm saarbrucken
Lgm saarbruckenLgm saarbrucken
Lgm saarbrucken
 
Methods available in WIEN2k for the treatment of exchange and correlation ef...
Methods available in WIEN2k for the treatment  of exchange and correlation ef...Methods available in WIEN2k for the treatment  of exchange and correlation ef...
Methods available in WIEN2k for the treatment of exchange and correlation ef...
 
Haplotype resolved structural variation assembly with long reads
Haplotype resolved structural variation assembly with long readsHaplotype resolved structural variation assembly with long reads
Haplotype resolved structural variation assembly with long reads
 
1406
14061406
1406
 
System overflow blocking-transients-for-queues-with-batch-arrivals-using-a-fa...
System overflow blocking-transients-for-queues-with-batch-arrivals-using-a-fa...System overflow blocking-transients-for-queues-with-batch-arrivals-using-a-fa...
System overflow blocking-transients-for-queues-with-batch-arrivals-using-a-fa...
 
Wideband Branch Line Coupler with Open Circuit Coupled Lines
Wideband Branch Line Coupler with Open Circuit  Coupled Lines Wideband Branch Line Coupler with Open Circuit  Coupled Lines
Wideband Branch Line Coupler with Open Circuit Coupled Lines
 
Coupling Neural Networks to GCMs
Coupling Neural Networks to GCMsCoupling Neural Networks to GCMs
Coupling Neural Networks to GCMs
 
Yangetal Efficient Letkf
Yangetal Efficient LetkfYangetal Efficient Letkf
Yangetal Efficient Letkf
 
Pham,Nhat_ResearchPoster
Pham,Nhat_ResearchPosterPham,Nhat_ResearchPoster
Pham,Nhat_ResearchPoster
 
Concurrent Ternary Galois-based Computation using Nano-apex Multiplexing Nibs...
Concurrent Ternary Galois-based Computation using Nano-apex Multiplexing Nibs...Concurrent Ternary Galois-based Computation using Nano-apex Multiplexing Nibs...
Concurrent Ternary Galois-based Computation using Nano-apex Multiplexing Nibs...
 
Practical Active Filter Design
Practical Active Filter Design Practical Active Filter Design
Practical Active Filter Design
 
Rare B strangeness decay
Rare B strangeness decayRare B strangeness decay
Rare B strangeness decay
 
CONCURRENT TERNARY GALOIS-BASED COMPUTATION USING NANO-APEX MULTIPLEXING NIBS...
CONCURRENT TERNARY GALOIS-BASED COMPUTATION USING NANO-APEX MULTIPLEXING NIBS...CONCURRENT TERNARY GALOIS-BASED COMPUTATION USING NANO-APEX MULTIPLEXING NIBS...
CONCURRENT TERNARY GALOIS-BASED COMPUTATION USING NANO-APEX MULTIPLEXING NIBS...
 
CONCURRENT TERNARY GALOIS-BASED COMPUTATION USING NANO-APEX MULTIPLEXING NIBS...
CONCURRENT TERNARY GALOIS-BASED COMPUTATION USING NANO-APEX MULTIPLEXING NIBS...CONCURRENT TERNARY GALOIS-BASED COMPUTATION USING NANO-APEX MULTIPLEXING NIBS...
CONCURRENT TERNARY GALOIS-BASED COMPUTATION USING NANO-APEX MULTIPLEXING NIBS...
 
Bharat 5th DRC ppt_ nuclear fusion theoretical calculations
Bharat 5th DRC ppt_ nuclear fusion theoretical calculationsBharat 5th DRC ppt_ nuclear fusion theoretical calculations
Bharat 5th DRC ppt_ nuclear fusion theoretical calculations
 
Ashg2017 workshop tg
Ashg2017 workshop tgAshg2017 workshop tg
Ashg2017 workshop tg
 
AGBT2017 Reference Workshop: Lindsay
AGBT2017 Reference Workshop: LindsayAGBT2017 Reference Workshop: Lindsay
AGBT2017 Reference Workshop: Lindsay
 
PCA-CompChem_seminar
PCA-CompChem_seminarPCA-CompChem_seminar
PCA-CompChem_seminar
 
Formulating Evolutionary Dynamics of Organism-Environment Couplings Using Gra...
Formulating Evolutionary Dynamics of Organism-Environment Couplings Using Gra...Formulating Evolutionary Dynamics of Organism-Environment Couplings Using Gra...
Formulating Evolutionary Dynamics of Organism-Environment Couplings Using Gra...
 

Recently uploaded

Manassas R - Parkside Middle School 🌎🏫
Manassas R - Parkside Middle School 🌎🏫Manassas R - Parkside Middle School 🌎🏫
Manassas R - Parkside Middle School 🌎🏫qfactory1
 
Transposable elements in prokaryotes.ppt
Transposable elements in prokaryotes.pptTransposable elements in prokaryotes.ppt
Transposable elements in prokaryotes.pptArshadWarsi13
 
Twin's paradox experiment is a meassurement of the extra dimensions.pptx
Twin's paradox experiment is a meassurement of the extra dimensions.pptxTwin's paradox experiment is a meassurement of the extra dimensions.pptx
Twin's paradox experiment is a meassurement of the extra dimensions.pptxEran Akiva Sinbar
 
Dashanga agada a formulation of Agada tantra dealt in 3 Rd year bams agada tanta
Dashanga agada a formulation of Agada tantra dealt in 3 Rd year bams agada tantaDashanga agada a formulation of Agada tantra dealt in 3 Rd year bams agada tanta
Dashanga agada a formulation of Agada tantra dealt in 3 Rd year bams agada tantaPraksha3
 
Temporomandibular joint Muscles of Mastication
Temporomandibular joint Muscles of MasticationTemporomandibular joint Muscles of Mastication
Temporomandibular joint Muscles of Masticationvidulajaib
 
Module 4: Mendelian Genetics and Punnett Square
Module 4:  Mendelian Genetics and Punnett SquareModule 4:  Mendelian Genetics and Punnett Square
Module 4: Mendelian Genetics and Punnett SquareIsiahStephanRadaza
 
THE ROLE OF PHARMACOGNOSY IN TRADITIONAL AND MODERN SYSTEM OF MEDICINE.pptx
THE ROLE OF PHARMACOGNOSY IN TRADITIONAL AND MODERN SYSTEM OF MEDICINE.pptxTHE ROLE OF PHARMACOGNOSY IN TRADITIONAL AND MODERN SYSTEM OF MEDICINE.pptx
THE ROLE OF PHARMACOGNOSY IN TRADITIONAL AND MODERN SYSTEM OF MEDICINE.pptxNandakishor Bhaurao Deshmukh
 
