coupled dynamical systems (coupled map, chaotic itinerancy,…) magic number 7+-2 dynamical systems...
TRANSCRIPT
• Coupled Dynamical Systems
(Coupled map, chaotic itinerancy,…)
magic number 7+-2
dynamical systems inspired by biology
• Cell Model with chemical networks
(universal statistical properties, evolution)
selection of dyn.systems through evolution biology study based on dyn.sys, statistical phys.
• Coupled Cell Dynamics
cell differentiation and development
biology study based on coupled dyn.sys.With growth
? Q: entanglement versus module:
Is there some limit in degrees of freedom so that the system behaves ‘’nicely ‘’?
background of this question;
(1) Magic number 7 +- 2 in psychology
--- number limit of chunk for a short-term memory, (e.g., phone number) Muller
(2) Use of modularity in biological systems
degrees of freedom are often clustered
??is there some critical number of degrees of freedom beyond which the ‘memory’ (separation into attractors) by dynamical system is unstable??
feed-forward network, globally coupled map
underlying universal logic?
Example 1:Feed-forward thresholdNetwork
M elements each layer
Neural network;Gene network with thresholdDynamics
Effect of entanglement?
Each element +1,-1 if β >1Attracted to +1,-1 dep. on input
input
Ishihara and KK 200 5 PRL
Random coupling
regarded as tim
e
• Outputs are clearly separated to +1, -1 for most inputs for most networks (coupling ε )
for M ≦ 6
・ For M ≧ 7、 outputs can take intermediate values. Chaotic Dynamics is observed along the layer
(stretching and folding dynamics are observed commonly)
・ This critical number ~ 7 is rather universal
The average fraction of x (with some bin), where orbits visited (i.e. x such that P(x)>0)
Model with zero threshold
Model with distributed threshold
The ratio of networks that show chaotic behaviorPlotted as a function of M
Computed from Lyapunov exp.
Why?
consider just 2-layer dynamics by first
fixing input x2,x3,..xM and study the map
from the input x1 to the output x1
σ 1>0 σ 2<0
σ 3>0
ν 1 ν 2 ν 3
Let us renumber ν as orderingν 1< ν 2<…Then, if σ 1、 σ 2、… changes itsSign alternately folding Process like logistic map if satisfied at all i=1,2,,,M-1 、 then folding at every region
x(inp)
X(out)
• probability to satisfy such sign changes
----- 2^{-M} • By change of input values x_2,…x_M
there can be a variety of ( M-1)! for such ordering for ν(using original numbering ν)
If >1 then, the folding occurs at every region at least for somewhere in the layer chaos dominates cover the phase space
‘’If Combinatorial Variety wins exponential growth,
then the clear separation by threshold dynamics
is not possible!”
Crossover occurs at M =6
Cluster: group of elements such that x(i)=x(j); Number of elements in each cluster; N1,N2,…,Nk•at some parameter region many attractors with different clusteringsDue to the symmetry there are
attractors of the same clusterings -- combinatorially many increase with the order of (N-1)! or so (KK,PRL89)
Another example: GCM
a
ε
• At some parameter region (partially ordered phase) Milnor attractors occupy large portion of basin (KK,PRL97 、 PhysicaD98)
• Milnor attractor ( i.e., Attractor in the sense of Milnor minus usual attractor with asymptotic stability); attractor and its basin boundary touches, i.e., any small perturbation from it can kick the orbit
out of the attractor, while it has a finite measure of basin ( orbits from many initial conditions are attracted to it)
Milnor attractor network chaotic itinerancy
The fraction of basin(i.e. initial values) forMilnor attractors,Plotted as a function ofLogistic map parameter
Note! Fraction is almost1 for some region
Result for N=10,50,100…. a
1
The Milnor attractors become dominant around N > ~ (5-8)
DependenceOn theNumber ofElements N
(accumulation over 1.55<a<1.72)
(kk、 PRE,200 2 )
• Why? Conjecture by combinatorial explosion of basin boundaries
Simple separation x(i)>x* or x(i)<x*; one can separate 2 ^N attractors by N planes.
