8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. A...

49
MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org MGS « encore un Modèle Général de Simula3on » and Spa&al Compu&ng 1 JeanLouis Giavi:o, Antoine Spicher, Olivier Michel [email protected] [email protected] [email protected]

Transcript of 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. A...

Page 1: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

MGS  «  encore  un  Modèle  Général  de  Simula3on  »  

and  Spa&al  Compu&ng  

1

Jean-­‐Louis  Giavi:o,  Antoine  Spicher,  Olivier  Michel  

[email protected] [email protected]

[email protected]

Page 2: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

Chronology  and  warnings  

• MGS:  start  September  2000  •  Spa3al  Compu3ng    

– Dagsthul  2006  –  SC  workshop  since  2008    

•  The  tutorial  is  –  “unconven3onal  programming”  oriented  –  examples  and  applica3on  based:  unconven-onal  algorithmic,  biology,  abstract  morphogenesis  

–  no  complexity  and  few  theory  

2

Page 3: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

A  topological  manifesto  Spa3al  compu3ng  proposes  to  celebrate  

corporeality  of  data  rather  than  trying  to  deny  it.  Simon  Greenworld  (MIT  medialab)  

   •  The  logical  approach  in  computer  science  

 computa3on  =  deduc3on    (the  Curry-­‐Howard  isomorphism)  

 •  Other  paradigms  can  be  fruiZull  :  topology  

 computa&on  =  moving  in  some  space    •  Try  to  perceive  space  (and  3me)  in  programs  

(rather  than  logical  opera3ons)  purposes:  technical,  heuris-c,  didac-cal  

3

Page 4: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

Material  

•  The  MGS  home  page  (new  home  page  for  the  end  of  the  year)  h:p://mgs.spa3al-­‐compu3ng.org  

•  MGS  download  (new  home  page  for  the  end  of  the  year)  h:p://www.spa3al-­‐compu3ng.org/mgs  

•  This  MGS  tutorial  h:p://www.spa3al-­‐compu3ng.org/mgs:tutorial  

•  MGS  publica3on  h:p://mgs.spa3al-­‐compu3ng.org/PUBLICATIONS    (old)  h:p://repmus.ircam.fr/giavi:o/publica3ons  h:p://www.lacl.fr/~michel/doku.php?id=research:start  h:p://www.lacl.fr/~michel/www/bib  h:p://www.lacl.fr/~aspicher/publica3ons.html  

 •  The  Spa3al  Compu3ng  home  page  

h:p://www.spa3al-­‐compu3ng.org  

4

Page 5: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

A  complete  schedule  

5

A.  Mo3va3on  &  Applica3on  domains  1.  (DS)2  and  morphogenesis  2.  Interac3on-­‐Based  Compu3ng  3.  Spa3al  Compu3ng  

B.  MGS  …  

C.  Applica3ons  …  

D.  Implementa3on  …  

E.  Current  Fron3ers  …  

Page 6: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

Lecture  I      (Tuesday,  September  4)  

A.  Mo3va3on  &  Applica3on  domains  1.  Modelling  Morphogenesis  2.  (DS)2:  dynamical  systems  with  a  dynamical  structure    

B.  MGS  1.  Collec3on  2.  Transforma3on  3.  A  Func3onal  Language  

 

C.  Applica3ons  1.  Algorithmic  Examples  2.  Monoidal  Collec3ons  3.  GBF  :  Cellular  and  Laice  Gaz  Automata  4.  Delaunay  :  the  growth  of  a  meristem  

6

Page 7: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

Lecture  II      (Wednesday,  September  5)  

A.  Mo3va3on  &  Applica3on  domains  1.  Spa3al  Compu3ng  2.  Interac3on-­‐Based  Compu3ng    

B.  MGS  1.  Abstract  Combinatorial  Complex  2.  Formaliza3on      

C.  Applica3ons  1.  Remeshing  2.  Growing  a  «  T  »  3.  Self-­‐Assembly  of  Musical  Spaces  4.  Growing  an  ontology  with  the  Li:le  Riding  Hood  5.  Growing  an  analogy  with  Paul  Ricœur  

7

Page 8: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

Lecture  III      (Thursday,  September  6)  

