Jochen Triesch, UC San Diego, 1 Pattern Formation Goal: See how globally ordered spatial structures can arise from local interactions Outline: chemical reaction systems: stirred and unstirred reactors diffusion equation: motivation and simple cases reaction diffusion systems: 1D activator/inhibitor system, Turing instability 2D cellular automata for modeling development of animal coat patterns L-Systems based on grammars
Jochen Triesch, UC San Diego, 2 Chemical Reactors Setup 1: well stirred reactor N different chemicals (reactants) with time-varying concentrations c i (t), i=1,..N vector notation: c(t) changes in concentrations due to reactions modeled by system of DEs: chemical reactor, well stirred
Jochen Triesch, UC San Diego, 3 Example System: three chemicals A, B, C; concentrations n A, n B, n C Reaction: A+B ↔ C assume rate of “→”reaction proportional to probability of encounter of an A and B molecule: k 1 n A n B assume rate of “←“ reaction is: k 2 n C A A B B C C a) b)
Jochen Triesch, UC San Diego, 4 Setup 2: unstirred reactor N different chemicals (reactants) with time-varying and space-varying concentrations c i (x,t), i=1,..N vector notation: c(x,t) changes in concentrations of one chemical due to reactions depends on concentrations of other chemicals at the same location changes in concentration due to diffusion depend on concentration at neighboring sites chemical reactor, unstirred
Jochen Triesch, UC San Diego, 5 Derivation, part 1: random walk time between collisions: T characteristic velocity v distance traveled l=T v current J : particles/surface area/time Diffusion Equation (1D)
Jochen Triesch, UC San Diego, 6 Derivation, part 2: conservation of particles, consider small volume Adx: From above: Together: “diffusion equation” “diffusion constant”
Jochen Triesch, UC San Diego, 7 Simple Analysis of Diffusion Equation Consider point source of strength S at x 0, decay with rate μ: decay point sourcediffusion
Jochen Triesch, UC San Diego, 8 Case 1: source inactive (S=0), look for homogenous solution In this case no diffusion: with the solution: The homogeneous solution exponentially decays to zero.
Jochen Triesch, UC San Diego, 9 Case 2: source active (S>0), look for stationary solution In this case: bi-exponential Ansatz: leads to solution: In particular: shape gets wider if μ smaller (slower decay) and/or D gets bigger (faster diffusion)
Jochen Triesch, UC San Diego, 10
Jochen Triesch, UC San Diego, 11 Reaction Diffusion System Consider 1D system with just two substances: Gierer&Meinhardt (1972), Haken (1990) inhibitor grows due to activator spontaneous creation diffusion “activator” “inhibitor” positive feedback limited by inhibitor spontaneous decay spontaneous decay
Jochen Triesch, UC San Diego, ) Simplify System: new set of variables to remove some constants 2) Find homogenous stationary solution 3) Linearize the system: Consider small deviations around homogeneous stationary solution in the linearized system 4) Check under what conditions instability occurs 5) Translate results from simplified system back into original system 6) Main Finding: inhibitor must diffuse faster than activator Analysis of System
Jochen Triesch, UC San Diego, 13 H. Meinhardt (1998), The algorithmic beauty of sea shells. Simulation of (similar) system: Lyria planicostata taiwanica
Jochen Triesch, UC San Diego, 14 More Sea Shell Patterns H. Meinhardt (1998), The algorithmic beauty of sea shells. Sea shells have 1D growing edge, pigmentation of shell can be seen as historical record of 1D reaction diffusion system. Amoria ellioti
Jochen Triesch, UC San Diego, 15 H. Meinhardt (1998), The algorithmic beauty of sea shells. Oliva porphyria