In this case the distance between attractor and the basin boundary does not change with N ---- Order of (N-1)!
The boundary makes combinatorial explosion
• The number of basin boundary planes has combinatorial explosion, as factorial wins over exponential ( around N ≒ 7) . Then, the basin boundary is ‘crowded’ in the phase space. Thus often attractors touch with basin boundaries
dominance of Milnor attractors here (complete symmetry is unnecessary)These two examples show that a separation of attractors
cannot work well when combinatorial variety wins over exponential increase of the phase space.
This explains psychological magic number 7 + - 2 ?
If elements more than 7 are entangled, clear separation behavior is difficult
A solution could be use of modules with each <7 elements
Constructive Biology Project : to understand basic biological features
theme experiment theory question
replicating
system
in vitro
replication with
enzymatic reaction
minority
control
origin of
heredity;
evolvability
cell
system
replicating cell
with internal
reactions
universal
statistics in
reaction dynamics
condition for
recursive
growth
cell differentiation.
development
differentiation
of E Coil
by interaction
emergence of
differentiation
rule from dynamics
irreversibility
robustness
Spontaneous
adaptationArtificial gene network
Adaptive attractor selection by noise
Robust adaptation without signalling
evolutionRelevance of
phenotypic fluctuation and dynamics
Genetic assimilation of phenotype fluct.an
d dynamics
geno-pheno type
relationship
ERATO Project Complex Systems Biology (2004 -2010, wiith Tetsuya Yomo (experimentalist)
How recursive production of a cell is sustained ? each cell complex reaction network
with diversity of chemicals;The number of molecules of each
species not so large
●●● ●●●●●●●●●●●
●●●● ●●●●●●●●
●●● ●●●●●●●●●●●
●●●● ●●●●●●●●
●●●● ●●●●●●●●
●●● ●●●●●●●● ●●
?Fluctuations
Universal characteristics as reproducingDynamical system
Toy Cell Model with Catalytic Reaction Network
‘Crude but a cell model that reproduces’
( resource)
reaction
catalyze
cell
medium
diffusion
k k species of chemicals species of chemicals 、、 XXoo……XX k-1k-1 number ---nnumber ---n 00 、 、 nn 11 … … nn k-1k-1
some chemicals are some chemicals are penetrablepenetrable
through the membrane with the through the membrane with the diffusion coefficient Ddiffusion coefficient D
resource chemicals are thus transforresource chemicals are thus transformed into impenetrable chemicals, leadimed into impenetrable chemicals, leading to the growth inng to the growth in
N= Σni, when it exceeds N when it exceeds Nmaxmax
the cell divides into twothe cell divides into two
random catalytic reaction networkrandom catalytic reaction network
with the path rate pwith the path rate p
for the reaction for the reaction XX ii +X+X jj ->X->X kk
+X+Xjj
modelmodel
C.Furusawa & KK 、 PRL2003
・・・ K >>1 species
dX1/dt X0X4; rate equation;∝Stochastic model here
(Cf. KK&Yomo 94,97)
In continuum description, the following rate eqn., but we mostly use stochastic simulation
Michalel-Menten kinetics, higher order catalysis same results
☆☆Growth speed and fidelity in replication Growth speed and fidelity in replication are maximum at Dcare maximum at Dc
0.00
0.05
0.10
0.15
0.20
0.001 0.01 0.10
0.2
0.4
0.6
0.8
1
grow
th sp
eed
(a.u
.)