Harmful and Useful Microorganisms Presentation
Harmful and Useful Microorganisms PresentationHarmful and Useful Microorganisms Presentation
Harmful and Useful Microorganisms Presentationtahreemzahra82
 
Heredity: Inheritance and Variation of Traits
Heredity: Inheritance and Variation of TraitsHeredity: Inheritance and Variation of Traits
Heredity: Inheritance and Variation of TraitsCharlene Llagas
 
Bentham & Hooker's Classification. along with the merits and demerits of the ...
Bentham & Hooker's Classification. along with the merits and demerits of the ...Bentham & Hooker's Classification. along with the merits and demerits of the ...
Bentham & Hooker's Classification. along with the merits and demerits of the ...Nistarini College, Purulia (W.B) India
 
Is RISC-V ready for HPC workload? Maybe?
Is RISC-V ready for HPC workload? Maybe?Is RISC-V ready for HPC workload? Maybe?
Is RISC-V ready for HPC workload? Maybe?Patrick Diehl
 
Call Girls in Aiims Metro Delhi 💯Call Us 🔝9953322196🔝 💯Escort.
Call Girls in Aiims Metro Delhi 💯Call Us 🔝9953322196🔝 💯Escort.Call Girls in Aiims Metro Delhi 💯Call Us 🔝9953322196🔝 💯Escort.
Call Girls in Aiims Metro Delhi 💯Call Us 🔝9953322196🔝 💯Escort.aasikanpl
 
Grafana in space: Monitoring Japan's SLIM moon lander in real time
Grafana in space: Monitoring Japan's SLIM moon lander  in real timeGrafana in space: Monitoring Japan's SLIM moon lander  in real time
Grafana in space: Monitoring Japan's SLIM moon lander in real timeSatoshi NAKAHIRA
 
Call Girls In Nihal Vihar Delhi ❤️8860477959 Looking Escorts In 24/7 Delhi NCR
Call Girls In Nihal Vihar Delhi ❤️8860477959 Looking Escorts In 24/7 Delhi NCRCall Girls In Nihal Vihar Delhi ❤️8860477959 Looking Escorts In 24/7 Delhi NCR
Call Girls In Nihal Vihar Delhi ❤️8860477959 Looking Escorts In 24/7 Delhi NCRlizamodels9
 
Speech, hearing, noise, intelligibility.pptx
Speech, hearing, noise, intelligibility.pptxSpeech, hearing, noise, intelligibility.pptx
Speech, hearing, noise, intelligibility.pptxpriyankatabhane
 
SOLUBLE PATTERN RECOGNITION RECEPTORS.pptx
SOLUBLE PATTERN RECOGNITION RECEPTORS.pptxSOLUBLE PATTERN RECOGNITION RECEPTORS.pptx
SOLUBLE PATTERN RECOGNITION RECEPTORS.pptxkessiyaTpeter
 
Spermiogenesis or Spermateleosis or metamorphosis of spermatid
Spermiogenesis or Spermateleosis or metamorphosis of spermatidSpermiogenesis or Spermateleosis or metamorphosis of spermatid
Spermiogenesis or Spermateleosis or metamorphosis of spermatidSarthak Sekhar Mondal
 
Call Girls in Munirka Delhi 💯Call Us 🔝9953322196🔝 💯Escort.
Call Girls in Munirka Delhi 💯Call Us 🔝9953322196🔝 💯Escort.Call Girls in Munirka Delhi 💯Call Us 🔝9953322196🔝 💯Escort.
Call Girls in Munirka Delhi 💯Call Us 🔝9953322196🔝 💯Escort.aasikanpl
 
Evidences of Evolution General Biology 2
Evidences of Evolution General Biology 2Evidences of Evolution General Biology 2
Evidences of Evolution General Biology 2John Carlo Rollon
 

Recently uploaded (20)

Manassas R - Parkside Middle School 🌎🏫
Manassas R - Parkside Middle School 🌎🏫Manassas R - Parkside Middle School 🌎🏫
Manassas R - Parkside Middle School 🌎🏫
 
Transposable elements in prokaryotes.ppt
Transposable elements in prokaryotes.pptTransposable elements in prokaryotes.ppt
Transposable elements in prokaryotes.ppt
 
Twin's paradox experiment is a meassurement of the extra dimensions.pptx
Twin's paradox experiment is a meassurement of the extra dimensions.pptxTwin's paradox experiment is a meassurement of the extra dimensions.pptx
Twin's paradox experiment is a meassurement of the extra dimensions.pptx
 
Dashanga agada a formulation of Agada tantra dealt in 3 Rd year bams agada tanta
Dashanga agada a formulation of Agada tantra dealt in 3 Rd year bams agada tantaDashanga agada a formulation of Agada tantra dealt in 3 Rd year bams agada tanta
Dashanga agada a formulation of Agada tantra dealt in 3 Rd year bams agada tanta
 
Temporomandibular joint Muscles of Mastication
Temporomandibular joint Muscles of MasticationTemporomandibular joint Muscles of Mastication
Temporomandibular joint Muscles of Mastication
 
Module 4: Mendelian Genetics and Punnett Square
Module 4:  Mendelian Genetics and Punnett SquareModule 4:  Mendelian Genetics and Punnett Square
Module 4: Mendelian Genetics and Punnett Square
 
THE ROLE OF PHARMACOGNOSY IN TRADITIONAL AND MODERN SYSTEM OF MEDICINE.pptx
THE ROLE OF PHARMACOGNOSY IN TRADITIONAL AND MODERN SYSTEM OF MEDICINE.pptxTHE ROLE OF PHARMACOGNOSY IN TRADITIONAL AND MODERN SYSTEM OF MEDICINE.pptx
THE ROLE OF PHARMACOGNOSY IN TRADITIONAL AND MODERN SYSTEM OF MEDICINE.pptx
 
Harmful and Useful Microorganisms Presentation
Harmful and Useful Microorganisms PresentationHarmful and Useful Microorganisms Presentation
Harmful and Useful Microorganisms Presentation
 
Heredity: Inheritance and Variation of Traits
Heredity: Inheritance and Variation of TraitsHeredity: Inheritance and Variation of Traits
Heredity: Inheritance and Variation of Traits
 
Bentham & Hooker's Classification. along with the merits and demerits of the ...
Bentham & Hooker's Classification. along with the merits and demerits of the ...Bentham & Hooker's Classification. along with the merits and demerits of the ...
Bentham & Hooker's Classification. along with the merits and demerits of the ...
 