B.  MGS  1.  Formalisa3on    2.  Stochas3c  strategies  3.  Discrete  Differen3al  Operators  

C.  Applica3ons  1.  A  generic  diffusion  operator  2.  Four  models  for  the  Paris  iGEM’07  project  

D.  Implementa3on  1.  Incremental  HOAS  implementa3on  of  a  DSL  2.  Beyond  algebraic  data  type  :  a  Generic  Advanced  API  for  Containers  3.  Generic  pa:ern  matching  

E.  Current  Fron3ers  From  Global  to  Local  ,  Mul3-­‐level  and  par3al  processes,  Hybrid  Modeling,  Time  

8

Page 9: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 9

Modeling  and  Simula&on  of  Morphogenesis  

Page 10: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

Environment characterized by its effects on the system

System described by a state (determined by observation)

10

Modelling  morphogenesis:  the  approach  of  A.  Turing  

Page 11: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

statet-­‐1   statet   statet+1   N  

statet-­‐dt   statet   statet+dt   R  

Specifying  a  dynamical  system  (for  simula3on)  

11  

statet  

Specifica&on  of  •   structure  of  state  •   structure  of  &me  •   evolu&on  func&on  

H   H  

H*  

∫H(t)dt  

state

time

evolution

Page 12: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 12

Morphogenesis  as  a  Dynamical  System  

Modelling a dynamical system •  state, including space (e.g. fields) •  time •  evolution function

C : continuous, D: discrete

PDE Coupled ODE

Iteration of functions

Cellular automata …

state C C C D …

time C C D D …

space C D D D …

Page 13: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 13

Modelling  morphogenesis:  the  approach  of  A.  Turing  

•  Uniform  maDer,  con&nuous-­‐oriented  system  descrip&on  è  SPACE  based  One  choice  is  to  ignore  cells  completely,  e.g.,  Physiome  models  3ssues  as  con3nua  with  bulk  mechanical  proper3es  and  detailed  molecular  reac3on  networks,  which  is  computa3onally  efficient  for  describing  dense  3ssues  and  non-­‐cellular  materials  like  bone,  extracellular  matrix  ,  fluids,  and  diffusing  chemicals,  but  not  for  situa-ons  where  cells  reorganize  or  migrate.  

versus    

•  Cell-­‐oriented  discrete  system  descrip&on  è  ENTITY  based  Mul3-­‐cell  simula3ons  are  useful  to  interpolate  between  single-­‐cell  and  con3nuum-­‐3ssue  extremes  because  cells  provide  a  natural  level  of  abstrac3on  for  simula3on  of  3ssues,  organs  and  organisms.    Trea3ng  cells  phenomenologically  reduces  the  millions  of  interac3ons  of  gene  products  to  several  behaviors:  most  cells  can  move,  divide,  die,  differen3ate,  change  shape,  exert  forces,  secrete  and  absorb  chemicals  and  electrical  charges,  and  change  their  distribu3on  of  surface  proper3es.    

(CompuCell3D  manual)    

Aggregate-­‐  vs.  En&ty-­‐based  models  

Page 14: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 14

Modeling  morphogenesis:  paDerning  vs.  development  

4 Spicher, Michel & Giavitto

data) and the content (chemical data) are clearly decoupled, making these modelsusually fall into two categories2:

• formalisms that focus on pattern formation in an initially homogeneous but staticsubstrate, and

• generative formalisms that specify the creation and the evolution of a dynamicshape, irrespective of the processes that may take place within the shape.

Examples of the first category are given by reaction-diffusion systems, activator-inhibitor models, or random Boolean networks. They rely on various model of dy-namical systems (Table 1). Examples of the second category include Lindenmayersystems (L-systems) [36], membrane computing [46], graph grammars [49], self-assembly systems, and others (Table 2).

Table 1 Formalisms that can be used to specify structured dynamical systems according to (a) theunderlying space in which the patterning process takes place, (b) a continuous or discrete repre-sentation of time, and (c) the state variables of the system’s components. “Numerical Solutions”refer to explicit numerical solutions of partial differential equations (PDE) or systems of coupledordinary differential equations (ODE).