simila
rity
H
diffusion coefficient D
growth speedsimilarity
D = Dc
※similarity is defined from inner products of composition vectors between mother and daughter cells
・・
・・
D
Dc
No Growth
(only resources)
Growth
Zipf’s Law is oberved at D = Dc
nnii (number(number ofof moleculesmolecules ))
rankrank
Furusawa &KK,2003,PRL
Average number of each chemical ∝ 1/(its rank)
(distribution of x : ρ (x)∝x )-2
number rankX1 300 5X2 8000 1X3 5000 2X4 700 4X5 2000 3…….. (for example)
1e-05
1e-04
1e-03
1e-02
1e-01
1e+00 1e+01 1e+02 1e+03
ratio
of m
RNA
rank
Human Liver
Human Kidney
Human Heart
Human Colorectal Cancer
1e-05
1e-04
1e-03
1e-02
1e-01
1e+00 1e+01 1e+02 1e+03 1e+04
ratio
of m
RNA
rank
1e-04
1e-03
1e-02
1e-01
1e+00 1e+01 1e+02 1e+03 1e+04
ratio
of m
RNA
rank
1e-04
1e-03
1e-02
1e-01
1e+00 1e+01 1e+02 1e+03ra
tio of
mRN
Arank
α= -1
Confirmed by gene expression data
Mouse ES cell
C. elegans
Mouse Fibroblast Cell
Yeast
1e-05
1e-04
1e-03
1e-02
1e-01
1e+00 1e+01 1e+02 1e+03 1e+04
ratio
of m
RNA
rank
1e-04
1e-03
1e-02
1e-01
1e+00 1e+01 1e+02 1e+03ra
tio o
f mRN
Arank
1e-05
1e-04
1e-03
1e-02
1e+00 1e+01 1e+02 1e+03 1e+04
ratio
of m
RNA
rank
1e-04
1e-03
1e-02
1e-01
1e+00 1e+01 1e+02 1e+03
ratio
of m
RNA
rank
α= -1
Formation of cascade catalytic reactionFormation of cascade catalytic reaction
Rank of nRank of ni i
Catalyze chemicals of higher rank mainly
1
10
100
1000
10000
100000
1 10 100 1000 10000 100000
栄養から直接生成される成分の数を多くなるようパラメータを変えて、階層構造が判りやすくした例
1122
3344
1:1: minority moleculesminority molecules
2:2: catalyzed by 1, synthesized by resourcecatalyzed by 1, synthesized by resource
3:3: catalyzed by 2catalyzed by 2
: :
★★
nini
With conservation law,The exponent -1 is explained
Mean-field type (self-consistent) calc.)
Evolution of Network to satisfy Zipf’s law? Yes Critical D value depends on connectivity in the network; mutation of network + selection approaches Zipf’s law
Furusawa,kk
Later, the connectivity in the network approaches “scale-free” network through evolution.statistical properties; embedded into network structure
Zipf’s law holds, irrespective of network structure, but
Furusawa 、 KK, submitted
initial
evolved
42
evolutionary embedding of dynamics into network
Dynamics (abundance) first, structure (equation for dynamics) later
probability for a path to chemical with abundances x is selected; q(x) transformation of abundance distrb. to connectivity distrib.
So far average quantity of all components;
Next question: fluctuation by cells: distribution of each Ni by cells
Each color givesdifferent chemical species
LOG SCALE
Furusawa,..KK,Biophysics2005
Log normal distribution !
