1. Multi-Agents Systems and environment modelling…
NetBioDyn,
an easy to use multi-agents engine
for ecosystems simulation
Vincent Rodin

2. Road map
Multi-Agents Systems (MAS)
From Biological environment simulation
Towards Ecosystems simulation
NetBioDyn software
Conclusions and futur works

3. Models’ autonomy Multi-Agents systems properties
Agent : perception-decision-action
Multi-agents System :
auto-organisation
emergence
robustness
adaptability
Models’ autonomy

4. Road map
Multi-Agents Systems (MAS)
From Biological environment simulation
Towards Ecosystems simulation
NetBioDyn software
Conclusions and futur works

5. Multi-Agent Systems and Biological modelling & simulation
From cell-agent
To systemic approach
Interaction-agent
Interface-Agent
Reaction-agent
Cell-agent

6. Model of located agents with
Cell-agent model
Mitosis
Activation
Internalisation
Expression of receptor
Apoptose
Basic behaviours
Behaviours
Model of located agents with
complex behaviors

7. An exemple of application Simulation of physiologic coagulation:
Cell-agent model:
An exemple of application
Simulation of physiologic coagulation:
Cell
Fibroblasts cells, endothelial cells, platelet
Procoagulant factors
TF, I, II, V, VII, VIII, IX, X, XI, …
Inhibitor factors
TFPI, AT3, 2M, PC, PS, PZ, …
Flux Ia Va
IIa Pa Ia Pa Ia
Coagulation
Cascade
 Blood clot Xa Xa
VIIa
IIa Xa
VIIIa
VIIIa Va
VIIIa
IXa Xa
VIIa
IXa Xa
IXa Xa
VIIa
VIIa
VIIa
Endothelial cell
F.W
Fibroblasts + Tissue Factors

8. Cell-agent model: An exemple of application
Simulation of physiologic coagulation:  blood clot…

9. An exemple of application
Cell-agent model:
An exemple of application
Elements of validation of the coagulation
multiagents model :
Comparison with Biological experiment
Curve of thrombin
Generation
[Hemker, 1995]
Coherence with respect to pathologies

10. An exemple of application Simulation of physiologic coagulation:
Cell-agent model:
An exemple of application
Simulation of physiologic coagulation:
Healthy patient, hemophiliac,
hemophiliac with treatment
Thrombin
generation
Normal
coagulation
Hemophilia B
+ NovoSeven
Time
Hemophilia A
in seconds

11. Multi-Agent Systems and Biological modelling & simulation
From cell-agent
To systemic approach
Interaction-agent
Interface-Agent
Reaction-agent
Cell-agent located agent

12. Reaction-agent model reaction « microscopic » level :
« macroscopic » level :
agent = cell/molecule
agent = reaction
1: reading of the concentrations
in reactants
perception
action
decision
2:calculation the reaction speed and then the quantity of reactant to be reacted
reaction
3: consequently, modification of the
concentrations in reactants and products

13. Reaction-agent model Non located agents Spatial … r1 r2
indiscernibility r1 … r2
Non located agents rm
C1, C2, …, Cn
Chemical reactor
Asynchronous phenomena and chaotic order
No Ordinary Differential Equation

14. Reaction-agent model Asynchrony Random permutation of MAS
time
tn+3
tn+2
tn+1 tn
tn(2)
tn(3)
tn(4)
tn(1)
Asynchrony
of MAS tn
n+1
n+2
n+3
cycles n
cycle of simulation r2 r3 r4 r5 r1 r1 r5 r3 r2 r4 r5 r2 r1 r4 r3
Pascal.
Maître de conférences de mon équipe en maths appli :
aspect preuve de la méthode:
Méthode agent : convergence ordre 2
Méthode classique : convergence ordre >= 3
Random permutation
Traditional approaches
Asynchronous and chaotic iterations