Is RISC-V ready for HPC workload? Maybe?
Is RISC-V ready for HPC workload? Maybe?Is RISC-V ready for HPC workload? Maybe?
Is RISC-V ready for HPC workload? Maybe?
 
Call Girls in Aiims Metro Delhi 💯Call Us 🔝9953322196🔝 💯Escort.
Call Girls in Aiims Metro Delhi 💯Call Us 🔝9953322196🔝 💯Escort.Call Girls in Aiims Metro Delhi 💯Call Us 🔝9953322196🔝 💯Escort.
Call Girls in Aiims Metro Delhi 💯Call Us 🔝9953322196🔝 💯Escort.
 
Grafana in space: Monitoring Japan's SLIM moon lander in real time
Grafana in space: Monitoring Japan's SLIM moon lander  in real timeGrafana in space: Monitoring Japan's SLIM moon lander  in real time
Grafana in space: Monitoring Japan's SLIM moon lander in real time
 
Call Girls In Nihal Vihar Delhi ❤️8860477959 Looking Escorts In 24/7 Delhi NCR
Call Girls In Nihal Vihar Delhi ❤️8860477959 Looking Escorts In 24/7 Delhi NCRCall Girls In Nihal Vihar Delhi ❤️8860477959 Looking Escorts In 24/7 Delhi NCR
Call Girls In Nihal Vihar Delhi ❤️8860477959 Looking Escorts In 24/7 Delhi NCR
 
Speech, hearing, noise, intelligibility.pptx
Speech, hearing, noise, intelligibility.pptxSpeech, hearing, noise, intelligibility.pptx
Speech, hearing, noise, intelligibility.pptx
 
SOLUBLE PATTERN RECOGNITION RECEPTORS.pptx
SOLUBLE PATTERN RECOGNITION RECEPTORS.pptxSOLUBLE PATTERN RECOGNITION RECEPTORS.pptx
SOLUBLE PATTERN RECOGNITION RECEPTORS.pptx
 
Spermiogenesis or Spermateleosis or metamorphosis of spermatid
Spermiogenesis or Spermateleosis or metamorphosis of spermatidSpermiogenesis or Spermateleosis or metamorphosis of spermatid
Spermiogenesis or Spermateleosis or metamorphosis of spermatid
 
Hot Sexy call girls in Moti Nagar,🔝 9953056974 🔝 escort Service
Hot Sexy call girls in  Moti Nagar,🔝 9953056974 🔝 escort ServiceHot Sexy call girls in  Moti Nagar,🔝 9953056974 🔝 escort Service
Hot Sexy call girls in Moti Nagar,🔝 9953056974 🔝 escort Service
 
Call Girls in Munirka Delhi 💯Call Us 🔝9953322196🔝 💯Escort.
Call Girls in Munirka Delhi 💯Call Us 🔝9953322196🔝 💯Escort.Call Girls in Munirka Delhi 💯Call Us 🔝9953322196🔝 💯Escort.
Call Girls in Munirka Delhi 💯Call Us 🔝9953322196🔝 💯Escort.
 
Evidences of Evolution General Biology 2
Evidences of Evolution General Biology 2Evidences of Evolution General Biology 2
Evidences of Evolution General Biology 2
 