C: continuousD: discrete PDE Coupled ODE Numerical

SolutionsCellular

Automata(a) Space C D D D(b) Time C C D D(c) States C C C D

Table 2 Generative formalisms that can be used to specify the evolution of a shape, accordingto the topology connecting the components of the shape. In a multiset, all elements are consideredconnected to each other. In a sequence, elements are ordered linearly: this case includes lists and ex-tends to tree-like structures (lists of lists). Uniform structures involve a regular neighborhood, e.g.,a rectangular lattice where each element has exactly four neighbors. Group-Based Fields (GBF,Section 2.1.4) are a powerful tool relying on mathematical groups to represent such structures.The first four cases describe spatial structures that can be accurately depicted by a graph. Beyondgraphs, nD combinatorial structures are used to define arbitrary connections between componentsof various dimensions. The MGS language (next section) can handle all these types of shapes.

Topology: multiset sequence uniform arbitrary graph nD combinatorialstructures

Formalism: membranesystem

L-systems GBF map L-systems,graph-grammars

MGS

2 This distinction is somewhat contrived. For instance, cellular automata have been devised tostudy self-reproduction of distinct entities, but these entities are represented by specific patterns ina predefined medium.

generative formalism

C : continuous, D: discrete

PDE Coupled ODE Iteration of functions

Cellular automata …

state C C C D …

time C C D D …

space C D D D …

(vs. dynamical systems)

Page 15: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 15

Modeling  morphogenesis:  pa:erning  in  a  predefined  medium  

Alexander  Gavrilovitch  Gurwitsch  (1875-­‐1954)  Morphogene3c  Fields  

Page 16: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 16

©  H.  Meinhardt,  c.1970  Fixed  shape,  diffusion  and  reac3on  

Page 17: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 17

Modeling  morphogenesis:  pa:erning  in  a  predefined  medium  

Compatible with •  the notion of morphogenetic field •  cell fate But •  there is evidence for

feedback loops between the shape and the process inhabiting the shape

from  E.  Haenkel  (cited  by  C.  Goodman-­‐Strauss):  example  of  a  nega3ve  curvature  surface.  Curvature  can  be  controlled  while  the  surface  is  growing  along  a  ‘front’  

Page 18: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 18  

The  medium/process  problem  

Page 19: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

•   (px,  py)  •   (vx,  vy)  •   (px’,  py’)  •   (vx’,  vy’)  

19  

The  medium/process  problem  

a  falling  ball  

at  any  -me  a  state  is  a  posi-on  and  a  speed  

A  dynamical  system  (DS)  

Page 20: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 20  

The  medium/process  problem  

a  falling  ball   a  developing  embryo  

the  structure  of  the  state  (chemical  and  mechanical  state  of  each  cell)  

is  changing  in  -me  A  dynamical  system  with  a  dynamical  structure  

(DS)2  

•   (px’,  py’)  •   (vx’,  vy’)  

at  any  -me  a  state  is  a  posi-on  and  a  speed  

A  dynamical  system  (DS)  

Page 21: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 21  

The  interplay  between  state  and  form  

a  developing  embryo  

the  state  as  well  as  the  structure  of  the  state  is  changing  in  3me  (chemical  and  mechanical  state  of  each  cell  as  well  as  the  arrangement  of  the  cells)  

A  dynamical  system  with  a  dynamical  structure  (DS)2  

Dynamics ON form Dynamics OF form F

orm

State

“Strange  loop”  

Page 22: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

Waddington  :  epigene3c  landscape  

Link  Li  &  Mhairi  Towler,  2012  

Conrad  Waddington  (1905-­‐1975)  

Page 23: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 23

Such  cycles  are  not  new  in  biology  

Tibor Ganti’s CHEMOTON (1971)

metabolic cycle

template molecules

membranogenic molecule

… a “chemical automaton”… (also called fluid automata)

Varela’s AUTOPOIESIS (1974)

“an autopoietic unity is able to self-generate owing to a reaction network taking place within its

own boundary” (the reaction network makes all components of the unit, including the boundary)

M. Eigen & P. Schuster: hypercycle, W. Fontana & L. Buss: organization …

Page 24: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

•  Complex  systems  •  Whose  structure  evolves  over  the  3me  

–  The  structure  constrains  the  behavior  of  the  system  that  modifies  the  structure  …  –  The  phase  space  cannot  be  defined  a  priori  –  The  evolu3on  func3on  cannot  be  specified  globally  