Jochen Triesch, UC San Diego, 16 H. Meinhardt (1998), The algorithmic beauty of sea shells. Conus marmoreus
Jochen Triesch, UC San Diego, 17 example of self-organization: positive and negative feedback global pattern from local interactions spontaneous symmetry breaking basis of many models of biological form sea shells animal coats (cheetah, zebra, giraffe, salamander, tropical fish, …) basis for thinking about some more advanced problems embryogenesis: not just diffusion but active transport, inter and intra cell communication growing of arms, legs, wings different organs, nervous system Discussion of Reaction Diffusion Systems
Jochen Triesch, UC San Diego, D Cellular Automaton Model
Jochen Triesch, UC San Diego, 19 Setup: Regular grid of “cells” each in binary state s i in{0,1} asynchronous updating: pick single unit at random, update, pick next, … update rule: Parameters: strength of activation/inactivation: J 1, J 2 range of activation/inactivation: R 1, R 2 “bias”: h
Jochen Triesch, UC San Diego, 20 % Parameters length = 200; % size of layer fraction = 0.2; % fraction of units updated each turn j1 = 1.0; % strength of activation j2 = -0.3; % strength of inactivation r1 = 3; % range of activation r2 = 6; % range of inactivation h = 0; % strength of bias % Initializations X = unidrnd(2,length,length)*2-3; f_size = 2*ceil(r2)+1; f = j2*ones(f_size); for(x=1:f_size) for(y=1:f_size) rr = (x-(f_size+1)/2.0)^2+(y-(f_size+1)/2.0)^2; if rr <= r1^2 f(x,y)=j1; elseif rr > r2^2 f(x,y)=0; end figure(1) imagesc(f); % display filter colormap(gray); axis equal; % Main loop while(1) figure(2) imagesc(X); % display state colormap(gray); axis equal; axis off; pause % synchronous update leading to oscillations, so update % only a fraction of the units New = ((h+conv2(X,f,'same'))>0)*2-1; Update = unifrnd(0,1,length,length)<fraction; X(Update)=New(Update); end
Jochen Triesch, UC San Diego, 21 j1=1, j2=-0.3, r1=3, r2=6, h=6j1=1, j2=-0.3, r1=3, r2=6, h=-6j1=1, j2=-0.3, r1=3, r2=6, h=0 Simulations Varying bias field h:
Jochen Triesch, UC San Diego, 22 j1=1, j2=-0.3, r1=3, r2=18, h=-6 Simulations Varying inactivation range r2: j1=1, j2=-0.3, r1=3, r2=12, h=-6j1=1, j2=-0.3, r1=3, r2=6, h=-6
Jochen Triesch, UC San Diego, 23 j1=1, j2=-0.01, r1=3.5, r2=6, h=0 j1=1, j2=-0.1, r1=3.5, r2=6, h=0 Using smaller J2:
Jochen Triesch, UC San Diego, 24 nice example of formation of patterns from local interactions fairly abstract compared to reaction diffusion DE same scheme of local activation/global inhibition range of qualitatively different patterns from same dynamics simple models for formation of animal coat patterns role of boundary conditions: periodic, non-square lattice number of extensions Discussion of 2D Cellular Automaton
Jochen Triesch, UC San Diego, 25 Idea (Aristid Lindenmayer, 1968): model development of tree-like structures w/ grammar Example Grammar: A→A B→AC C→ [B][B] (brackets mean branching) Example updates: 1.[B] 2.[AC] 3.[A[B][B]] 4.[A[AC][AC]] 5.[A[A[B][B]] [A[B][B]]] L-Systems
Jochen Triesch, UC San Diego, 26 Example: ( This is a simple two-dimensional L-systems graphical representation of an oppositely branched red alga: Axiom = A A → F[-A][+A]FA F → FF Parameters: Initial angle = 90 degrees Angle increment = 30 degrees Iteration depth = 8
Jochen Triesch, UC San Diego, 27 Other Examples: Developmental sequence of Mycelis muralis. Copyright © 1987 P. Prusinkiewicz and J. Hanan. The Garden of L. Copyright © P. Prusinkiewicz, F.D. Fracchia, J. Hanan, and D. Fowler
Jochen Triesch, UC San Diego, 28 Modeling Neurons: Ascoli and Krichmar (2000): top: real hippocampal pyramidal neurons, bottom: artificial “L-neurons”