Design principles in pattern formation: Robustness and equivalences

  • 1. Design principles in pattern formation: Robustness and equivalences Michael P.H. Stumpf Theoretical Systems Biology, Melbourne Integrative Genomics School of BioSciences & School of Mathematics and Statistics
  • 2. Engineering Turing Patterns Development- tal Biology Systems Biology Synthetic Biology Mathematical Modelling James Briscoe Michael Stumpf Heike Siebert Mark Isalan Statistical Inference Model Reduction Molecular Biology Control Engineering Multi-scale Modelling Genome Editing In Silico Design Stem Cell Biology Bio- informatics Natalie Scholes David Schnoerr Sean Vittadello
  • 3. Turing Patterns Alan Turing A B Local Activation & Lateral Inhibition
  • 4. Turing Patterns are Robust Gierer-Meinhardt Model (A) (B) (D) (C) Vittadello et al., Phil Trans Roy Soc B (2021); Leyshon et al., J Theo Biol (2021).
  • 5. Positional Information ( aka French Flag Model) n or bition Concentration T1 T2 Concentration Space Positional information creating the French Flag pattern B Wolpert – PI i Green & Sharpe, Development (2015) Lewis Wolpert
  • 10. Why the Acrimony? Stumpf†, Nature (1966) Raspopovic et al., Science (2014) †Hildegard F. Stumpf (1924-1967) – No relation.
  • 11. Turing Patterns vs Positional Information The simple dichotomy is wrong. There is room and need for both mechanisms. We are interested in The design principles of Turing Pat- terns. The design principles of Positional Information Mechanisms. The differences between the two sets of design principles. Concentration Space Concentration Space Concentration Space Concentration Space Concentration Space Auto-activation Activation of inhibitor Lateral inhibition of activator Diffusion of inhibitor Consistent ‘wavelength’ controlled by global parameters i ii iii iv v Molecular fluctuations Concentration T1 T2 Concentration Space Positional information creating the French Flag pattern T7 Concentration Space T6 T5 T4 T3 T2 T1 B Wolpert – PI i ii Positional information creating a periodic pattern Green & Sharpe, Development (2015)
  • 12. Turing Generating Mechanisms and Their Properties We considered > 5,000 network structures. > 1011 parameter combina- tions. We sampled neighbouring 4–node motifs. Scholes et al., Cell Systems (2019)
  • 13. Turing Generating Mechanisms and Their Properties We need two ingredients: Non-spatial dynamics allow for a non-trivial stationary state. Spatial dynamics (≡ spatial dynamics but with diffusion of molecules turned on) are unstable. C A “Fast inhibitor” “Slow activator” Non- homogeneous initial state Reaction & Diffusion Pattern formation time time 1 A B 2 A B 3 A B 4 A B 5 8 A B 9 A B 10 A B 11 A B 12 15 A B 16 A B 17 A B 18 A B 19 E Non-Competitive time activating inhibiting D Topology B @B @t = VB · 1 1 + (kAB A )n + bB µBB (2) @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA (3) @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB (4) J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 (5) @A @t = ↵ + · A2 B µAA (6) @B @t = · A2 µBB (7) 2 @B @t = VB · 1 1 + (kAB A )n + bB µBB (2) @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA (3) @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB (4) J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 (5) @A @t = ↵ + · A2 B µAA (6) @B @t = · A2 µBB (7) 2
  • 14. B 2-Node: 3-Node: 0 1 -1 Reduce according to symmetry Only consider connectednetworks … Adjacency matrix representation Translating networks into ODE equations 1 1 −1 0 C Estimating Turing Pattern capability Competitive Non-Competitive Graph representation Parameter Range V 0.1-100 Specify parameter boundaries Stability analysis: A B A B 1 A B 2 A B 3 A B 4 A 5 A B 6 A B 7 A B 8 A B 9 A 10 A B 11 A B 12 A B 13 A B 14 A 15 A B 16 A B 17 A B 18 A B 19 A 20 A B 21 A B C composes the central node and mediates the signals between both HGF and NK4 (see Fig 7 C). With this model, the network equals the core topology #28 with nodes B and C diffusing. Indeed, this topology is amongst the top 10 % of the most robust topologies according to our results. Even so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely to diffuse at quite similar rates, making this network only implementable if HGF could be modified to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain the original single promoter design, the network atlas reveals that including a positive feedback loop on top of the receptor would suffice to make equal diffusivity accessible for given network (# 63 BC). This, however, though improving robustness with respect to extracellular parameters does not improve intracellular robustness significantly and thus in total, robustness is only marginally improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interaction between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this shows how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA (1) @B @t = VB · 1 1 + (kAB A )n + bB µBB (2) @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA (3) @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB (4) 2@fA D q2 @fA 3 so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely t diffuse at quite similar rates, making this network only implementable if HGF could be modifie to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain th original single promoter design, the network atlas reveals that including a positive feedback loo on top of the receptor would suffice to make equal diffusivity accessible for given network (# 6 BC). This, however, though improving robustness with respect to extracellular parameters do not improve intracellular robustness significantly and thus in total, robustness is only marginal improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interactio between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this show how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA (1 @B @t = VB · 1 1 + (kAB A )n + bB µBB (2 @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA (3 @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB (4 J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 (5 so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely to diffuse at quite similar rates, making this network only implementable if HGF could be modified to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain the original single promoter design, the network atlas reveals that including a positive feedback loop on top of the receptor would suffice to make equal diffusivity accessible for given network (# 63 BC). This, however, though improving robustness with respect to extracellular parameters does not improve intracellular robustness significantly and thus in total, robustness is only marginally improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interaction between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this shows how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA (1) @B @t = VB · 1 1 + (kAB A )n + bB µBB (2) @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA (3) @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB (4) J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 (5) on top of the receptor would suffice to make equal diffusivity accessible for given network (# 6 BC). This, however, though improving robustness with respect to extracellular parameters do not improve intracellular robustness significantly and thus in total, robustness is only marginal improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interactio between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this show how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA ( @B @t = VB · 1 1 + (kAB A )n + bB µBB ( @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA ( @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB ( J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 ( @A @t = ⇢ · A2 B µAA ( so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely diffuse at quite similar rates, making this network only implementable if HGF could be modifi to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain original single promoter design, the network atlas reveals that including a positive feedback lo on top of the receptor would suffice to make equal diffusivity accessible for given network (# BC). This, however, though improving robustness with respect to extracellular parameters d not improve intracellular robustness significantly and thus in total, robustness is only margina improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interact between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this sho how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA @B @t = VB · 1 1 + (kAB A )n + bB µBB @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 Initial conditions:
  • 15. B 2-Node: 3-Node: 0 1 -1 Reduce according to symmetry Only consider connectednetworks … Adjacency matrix representation Translating networks into ODE equations 1 1 −1 0 C Estimating Turing Pattern capability Competitive Non-Competitive Graph representation Parameter Range V 0.1-100 Specify parameter boundaries Stability analysis: A B A B 1 A B 2 A B 3 A B 4 A 5 A B 6 A B 7 A B 8 A B 9 A 10 A B 11 A B 12 A B 13 A B 14 A 15 A B 16 A B 17 A B 18 A B 19 A 20 A B 21 A B C composes the central node and mediates the signals between both HGF and NK4 (see Fig 7 C). With this model, the network equals the core topology #28 with nodes B and C diffusing. Indeed, this topology is amongst the top 10 % of the most robust topologies according to our results. Even so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely to diffuse at quite similar rates, making this network only implementable if HGF could be modified to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain the original single promoter design, the network atlas reveals that including a positive feedback loop on top of the receptor would suffice to make equal diffusivity accessible for given network (# 63 BC). This, however, though improving robustness with respect to extracellular parameters does not improve intracellular robustness significantly and thus in total, robustness is only marginally improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interaction between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this shows how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA (1) @B @t = VB · 1 1 + (kAB A )n + bB µBB (2) @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA (3) @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB (4) 2@fA D q2 @fA 3 so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely t diffuse at quite similar rates, making this network only implementable if HGF could be modifie to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain th original single promoter design, the network atlas reveals that including a positive feedback loo on top of the receptor would suffice to make equal diffusivity accessible for given network (# 6 BC). This, however, though improving robustness with respect to extracellular parameters do not improve intracellular robustness significantly and thus in total, robustness is only marginal improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interactio between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this show how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA (1 @B @t = VB · 1 1 + (kAB A )n + bB µBB (2 @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA (3 @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB (4 J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 (5 so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely to diffuse at quite similar rates, making this network only implementable if HGF could be modified to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain the original single promoter design, the network atlas reveals that including a positive feedback loop on top of the receptor would suffice to make equal diffusivity accessible for given network (# 63 BC). This, however, though improving robustness with respect to extracellular parameters does not improve intracellular robustness significantly and thus in total, robustness is only marginally improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interaction between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this shows how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA (1) @B @t = VB · 1 1 + (kAB A )n + bB µBB (2) @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA (3) @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB (4) J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 (5) on top of the receptor would suffice to make equal diffusivity accessible for given network (# 6 BC). This, however, though improving robustness with respect to extracellular parameters do not improve intracellular robustness significantly and thus in total, robustness is only marginal improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interactio between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this show how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA ( @B @t = VB · 1 1 + (kAB A )n + bB µBB ( @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA ( @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB ( J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 ( @A @t = ⇢ · A2 B µAA ( so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely diffuse at quite similar rates, making this network only implementable if HGF could be modifi to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain original single promoter design, the network atlas reveals that including a positive feedback lo on top of the receptor would suffice to make equal diffusivity accessible for given network (# BC). This, however, though improving robustness with respect to extracellular parameters d not improve intracellular robustness significantly and thus in total, robustness is only margina improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interact between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this sho how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA @B @t = VB · 1 1 + (kAB A )n + bB µBB @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 Initial conditions: C qmax 2π /qmax A “Fast inhibitor” “Slow activator” Non- homogeneous initial state Reaction & Diffusion Pattern formation time time 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 E Competitive Non-Competitive time activating inhibiting D Topology Regulatory function Function parameters Diffusion constants B @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB (4) J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 (5) @A @t = ↵ + · A2 B µAA (6) @B @t = · A2 µBB (7) 2 @B @t = VB · (kAB ) 1 + ( A kAB )n + bB µBB (4) J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 (5) @A @t = ↵ + · A2 B µAA (6) @B @t = · A2 µBB (7) 2
  • 16. B 2-Node: 3-Node: 0 1 -1 Reduce according to symmetry Only consider connectednetworks … Adjacency matrix representation Translating networks into ODE equations 1 1 −1 0 C Estimating Turing Pattern capability Competitive Non-Competitive Graph representation Parameter Range V 0.1-100 Specify parameter boundaries Stability analysis: A B A B 1 A B 2 A B 3 A B 4 A 5 A B 6 A B 7 A B 8 A B 9 A 10 A B 11 A B 12 A B 13 A B 14 A 15 A B 16 A B 17 A B 18 A B 19 A 20 A B 21 A B C composes the central node and mediates the signals between both HGF and NK4 (see Fig 7 C). With this model, the network equals the core topology #28 with nodes B and C diffusing. Indeed, this topology is amongst the top 10 % of the most robust topologies according to our results. Even so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely to diffuse at quite similar rates, making this network only implementable if HGF could be modified to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain the original single promoter design, the network atlas reveals that including a positive feedback loop on top of the receptor would suffice to make equal diffusivity accessible for given network (# 63 BC). This, however, though improving robustness with respect to extracellular parameters does not improve intracellular robustness significantly and thus in total, robustness is only marginally improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interaction between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this shows how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA (1) @B @t = VB · 1 1 + (kAB A )n + bB µBB (2) @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA (3) @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB (4) 2@fA D q2 @fA 3 so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely t diffuse at quite similar rates, making this network only implementable if HGF could be modifie to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain th original single