•  Examples  –  Biology  

•  Molecular  biology  •  Developmental  biology  

–  Physics  •  Soz  ma:er  mechanics,  

mul3-­‐scale  systems  •  General  rela3vity  

–  Urbanism  City  growth,  

traffic  control,  …  –  Computer  science  

Internet,  sensor  network,  reconfigurable  nanobots,  knowledge  network,  …  

DS²  :  Dynamical  System  with  a  Dynamical  Structure  

© David S. Goodsell 1999

© Tecplot

© L. Sanders - EUROSIM

© C. Harrison - Clusterball project 24  

Page 25: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

What  has  changed  since  Turing’s  &me  

25

Page 26: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 26

What  has  changed  since  Turing’s  &me  

P.  Prusinkiewicz,  c.2003  Diffusion  and  reac3on  in  a  deformable  surface  (E.  Coen’s  expanding  canvas  metaphor).  Spring-­‐mass  system.  No  topological  change.  

Page 27: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

Sozware  as  Science  ?    

•  Intelligibility    The  en-re  process  should  be  accessible  for  analysis  into  a  finite,  not  very  large  number  of  stages,  each  stage  being  represented  as  a  monotonic  func-on  of  some  definite  ini-al  condi-ons  and  a  single  variable  such  as  -me,  or  distance,  etc.  (Gurwitsch,  1944)  à  compress  behavior  or  shape  in  few  rules  

•  Simula3on  is  only  a  first  step:  models  must  enable  reasoning  à  stay  close  to  mathema3cal  formalism  

27

M.-P. C

ani & F. B

ertails

A program is a formal object (and some form of reasoning on it is possible) but a 106 lines of codes is not an explanation !

Page 28: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

(…)  An  invagina3on  of  a  germ  layer  may  be  explained  on  a  basis  of  a  pressure  difference  between  the  two  surfaces  (or  sides)  or  by  cell  movements,  and  so  forth.  This  can  be  considered  as  an  'explana3on'  un3l  we  ask  about  the  origin  of  pressure  differences,  or  the  mechanisms  involved  in  cell  movement,  etc.  However,  ques3ons  of  this  kind  become  trivial  when  a  larger  process,  rather  than  its  individual  components  becomes  the  main  problem.  Suppose  for  a  moment  that  each  element  in  the  succession  A,  a,  C…  can  be  explained  separately,  e.g.  A  as  a  swelling,  a  as  a  chemical  reac3on,  etc.  Interes3ng  as  they  may  be,  these  explana3ons  are  of  subordinate  importance  when  related  to  the  main  ques3on:  Why  indeed  is  a  regular  (emphasis  added)  succession  of  these  obviously  quite  different  processes  taking  place  at  all?  Most  biological  problems  are  of  this  kind  and  all  of  embryogenesis  is  just  such  a  single  problem.  Here  we  require  a  peculiar  or,  maybe,  original  explanatory  principle...  A  process  may  become  accessible  to  explana3on  only  insofar  as  one  can  succeed  in  subs3tu3ng  [understanding  of]  a  purely  phenomenological  mul3plicity  and  diversity  of  events  [for  understanding]  of  a  less  diverse  and  less  arbitrarily  created  picture  correctly  reflec3ng  reality.  The  main  aim  of  such  a  construc3on  would  be  as  follows.  The  en3re  process  should  be  accessible  for  analysis  into  a  finite,  not  very  large  number  of  stages,  each  stage  being  represented  as  a  monotonic  func3on  of  some  definite  ini3al  condi3ons  and  a  single  variable  such  as  3me,  or  distance,  etc.  If  this  cannot  be  realized,  we  consider  a  given  set  of  events  as  scien3fically  inaccessible.  On  the  other  hand,  even  a  par3al  success  of  such  an  enterprise  is  an  obvious  step  forward."  