15. Reaction-agent model E1 + S1  E1 + P1
d[S1]/dt = – kcat1[E1][S1]/(Km1+[S1])
– kcat3[E1][S1]/(Km3+[S1])
d[S2]/dt = + kcat2[E2][S2]/(Km2+[S2])
d[P1]/dt = – kcat1[E1][S1]/(Km1+[S1])
– kon3[P1][P2]
+ kcat3[E1][S1]/(Km3+[S1])
d[P2]/dt = + kcat2[E2][S2]/(Km2+[S2])
d[P1P2Complex]/dt = kon3[P1][P2]
E1 + S1  E1 + P1
E2 + S2  E2 + P2
P1 + P2  P1P2Complex
E3 + S1  E3 + P1
new EnzimaticReaction(plasma, E1, S1, P1, kcat1, Km1);
new EnzimaticReaction(plasma, E2, S2, P2, kcat2, Km2);
new ComplexFormationReaction(plasma, P1, P2, P1 P2Complex, kon3);
new EnzimaticReaction(plasma, E3, S1, P1, kcat3, Km3);

16. Reaction-agent model: An exemple of application
VIIIa + IXa -> VIIIa-IXa
VIIIa-IXa + X -> VIIIa-IXa + Xa
Va-Xa + II -> Va-Xa + IIa
AT3 + (IXa, Xa, XIa, IIa) -> 0
alpha2M + IIa -> 0
I + IIa -> Ia + IIa
TM (endo cell) + IIa -> IIi
AT3(endo cell) + (IXa, Xa, XIa, IIa) -> 0
ProC + IIi -> PCa + IIi
PCa + Va -> 0
PCa + Va-Xa -> Xa
PCa + PS -> PCa-S
PCa + PCI -> 0
PCaS + Va -> 0
PCaS + V -> PCa-S-V
PCa-S-V + VIIIa-IXa -> IXa
PCa-S-V + VIIIa -> 0
Fibroblaste + VIIa -> VII-TF
VII + VIIa -> VIIa + VIIa
VII + VII-TF -> VIIa + VII-TF
IX + VII-TF -> IXa + VII-TF
X + VII-TF -> Xa + VII-TF
TFPI + Xa -> TFPI-Xa
TFPI-Xa + VII-TF -> 0
VII + IXa -> VIIa + IXa
IXa + X -> IXa + Xa
II + Xa -> IIa + Xa
XIa + IX -> XIa + IXa
XIa + XI -> XIa + XIa
IIa + XIa -> IIa + XIa
IIa + VIII -> IIa + VIIIa
IIa + V -> IIa + Va
Xa + V -> Xa + Va
Xa + VIII -> Xa + VIIIa
Xa + Va -> Va-Xa
Simulation of physiologic coagulation:
Healthy patient, hemophiliac,
hemophiliac with treatment
½ seconds
Thrombine
generation
Normal
coagulation
Hemophilia B
+ NovoSeven
Time in
Hemophilia A
42 reactions

17. Multi-Agent Systems and Biological modelling & simulation
From cell-agent
To systemic approach
Interaction-agent
Interface-Agent
Reaction-agent non located agent
Cell-agent located agent

18. Interface-agent model
Transport
phenomena
Interface
advection
diffusion
Chemical
reactor
Chemical
reactor
Medium
Interface
Intérêt : on ne touche pas aux agents réactions  multi-modèles
Démon de Maxwell
Medium
Relaxation
phenomena
Chemical
reactor
Medium

19. « Classical » point of view:
Interface-agent model
« Classical » point of view:
transport
interface
/r
Chemical
reactor
Chemical
reactor
medium
/t
Variables = concentrations in reactants in each chemical reactor (mesh of the medium)
All the phenomena are supposed simultaneous
Resolution of partial differential equations

20. An « agent » point of view:
Interface-agent model
An « agent » point of view:
Interface-agent
interaction between the mediums
Asynchronous phenomena and chaotic order
No Partial Differential Equation

21. Interface-agent model: An exemple of application
Coagulation :
HOLE
Flux
3D Vessel
Chimical Reactors : 42 reactions (coagulation)

22. Multi-Agent Systems and Biological modelling & simulation
From cell-agent
To systemic approach
Interaction-agent
Interface-agent transport agent
Reaction-agent non located agent
Cell-agent located agent

23. Generic model of interaction-agent
Main principle :
Models’ autonomy
Interaction between models
of different natures
Multi-modeling Simulation
Systemic paradigm
Data Matter exchange
between organisations