promoter design, the network atlas reveals that including a positive feedback loo on top of the receptor would suffice to make equal diffusivity accessible for given network (# 6 BC). This, however, though improving robustness with respect to extracellular parameters do not improve intracellular robustness significantly and thus in total, robustness is only marginal improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interactio between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this show how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA (1 @B @t = VB · 1 1 + (kAB A )n + bB µBB (2 @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA (3 @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB (4 J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 (5 so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely to diffuse at quite similar rates, making this network only implementable if HGF could be modified to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain the original single promoter design, the network atlas reveals that including a positive feedback loop on top of the receptor would suffice to make equal diffusivity accessible for given network (# 63 BC). This, however, though improving robustness with respect to extracellular parameters does not improve intracellular robustness significantly and thus in total, robustness is only marginally improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interaction between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this shows how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA (1) @B @t = VB · 1 1 + (kAB A )n + bB µBB (2) @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA (3) @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB (4) J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 (5) on top of the receptor would suffice to make equal diffusivity accessible for given network (# 6 BC). This, however, though improving robustness with respect to extracellular parameters do not improve intracellular robustness significantly and thus in total, robustness is only marginal improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interactio between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this show how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA ( @B @t = VB · 1 1 + (kAB A )n + bB µBB ( @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA ( @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB ( J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 ( @A @t = ⇢ · A2 B µAA ( so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely diffuse at quite similar rates, making this network only implementable if HGF could be modifi to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain original single promoter design, the network atlas reveals that including a positive feedback lo on top of the receptor would suffice to make equal diffusivity accessible for given network (# BC). This, however, though improving robustness with respect to extracellular parameters d not improve intracellular robustness significantly and thus in total, robustness is only margina improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interact between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this sho how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA @B @t = VB · 1 1 + (kAB A )n + bB µBB @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 Initial conditions: C qmax 2π /qmax A “Fast inhibitor” “Slow activator” Non- homogeneous initial state Reaction & Diffusion Pattern formation time time 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 E Competitive Non-Competitive time activating inhibiting D Topology Regulatory function Function parameters Diffusion constants B @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB (4) J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 (5) @A @t = ↵ + · A2 B µAA (6) @B @t = · A2 µBB (7) 2 @B @t = VB · (kAB ) 1 + ( A kAB )n + bB µBB (4) J = 2 4 @fA @A D1q2 @fA @B @fB @A @fB @B D2q2 3 5 (5) @A @t = ↵ + · A2 B µAA (6) @B @t = · A2 µBB (7) 2 B A Defining network structure 2-Node: 3-Node: 0 1 -1 Reduce according to symmetry Only consider connectednetworks … Adjacency matrix representation Translating networks into ODE equations 1 1 −1 0 Competitive Non-Competitive Graph representation A B A B 1 A B 2 A B 3 A B 4 A A B 6 A B 7 A B 8 A B 9 A A B 11 A B 12 A B 13 A B 14 A A B 16 A B 17 A B 18 A B 19 A A B 21 A B C MMP-1 promoter construct. In a previous study, this system was analyzed as a classical 2-node system and the authors suggested that a single promoter construct driving the expression of both activating and inhibiting species could suffice to create Turing Patterns [?]. This system would, however, require differential diffusion (10 fold) and a co-operativity factor of >= 2 for feasible Turing pattern generation. In reality, the network can be seen as a 3-node network in which c-Met composes the central node and mediates the signals between both HGF and NK4 (see Fig 7 C). With this model, the network equals the core topology #28 with nodes B and C diffusing. Indeed, this topology is amongst the top 10 % of the most robust topologies according to our results. Even so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely to diffuse at quite similar rates, making this network only implementable if HGF could be modified to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain the original single promoter design, the network atlas reveals that including a positive feedback loop on top of the receptor would suffice to make equal diffusivity accessible for given network (# 63 BC). This, however, though improving robustness with respect to extracellular parameters does not improve intracellular robustness significantly and thus in total, robustness is only marginally improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interaction between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this shows how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA (1) @B @t = VB · 1 1 + (kAB A )n + bB µBB (2) @A ( A k )n Turing pattern generation. In reality, the network can be seen as a 3-node network in which composes the central node and mediates the signals between both HGF and NK4 (see Fig With this model, the network equals the core topology #28 with nodes B and C diffusing. In this topology is amongst the top 10 % of the most robust topologies according to our results. so, the classical requirements found for the 2-node system persist HGF and NK4 are more lik diffuse at quite similar rates, making this network only implementable if HGF could be mo to diffuse much more slowly. To decrease the necessity of differential diffusion, and mainta original single promoter design, the network atlas reveals that including a positive feedback on top of the receptor would suffice to make equal diffusivity accessible for given network BC). This, however, though improving robustness with respect to extracellular parameter not improve intracellular robustness significantly and thus in total, robustness is only marg improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct inter between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA @B @t = VB · 1 1 + (kAB A )n + bB µBB @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA A n Turing pattern generation. In reality, the network can be seen as a 3-node network in which c-Met composes the central node and mediates the signals between both HGF and NK4 (see Fig 7 C). With this model, the network equals the core topology #28 with nodes B and C diffusing. Indeed, this topology is amongst the top 10 % of the most robust topologies according to our results. Even so, the classical requirements found for the 2-node system persist HGF and NK4 are more likely to diffuse at quite similar rates, making this network only implementable if HGF could be modified to diffuse much more slowly. To decrease the necessity of differential diffusion, and maintain the original single promoter design, the network atlas reveals that including a positive feedback loop on top of the receptor would suffice to make equal diffusivity accessible for given network (# 63 BC). This, however, though improving robustness with respect to extracellular parameters does not improve intracellular robustness significantly and thus in total, robustness is only marginally improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct interaction between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this shows how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA (1) @B @t = VB · 1 1 + (kAB A )n + bB µBB (2) @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA (3) so, the classical requirements found for the 2-node system persist HGF and NK4 are more lik diffuse at quite similar rates, making this network only implementable if HGF could be mo to diffuse much more slowly. To decrease the necessity of differential diffusion, and mainta original single promoter design, the network atlas reveals that including a positive feedbac on top of the receptor would suffice to make equal diffusivity accessible for given network BC). This, however, though improving robustness with respect to extracellular parameter not improve intracellular robustness significantly and thus in total, robustness is only marg improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct inter between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, this how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA @B @t = VB · 1 1 + (kAB A )n + bB µBB @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA @B @t = VB · ( A kAB )n 1 + ( A kAB )n + bB µBB 2 3 Turing pattern generation. In reality, the network can be seen as a 3-node network in which composes the central node and mediates the signals between both HGF and NK4 (see Fi With this model, the network equals the core topology #28 with nodes B and C diffusing. this topology is amongst the top 10 % of the most robust topologies according to our result so, the classical requirements found for the 2-node system persist HGF and NK4 are more l diffuse at quite similar rates, making this network only implementable if HGF could be m to diffuse much more slowly. To decrease the necessity of differential diffusion, and maint original single promoter design, the network atlas reveals that including a positive feedba on top of the receptor would suffice to make equal diffusivity accessible for given network BC). This, however, though improving robustness with respect to extracellular paramete not improve intracellular robustness significantly and thus in total, robustness is only mar improved (1.3 fold). To significantly improve the design (4.6 fold), an additional direct inte between HGF and NK4 would have to be engineered (NK4 activating HGF). Overall, thi how the network atlas can help decipher engineering options. 2 formulas @A @t = VA · 1 1 + (kAA A )n · 1 1 + ( B kBA )n + bA µAA @B @t = VB · 1 1 + (kAB A )n + bB µBB @A @t = VA · ( A kAA )n 1 + ( A kAA )n + ( B kBA )n + bA µAA A n
  • 17. Turing Pattern are CommonA B A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 complexity activating inhibiting A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 C D Non-Competitive Competitive ID 8 Level 1 1 2 3 61% of 3-node networks can generate Turing Patterns. But only 5 out of 21 2-node networks can generate Turing Patterns
  • 18. Turing Pattern are CommonA B A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 complexity activating inhibiting A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 C D Non-Competitive Competitive ID 8 Level 1 1 2 3 61% of 3-node networks can generate Turing Patterns. But only 5 out of 21 2-node networks can generate Turing Patterns B A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 complexity activating inhibiting A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21
  • 19. 2-Node Turing Pattern Generating Mechanisms B A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 complexity activating inhibiting A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 C D complexity A B A B Non-Competitive Competitive ID 8 9 15 16 20 Level 1 Level 2 Level 3 1 2 4 5 3 6 7 8 9 10 11 12 13 17 14 16 15 20 18 19 21 No pattern No pattern C D complexity A B A B Non-Competitive Competitive ID 8 9 15 16 20 Level 1 Level 2 Level 3 1 2 4 5 3 6 7 8 9 10 11 12 13 17 14 16 15 20 18 19 21 No pattern No pattern
  • 20. 2-Node Turing Pattern Generating Mechanisms B A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 complexity activating inhibiting A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 C D complexity A B A B Non-Competitive Competitive ID 8 9 15 16 20 Level 1 Level 2 Level 3 1 2 4 5 3 6 7 8 9 10 11 12 13 17 14 16 15 20 18 19 21 No pattern No pattern C D complexity A B A B Non-Competitive Competitive ID 8 9 15 16 20 Level 1 Level 2 Level 3 1 2 4 5 3 6 7 8 9 10 11 12 13 17 14 16 15 20 18 19 21 No pattern No pattern A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 complexity A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 C D complexity A B A B Non-Competitive Competitive ID 8 9 15 16 20 Level 1 Level 2 Level 3 1 2 4 5 3 6 7 8 9 10 11 12 13 17 14 16 15 20 18 19 21 No pattern No pattern
  • 21. Turing Pattern are Common We consider all possible 2– and 3–node networks and identify shared core topologies that are capable of generating Turing pat- terns. 3 activating inhibiting - D A B C 3774 A B C 3772 B C 4564 B C 4147 B C 1291 B C 733 B C 67 4030 1372 475 126 A B C 4147 A B C 1291 A B C 733 A B C 67 high low concentration A B C Scholes et al., Cell Systems (2019) A B C 49 A B C 50 A B C 52 A B C 53 A B C 65 A B C 86 A B C 87 A B C 93 A B C 131 A B C 132 A B C 16 A B C 17 A B C 64 A B C 67 A B C 68 A B C 92 A B C 119 A B C 120 A B C 49 A B C 50 A B C 52 A B C 53 A B C 65 A B C 86 A B C 87 A B C 93 A B C 131 A B C 132 A B C 16 A B C 17 A B C 64 A B C 67 A B C 68 A B C 92 A B C 119 A B C 120 A B C 46 A B C 47 A B C 83 A B C 84 A B C 125 A B C 126 A B Competitive Turing topologies Non-Competitive Turing topologies activating inhibiting In-Phase Core Topologies Out-of-Phase Core Topologies
  • 22. Turing Pattern are Common, but Robust They are Not A B
  • 23. Turing Pattern are Common, but Robust They are Not A B Ann Babtie Paul Kirk
  • 24. Turing Pattern are Common, but Robust They are Not A B
  • 25. Evolutionary Robustness of Turing Pattern Generators B C 3 4 Top 10 Non-Competitive Networks: activating inhibiting - D A B C 4030 A B C 1372 A B C 1528 A B C 4491 A B C 3774 A B C 3775 A B C 4300 A B C 3881 A B C 1331 A B C 3772 A B C 5651 A B C 4564 A B C 4147 A B C 1291 A B C 733 A B C 67 B C 5703 B C 4491 B C 4030 B C 1372 B C 475 B C 126 A B C 5651 A B C 4564 A B C 4147 A B C 1291 A B C 733 A B C 67 0 high low concentration A B C A B C A B A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 complexity activating inhibiting A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 C D complexity A B A B Non-Competitive Competitive ID 8 9 15 16 20 Level 1 Level 2 Level 3 1 2 4 5 3 6 7 8 9 10 11 12 13 17 14 16 15 20 18 19 21 No pattern No pattern B C D Top 10 Competitive Networks: A B C 1291 A B C 733 A B C 725 A B C 1807 A B C 4147 A B C 1261 A B C 1423 A B C 726 A B C 1474 A B C 2924 B C 5651 B C 4564 B C 4147 0 A B C A B C Frequency vs Robustness There are more non-competitive than competitive networks that can generate Turing patterns. However, competitive networks tend to be more robust.
  • 26. Evolutionary Robustness of Turing Pattern Generators B C 3 4 Top 10 Non-Competitive Networks: activating inhibiting - D A B C 4030 A B C 1372 A B C 1528 A B C 4491 A B C 3774 A B C 3775 A B C 4300 A B C 3881 A B C 1331 A B C 3772 A B C 5651 A B C 4564 A B C 4147 A B C 1291 A B C 733 A B C 67 B C 5703 B C 4491 B C 4030 B C 1372 B C 475 B C 126 A B C 5651 A B C 4564 A B C 4147 A B C 1291 A B C 733 A B C 67 0 high low concentration A B C A B C A B A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 complexity activating inhibiting A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 A B 1 A B 2 A B 3 A B 4 A B 5 A B 6 A B 7 A B 8 A B 9 A B 10 A B 11 A B 12 A B 13 A B 14 A B 15 A B 16 A B 17 A B 18 A B 19 A B 20 A B 21 C D complexity A B A B Non-Competitive Competitive ID 8 9 15 16 20 Level 1 Level 2 Level 3 1 2 4 5 3 6 7 8 9 10 11 12 13 17 14 16 15 20 18 19 21 No pattern No pattern B C D Top 10 Competitive Networks: A B C 1291 A B C 733 A B C 725 A B C 1807 A B C 4147 A B C 1261 A B C 1423 A B C 726 A B C 1474 A B C 2924 B C 5651 B C 4564 B C 4147 0 A B C A B C Frequency vs Robustness There are more non-competitive than competitive networks that can generate Turing patterns. However, competitive networks tend to be more robust. Evolutionary Change loss of interactions 3 gain of interaction 3 change of interaction 7 change of regulation 3/7 Simple prototypical Turing pattern generating mecha- nisms are probably frequently encountered, but easily lost as well. Core-motifs can be stabilized by additional interactions with external nodes.