(Gurwitsch,  1944)  cited  by  Beloussov  in  “Life  of  Alexander  G.  Gurwitsch  and  his  relevant  contribu3on  to  the  theory  of  morphogene3c  fields”,  Int.  J.  Dev.  Biol.  41,  771-­‐779  (1997)  

28

Page 29: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

COMBINING  PATTERNING  AND  DEVELOPMENT  

29

Page 30: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

Pa:erning  vs.  Growth  

•  Dynamics  ON  form    pa:erning,  mo3f  forma3on  in  a  predefined  form  – Diffusion,  reac3on-­‐diffusion,  transport  (con3nuous  models)  –  cellular  automata  (discrete  models)  

 

•  Dynamics  OF  form    growth,  deforma3on  of  a  shape    – Deforma3on  of  elas3c  bodies  (con3nuous  models)  –  Lindenmayer  systems  (discrete  models)    

30

Page 31: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

The  interplay  between  state  and  form  a  developing  embryo  

the  state  as  well  as  the  structure  of  the  state  is  changing  in  3me  

 (chemical  and  mechanical  state  of  each  cell  as  

well  as  the  arrangement  of  the  cells)  

A  dynamical  system  with  a  dynamical  structure  (DS)2  

Dynamics ON form Dynamics OF form F

orm

State

31

Page 32: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

(DS)2  versus  DS  

•  Processes  in  the  form  are  topologically  (geometrically)  meaningful  e.g.  growth  rate  

 •  Topological  (geometric)  informa3on  is  meaningful  e.g.  domain  of  diffusion,  informa3on  transfert  

32

Page 33: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

Topological  informa3on  is  meaningful  

A  basic  example  [T.  Gross,  Adapta3ve  Networks]:  –  Infec3ous  disease  spreading  across  a  network  – Node  are  either  suscep-ble  or  infected      

–  Suscep3ble  nodes  linked  to  an  infected  node  become  infected  with  probability  α

–  Infected  nodes  recover  with  probability  β

33

Page 34: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

Topological  informa3on  is  meaningful  

The  disease  persists  in  the  network  depending:  –  of  the  probability  α  

–  of  the  probability  β  

–  of  the  mean  excess  degree  d    the  average  number  of  links  that  one  finds  connected  to  a  node  that  is  reached  by  following  a  random  link:  a  global  topological  informa3on  

•  d  >  threshold:  every  node  will  be  infected  once  in  a  while  (in  a  large  network)  

•  By  watching  the  behavior  of  a  single  node  (local  observa-on)  for  a  sufficiently  long  period  of  3me,  we  can  es3mate  if  the  mean  excess  degree  (global  topological  informa-on)  exceeds  the  threshold  

•  Simple  mechanism  can  put  the  system  at  the  cri-cal  point!  (no  special  reason  to  be  in  a  generic  state)  

34

Page 35: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

The  interplay  between    form  and  topology  is  meaningful  

•  A  suscep3ble  node  is  allowed  to  cut  a  link  with  an  infected  node  with  a  certain  probability  decreasing  when  the  delay  between  its  own  infec3ons  increase  

•  Now  there  is  a  feedback  loop    between  topologies  and  states  of  the  network  the  dynamics  of  the  prevalence  of  the  disease  depends  on  the  network  topology,  but  the  evolu3on  of  the  network  topology  also  depends  on  the  prevalence  of  the  disease  

Simple  mechanism  can  put  the  system  at  the  cri@cal  point!  no  special  reason  to  be  in  a  generic  state  

e.g.  synchroniza3on  of  growth  rate  in  flat  leaf  

(and  it  can  be  worst  if  there  is  no  3mescale  separa3on)  

35

Page 36: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

OK:  the  coupling  is  important.  What  can  we  do?  

1.   “Dynamics  ON  form”  toward  DS2  –  parameterize  (control)  the  shape  by  some  quan33es  

(e.g.  curvature  for  a  manifold  or  adjacency  matrix  for  a  graph)  

–  link  these  quan33es  with  processes  in  the  shape  

•  growth  depending  on  concentra3on  

 

36

Page 37: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

Differen&al  geometry  is  not  enough  

•  Encoding  a  shape  into  con3nuous  parameters  is  difficult  

•  It  does  not  handle  topological  changes  very  well  

apical  elonga3on  depends  of  3me  

•  Ok,  some3mes  they  are  some  tricks  (e.g.  level  set)  but  they  are  very  smart  tricks  

37

Page 38: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

OK:  the  coupling  is  important.  What  can  we  do?  