24. Generic model of interaction-agent
Simulator *
Activity
Phenomenon
(organisation)
Component
Interaction
agent *
Cell
agent
Organ *
Interface
agent
Reaction

25. Generic model of interaction-agent : An exemple of application
Medium
Cell
Nucleus
Cytoplasm
Membrane

26. Generic model of interaction-agent : An exemple of application
Keratinocyte
Macrophage
Mast cell
Capillary vessel

27. Multi-Agent Systems and Biological modelling & simulation
From cell-agent
To systemic approach
Interaction-agent multimodeling
Interface-agent transport agent
Reaction-agent non located agent
Cell-agent located agent

28. Interests in the field of biology
To help the reflection
Test and validate hypotheses
Analyse parameters influence
Understand complex phenomena
To prepare experiments
To accelerate the drug discovery
Biology
Compute Sciences MA S

29. Road map
Multi-Agents Systems (MAS)
From Biological environment simulation
Towards Ecosystems simulation
NetBioDyn software
Conclusions and futur works

30. Towards Ecosystems simulation
Cells  Entities (insects, etc.)
&
Low level interaction
Reaction  High level interaction
between entities
Interface/Interaction  Physical environment

31. Road map
Multi-Agents Systems (MAS)
From Biological environment simulation
Towards Ecosystems simulation
NetBioDyn software
Conclusions and futur works

32. an easy to use multi-agents engine biological & ecosystem simulations
NetBioDyn software :
an easy to use multi-agents engine
Goal:
Rapid prototyping of
biological & ecosystem simulations

33. an easy to use multi-agents engine
NetBioDyn software :
an easy to use multi-agents engine
Key concepts:
Environment: a grid
Entities: colored squares
Interaction with environment : simple rules
Interaction between entities : simple rules
Easy to use… No programming skill requires !!!!

34. NetBioDyn software :
How to model…?
1st : take a grid
2nd : decide which entities to use
…..
3rd : define rules (with a probility of activation)
to give entities movements and interactions + Ø  Ø +
P=1 (go to the right) +  +
P=0.5 (contamination)
…..

35. NetBioDyn software :
How to model…?
4th : place entities
and : run +  +
(go to the right)
(contamination) +  +
…..

36. NetBioDyn software :
How to model…?
4th : place entities
and : run +  +
(go to the right)
(contamination) +  +
…..

37. NetBioDyn : 1st example, randomwalk
Entities: or
Rd_Walk 
P=1 or

38. NetBioDyn : 1st example, randomwalk
Entities: or
Rd_Walk 
P=1 or

39. NetBioDyn : 2nd example, sandglass
Entities: grain and border 
Go_Down
Go_LeftRight or
P=1
P=0.5
Go_Border
P=0.75

40. NetBioDyn : 3rd example, prey-predator
Entities: prey and predator
Prey ½ life time : 2000 cycles + Rd_Walk
Predator ½ life time : 200 cycles + Rd_Walk

41. NetBioDyn : 3rd example, prey-predator
Entities: prey and predator or
Prey_Birth 
P=0.01 or

42. NetBioDyn : 3rd example, prey-predator
Entities: prey and predator
Predator_Prey 
P=0.6    

43. NetBioDyn : 4th example, fish farm
Entities: fish , shrimp , algae ,
fisherman , border

44. NetBioDyn : 4th example, fish farm
Entities: fish , shrimp , algae ,
fisherman , border

45. Road map
Multi-Agents Systems (MAS)
From Biological environment simulation
Towards Ecosystems simulation
NetBioDyn software
Conclusions and futur works

46. NetBioDyn : Conclusion
Advantage:  very simple (entities, rules)
 no programming !
Drawback:  very simple (entities, rules)
 no entity’s state !
Example: Ø  Ø (movement)
……………
Example: *  * (new entity’s state)
……………
 new entity !

47. NetBioDyn : futur works
Self-ajusting parameters…
 a great challenge !
Entity’s state (simply an integer)…
 Building a model would be simpler by
minimizing the nb of « different » entities

48. Road map
Multi-Agents Systems (MAS)
From Biological environment simulation
Towards Ecosystems simulation
NetBioDyn software
Conclusions and futur works