  • 27. Positional Information in Mathematical Terms Schaerli et al., Nature Communications (2014) There is a notable lack of rigorous mathemat- ical definitions, especially in the early litera- ture, related to positional information.
  • 28. Turing Pattern vs Stripe Formation Mechanisms 10 most robust non-competitive networks B C 3 4 Top 10 Non-Competitive Networks: activating inhibiting - D A B C 4030 A B C 1372 A B C 1528 A B C 4491 A B C 3774 A B C 3775 A B C 4300 A B C 3881 A B C 1331 A B C 3772 5651 4564 4147 1291 733 67 A B C 5651 A B C 4564 A B C 4147 A B C 1291 A B C 733 A B C 67 A B C 5703 A B C 4491 A B C 4030 A B C 1372 A B C 475 A B C 126 A B C 5651 A B C 4564 A B C 4147 A B C 1291 A B C 733 A B C 67 0 A B C A B C 10 most robust competitive networks activating inhibiting D 3 4 5 6 - op 10 Competitive Networks: A B C 1291 A B C 733 A B C 725 A B C 1807 A B C 4147 A B C 1261 A B C 1423 A B C 726 A B C 1474 A B C 2924 B C 5651 B C 4564 B C 4147 B C 1291 B C 733 B C 67 A B C 5703 A B C 4491 A B C 4030 A B C 1372 A B C 475 A B C 126 A B C 5651 A B C 4564 A B C 4147 A B C 1291 A B C 733 A B C 67 A B C 5703 A B C 4491 A B C 4030 A B C 1372 A B C 475 A B C 126 A B C 5651 A B C 4564 A B C 4147 A B C 1291 A B C 733 A B C 67 A B C 5703 A B C 4491 A B C 4030 A B C 1372 A B C 475 A B C 126 A B C 5651 A B C 4564 A B C 4147 A B C 1291 A B C 733 A B C 67 0 Hierarchy of stripe forming networks
  • 29. How Different are Models? Positional Information Model The morphogen concentrations r(x, t) and s(x, t) can be modelled as 𝜕r 𝜕t = Dr∇2 r − krr − krs 𝜕s 𝜕t = Ds∇2 s − kss − krs Turing–Gierer–Meinhardt Model This model can be represented mathematically as 𝜕r 𝜕t = Dr∇2 r + 𝜌r2 s(1 + 𝜇rr2) − krr + 𝜌r 𝜕s 𝜕t = Ds∇2 s + 𝜌r2 − kss + 𝜌s
  • 30. How Different are Models? Positional Information Model The morphogen concentrations r(x, t) and s(x, t) can be modelled as 𝜕r 𝜕t = Dr∇2 r − krr − krs 𝜕s 𝜕t = Ds∇2 s − kss − krs with Dr, Ds diffusivities, and kr, ks degradation rates (with Neumann boundary conditions). Turing–Gierer–Meinhardt Model This model can be represented mathematically as 𝜕r 𝜕t = Dr∇2 r + 𝜌r2 s(1 + 𝜇rr2) − krr + 𝜌r 𝜕s 𝜕t = Ds∇2 s + 𝜌r2 − kss + 𝜌s with Dr, Ds diffusivities, 𝜌r, 𝜌s basal production rates, kr, ks degradation rates, and 𝜇r a saturation con- stant; 𝜌 is the source density (zero flux boundary conditions).
  • 31. Setting Out the Problem Model Space Dimension 1 Model Space Dimension 2 M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 M13 M11 M12 M17 M14 M15 M16 M18 M19 M20 Define a metric space, (ℳ, Δ), via a distance function, Δ(Mi, Mj), over the set of models, ℳ={M1, … , Mn}.
  • 32. Setting Out the Problem Model Space Dimension 1 Model Space Dimension 2 M1 M13 M3 M14 M15 M6 M7 M8 M9 M20 M2 M11 M12 M17 M4 M5 M16 M18 M19 M10 Define a metric space, (ℳ, Δ), via a distance function, Δ(Mi, Mj), over the set of models, ℳ={M1, … , Mn}.
  • 33. Distances between Mathematical Models To our surprise there was not much useful in the sense of distances between models. There were some candidates: We are looking for a distance not based on model outputs. Candidate a priori Distances Graph distances: Network representations of mathematical models fail to capture fundamental aspects of the dynamics. Semantic distances: This is promising route if we can describe differ- ent models adequately and consistently. … : There are a host of other candidates; most of these cannot treat different models to the same degree.
  • 34. Simplicial Complexes dx dt = x(𝛼 − 𝛽y) dy dt = y(𝛾x − 𝛿) Prey Prey growth Predation Oscillatory population Predator Predator growth Predator death Prey Y Y Y Prey growth Y Y Y Predation Y Y Y Oscillatory popu- lation Y Y Y Y Y Y Predator Y Y Y Y Predator growth Y Y Y Y Predator death Y Y Y Vittadello & Stumpf, Roy Soc Open Sci (2021).
  • 35. Model Structures 𝜕r 𝜕t = Dr∇2 r − krr − krs, 𝜕s 𝜕t = Ds∇2 s − kss − krs, v1 ⟷ Morphogen 1 (r) v2 ⟷ Diffusion 1 v3 ⟷ Degradation 1 v6 ⟷ Influx 1 v9 ⟷ Morphogen 2 (s) v10 ⟷ Diffusion 2 v11 ⟷ Degradation 2 v13 ⟷ Influx 2 v26 ⟷ Annihilation between morphogens 1 and 2 v40 ⟷ Monotonic gradient v43 ⟷ Global scale-invariance
  • 36. Model Structures and Topological Data Analysis
  • 37. Model Distances There is no clear way to separate Turing Pattern models from Positional Information models. Are PI and TP models fundamentally different? Apparently not. Can we get PI behaviour from a TP model or vice versa? Mathematically: Yes Biologically: ?
  • 38. Model Distances There is no clear way to separate Turing Pattern models from Positional Information models. Are PI and TP models fundamentally different? Apparently not. Can we get PI behaviour from a TP model or vice versa? Mathematically: Yes Biologically: ?
  • 39. Equivalences ``≡" A more useful concept to study differences is based on the concept of equivalence. The key here is to identify which aspects that we may regard as essentially identical. Then we can see if these equivalences correspond to biologically meaningful concepts or interventions.
  • 40. Equivalences ``≡" A more useful concept to study differences is based on the concept of equivalence. The key here is to identify which aspects that we may regard as essentially identical. Then we can see if these equivalences correspond to biologically meaningful concepts or interventions. Example equivalences between PI and TPG Models Positional information Turing Pattern Influx 1 ⟺ {Basal production 1, Self-activation of morphogen 1} Influx 2 ⟺ Basal production 2 Monotonic gradient ⟺ Periodic gradient field Annihilation between morphogens 1 and 2 ⟺ {Activation of morphogen 2 by morphogen 1, Inhibition of morphogen 1 by morphogen 2}
  • 42. Patterning Mechanisms are Not that Different It is possible to define a distance between models using simplicial complexes. This distance takes into account the structure of the network, but also properties of the nodes. Based on this distance there is no clear way to separate Turing pattern mechanisms and positional information mechanisms. Equivalences allow us to determine systematically where differences between models are only superficial. Vittadello &Stumpf, arXiv (2021); Scholes & Isalan, Curr Opin Chem Biol (2017).
  • 43. Patterning Mechanisms are Not that Different Scholes & Isalan, Curr Opin Chem Biol (2017) Scholes et al., Cell Systems (2019) 9:243–257; Vittadello & Stumpf, Roy Soc Open Sci (2021) 8:211361; Leyshon et al., J Theo Biol (2021) 531:110901; Vittadello et al., Phil Trans Roy Soc B (2021) 379:20200272. Natalie Scholes David Schnoerr Sean Vittadello
  • 44. Patterning Mechanisms are Not that Different Scholes & Isalan, Curr Opin Chem Biol (2017) Boundary conditions, not motif structures, are the biggest difference between PI and TP systems. Scholes et al., Cell Systems (2019) 9:243–257; Vittadello & Stumpf, Roy Soc Open Sci (2021) 8:211361; Leyshon et al., J Theo Biol (2021) 531:110901; Vittadello et al., Phil Trans Roy Soc B (2021) 379:20200272. Natalie Scholes David Schnoerr Sean Vittadello
  • 45. Patterning Mechanisms are Not that Different Scholes & Isalan, Curr Opin Chem Biol (2017) Boundary conditions, not motif structures, are the biggest difference between PI and TP systems. Mathematical equivalences identify routes for interconversion between PI systems and TP systems. Scholes et al., Cell Systems (2019) 9:243–257; Vittadello & Stumpf, Roy Soc Open Sci (2021) 8:211361; Leyshon et al., J Theo Biol (2021) 531:110901; Vittadello et al., Phil Trans Roy Soc B (2021) 379:20200272. Natalie Scholes David Schnoerr Sean Vittadello
  • 46. Patterning Mechanisms are Not that Different Scholes & Isalan, Curr Opin Chem Biol (2017) Boundary conditions, not motif structures, are the biggest difference between PI and TP systems. Mathematical equivalences identify routes for interconversion between PI systems and TP systems. Patterning systems are robust. The mechanism may change, however. Scholes et al., Cell Systems (2019) 9:243–257; Vittadello & Stumpf, Roy Soc Open Sci (2021) 8:211361; Leyshon et al., J Theo Biol (2021) 531:110901; Vittadello et al., Phil Trans Roy Soc B (2021) 379:20200272. Natalie Scholes David Schnoerr Sean Vittadello