1.   “Dynamics  ON  form”  toward  DS2  –  parameterize  (control)  the  shape  by  some  quan33es  

(e.g.  curvature  for  a  manifold  or  adjacency  matrix  for  a  graph)  

–  link  these  quan33es  with  processes  in  the  shape  

•  growth  depending  on  concentra3on  

 

2.   “Dynamics  OF  form”    toward  DS2    –  enhance  the  form  by  parameters    

–  put  a  dynamic  on  these  parameters  example:  module  in  Lindenmayer  systems  

38

Page 39: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

Lindenmayer  systems  

39  

The  growth  of  Anabaena  Catenula  •  two  cell  types  :  vegeta3ve  and  heterocyst  •  asymetric  division  :  one  is  smaller  than  the  other  •  polarized  cell  (lez/right  orienta3on)  

A a B b

a -> aB A -> bA b -> aB B -> bA a a B a B b A a B b A a B b A a B b A a B b A a B b A a B b A ...

Page 40: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 40

A  good  example  of  PaDerning+Development  in  a  declara&ve  formalism:  Lindenmayer  systems  with  modules  

•  The  structure  of  a  tree  can  be  coded  by  a  string  of  parenthe3sed  symbols  

•  A  symbol  is  an  elementary  part  of  the  plant  

•  The  symbol  between  [  and  ]  represents  a  sub-­‐tree  

•  Addi3onal  conven3ons  are  used  to  represent  main  axis,  orienta3on,  depth,  etc.    

•  A  rule    s0  →  s1  s2  s3  …  

represents  the  evolu3on  of  s0   P. P

rusi

nkie

wic

z

Page 41: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 41

Diffusion  and  reac3on  in  a  linear  growing  medium    M.  Hammel  and  P.  Prusinkiewicz  (1996)  

The  following  rules  state  that  a  differen3ated  cell  (heterocyst)  returns  to  a  vegeta3ve  state  if  the  concentra3on  of  the  ac3vator  is  too  low.  In  addi3on,  if  the  cell  is  large  enough,  it  con3nues  to  grow.  

e / (D(e) & (e.a < thr) | (e.x >= shorter*gr)) => {type ="C", a=e.a/gr, h=e.h/gr, x=e.x*gr, p=e.p};

The  following  rule  specifies  when  a  cell  with  a  lez  polarity  divides.  Only  vegeta3ve  cells  can  divide  (hence  the  predicate  C  in  the  rule  guard)  and  it  must  be  large  enough.  The  volume  of  the  two  daughter  cells  remains  the  same,  so  there  is  no  varia3on  in  the  concentra3on.

e / (C(e) & (e.x >= lm) & (e.p == L))

=> {type="C", a=e.a, h=e.h, x=e.x*shorter, p=L},

{type="C", a=e.a, h=e.h, x=e.x*longer, p=R};

Linear cell structure

morphogene concentration

time

Page 42: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 42

•  Rewri3ng  system  –  Used  to  formalize  equa3onnal  reasoning  –  A  genera3ve  device  (grammar)  –  Replace  a  sub-­‐part  of  an  en3ty  by  an  other  –  Set  of  rewri3ng  rules  α  à  β  

•  α:  pa:ern  specifying  a  sub-­‐part  •  β:  expression  evalua3ng  a  new  sub-­‐part  

•  Example:  arithme3c  expressions  simplifica3on  

Rewri&ng  systems  (and  abstract  transi3on  systems)  

x

+

0

x

y

+

x x

+

y

Page 43: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

A  non-­‐standard  interpreta3on  

43

e1  +  e2  →  …    e1  can  be  a  cell  and  e2  a  signal  e1  and  e2  can  be  interac3ng  cell  

+  is  the  possibility  of    interac-on  between  en33es  (or  some  other  rela3onships)  

→  is  the  passing  of  3me,  a  local  evolu3on,  a  transi3on,  the  concre3za3on  of  the  interac3on  

 Examples:  if  e  is  a  cell  and  i  a  biochemical  signal  

 e  +  i  →  e’  growth  (evolu3on  of  e  on  signal  i)      e  +  i  →  e+i’  quorum  sensing  e  +  i  →  e’  +  e’’  division  e  +  i  →  .    apoptose  

Page 44: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 44

Rewri3ng  systems  in  simula3on  

Complex  systems       Rewri&ng  techniques  

Modelling   Specifica3on  State  (space)   Data  structure  

hierarchical  and  tree  organiza3ons  arbitrary  complex  organiza-ons  

formal  trees  (or  terms)    ?  