e.g. cell1 X1 10000 cell2 8000 cell3 15000 cell4 20000….. histograrm
Experiment; protein abundances measured by fluorescence
Log-normalDistributionConfirmed experimentally
Furusawa,Kashiwagi,,Yomo,KK
+flow-cytometry
Also studied in GFP synthesis in liposome
☆☆Heuristic explanation of log-normal distributionHeuristic explanation of log-normal distributionConsider the case that a component X is catalyzed by Consider the case that a component X is catalyzed by other component A, and replicate; the number --other component A, and replicate; the number --NNXX 、、 NNAA
d Nd NXX /dt = N /dt = NXX N NAA
thenthen
d log( Nd log( NXX )/dt = N )/dt = NAA
IfIf 、 、 NNA A fluctuates around its mean < NNAA > , with fluct. η(t) d log( Nd log( NXX )/dt = )/dt = < NNAA > + + η (t)
log( Nlog( NXX ) shows Brownian motion ) shows Brownian motion NNX X log-normal distributionlog-normal distribution
too, simplified, since no direct self-replication exists here
But with cascade catalytic reactions, fluctuations are successively multiplied, (cf addition in central limit theorem.);Hence after logarithm, central limit th. applied
Significance of fluctuations phenotypic fluctuation ∝ evolution speed (experime
nt, cell model) extending fluctuation dissipation theorem in physics (cf)recall in standard evolutionary genetics, only the distribution of gene is discussed, by assuming
unique phenotype from a given genotype--in terms of dynamical systems genetic fluctuation (mutation): change in equation of dy
namical systems, parameter,Phenotype: variable according to the dynamical system
s (with fluctuations)Relationship! From evolutionary stability Sato et al., (PNAS, 2003), KK&Furusawa JTB2005
Cell differentiation: Isologous Diversification: (KK,Yomo1997)
internal dynamics and interaction : development phenotype
instability
distinct phenotypes
interaction-induced
Example: chemical reaction network
specialize in the use of some path
1 2( , ,.., )m
km
dxf x x x
dt
Universal characteristics as Coupledreproducing dynamical systems Assuming oscillatory reaction dynamics (gene expression) Study of coupled dynamical systems (globally coupled map) etc., +GROWTH differentiation??
GROWTH
Dilution by the increaseOf cell volume
Catalytic reaction dynamics (or gene expression)
Diffusion to/from media; resource included
synchronous division: no differentiation
Instability of homogeneous statethrough cell-cell interaction
formation of discrete types with different chemical compositions: stabilize each other
recursive production
Assuming oscillatory dynamics as a single cell
Hierarchical differentiation from ‘stem cell’;by taking initially dynamics with instability (e.g., chaotic)
(higher order catalysis) Furusawa&KK 98 BMathBiol
Hierarchical differentiation from ‘stem cell’;by taking initially dynamics with instability (e.g., chaotic)(higher order catalysis)
Irreversible loss of omnipotency characterized by the decrease in chemical diversity and phenotype variation
Generated Rule of Differentiation (example) A1 A A2 A3 S B
(1)hierarchical differentiation: stem cell system(2) Stochastic Branching: stochastic model proposed in hematopoietic system(3) probability depends on # distrib. of cell types with prob. pA for S A if #(A) then pA ―― global info. is embedded into internal cell statesSTABILITY
(4) Differentiation of cell ensemble (tissue) ――multiple stable distrib. { Ni }
Cf hematopoietic system
Stem cell= chaotic Milnor attractor hypothesis?
As long as the stem cell state is stable, it reproduces itself
With the increase of the number, it becomes unstable differentiate to other types
If the number of differentiated cells increases
then the stability of the stem cell is recovered,
and it reproduces itself
Stem cell stays at a marginally stable state
*recall that gene network mutually connected,
with more than 7(?) elements, easily expected
Universality? checked a huge number of networks; only some fraction of them show chaotic dynamics & differentiation
Cells with such networks with differentiation higher growth speed as an ensemble
Such networks are selected Furusawa,KK 2001PRL
Explained:
Robustness in development under large fluctuation in molecule numbers
Recall: (signal) molecules of few number -- relevant;
Loss of potency from totipotent cell (ES), to multipotent stem cell, and to determination
Irreversibility in cell differentiation process characterized by the loss of phenotypic variation
• Irreversibility: Loss of multipotency
• diversity in chemicals
• stability of the cellular state
• growth speed
Prediction:gene expression of stem cells diverse, weakregulation of proliferation/differentiation dep. on cell distributionIf #(stem) proliferation decreased externally then proliferation increases
• Summary:(1) Magic number 7?: crossover in (globally coup
led) dynamical systems at that number, combinatorial-exponential crossover (general?)(2) Cell model with reaction network; evolution to a ‘critical state’; universal statistical
laws on gene expression: log-normal distrib. evolution --- phenotype fluctuation (3) Coupled cell model: robust hierarchical differentiation from stem cell Collaborators: Ishihara(magic7) Furusawa (the
ory),Yomo and Kashiwagi (expriment), See http://chaos.c.u-tokyo.ac.jp for most papers