Evolu&on  func&on   Set  of  rules  interac@ons  à  evolu3on  

local  evolu&on  laws  α  :  pa:ern  à  β  :  expression  rewri3ng  rules  

Simula3on   Applica3on  Trajectories  

Time  management  Deriva3ons  Rule  applica3on  strategy  

discrete,  event-­‐based,  synchronous  vs.  asynchronous  

maximal  parallel,  sequen3al,  determinis3c,  stochas3c  

ßà

Page 45: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 45

•  local  evolu-on  rules  mandatory  when  you  cannot  express  a  global  func3on/rela3on  because  the  domain  of  the  func3on/rela3on  is  changing  in  3me  

•  interac3on  based  approch  the  l.h.s.  of  a  rule  specifies  a  set  of  elements  in  interac-on,  the  r.h.s.  the  result  of  the  interac3on  

•  the  phase  space  is  well  defined  but  not  well  known  a  genera3ve  process  enumerates  the  elements  but  membership-­‐test  can  be  very  hard  

•  various  kind  of  -me  evolu-on  (for  the  same  set  of  rules)  

•  demonstra-on  by  induc-on  on  the  rules  or  on  the  deriva3on  (e.g.  growth  func3on  in  L  system)  

Proper3es  

Page 46: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 46

How  to  extend  to  arbitrary  spa3al  structure?  

•  Anabaena  was  «  easy  »  because  of  the  linear  uniform  structure  •  How  to  handle  the  complex  spa3al  structure  of  a  cell?  

David S. Goodsell

Page 47: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org 47

Three  challenges  and  some  tools  

compilation, engineered emergence

validation, analysis, collective

properties

limits, non-standard analysis

approximation, numerical resolution,

partitionning

DNA

cellular machinery

Cellular machinery

DNA

global local

discrete

software/data hardware/programme

population

individual Molecules, Compartments,

Cells

concentration, time, fields

continuous

Page 48: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

Acknowledgements  MGS:  Antoine  Spicher,  Olivier  Michel,  Julien  Cohen  S&S  Bio  :  Hanna  Klaudel,  Franck  Delaplace,  Hugues  Berry,  Przemek    Prusinkiewicz,  Annick  Lesne…  Spa3al   Compu3ng:   Jacob   Beal,   Fréderic   Gruau,  René  Doursat…  

Examples:   Pierre   Barbier   de   Reuille,   Christophe  Godin,  Samuel  Bo:ani,  the  Paris  iGEM’07  team…    

Some  figures  are  borowed  from  Olivier  Michel,  Antoine  Spicher,  Pierre   Barbier   de   Reuille,   Franck   Delaplace,   Hugues   Berry  (INRIA),  the  iGEM  Paris  2007  and  many  others.  

 

48  

Page 49: 8 intro ds2 - spatial-computing.org · MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher.  A topological(manifesto(Spaal! compung !proposes!to!celebrate!

MGS @ UCNC’2012. J.-L. Giavitto, A. Spicher. http://mgs.spatial-computing.org

•   Antoine  Spicher  •   Olivier  Michel  •   PhD  and  other  students              J.  Cohen,  P.  Barbier  de  Reuille,              E.  Delsinne,  V.  Larue,  F.  Le3erce,  B.  Calvez,                F.  Thonerieux,  D.  Boussié    and  the  others...  •   Collabora&ons  •  A.  Lesne  (IHES,  stochas3c  simula3on)  •  P.  Prusinkiewicz  (Calgary,  declara3ve  modeling)        •  P.  Barbier  de  Reuille  (meristeme  model)  •  C.  Godin  (CIRAD,  biological  modeling)                            •  H.  Berry  (LRI,  stochas3c  simula3on)  •  G.  Malcolm  (Liverpool,  rewri3ng)  •  J.-­‐P.  Banâtre  (IRISA,  programming)                  •  F.  Delaplace  (IBISC,  synthe3c  biology)  •  P.  Di:rich  (Jena,  chemical  organiza3on)  •  E.  Goubault  (CEA,  topological  formaliza3on)  •  F.  Gruau  (U.  PXI,  language  and  hardware)                    •  P.  Liehnard  (Poi3er,  CAD,  Gmap  and  quasi-­‐manifold)    

49

Thanks  http://mgs.spatial-computing.org