Stem cell research has been the biggest growth area of medical science in recent years. However, experiments are very limited as it is not possible to track stem cells in the adult human body. In this Complex world, we use models in designing and developing high functionality systems, which in turn contains numerous objects. Many objects with simple ability integrate to perform complex tasks. We use models to develop and understand such behavior to understand the mechanisms involved.
“If you grow the phenomena, you’ll understand how it works” – Joshua Epstein
As said above, I am specifically interested in using the agent based models in establishing the key mechanisms of stem cells in an artificial environment to understand its process of emergence as organs. Stem cells are identical cells with ability to develop into any kind of body cell and divide to develop tissues for skin, heart, nerves, blood, etc). Simulation models for such biological systems run on computer systems can be used to perform classical experiments in a long time period.
2.1 Motivation
I view Agent Based Modeling as a type of individual based modeling that encapsulates a given individual (agent) as a software object with methods and properties. It is perfect for studying complex behavior and allows in creating artificial population of agents and running many experiments in identical settings in a long time period.
2.2 Aims & Objectives
The goal of this project is to generate populations of the system components and simulate their interactions in a virtual world. The main aim is to build agent based modeling framework in which theories of stem cell behavior can be modeled. Agent based modeling initially begins with the certain rules of their behavioral and choose to reconstruct from computational instantiation of their behavioral rules, the observe patterns of behavior. The main challenges of the project are mathematical modeling and formulating the properties of stem cell, conceptualization and developing the simulation of stem cell. To develop mathematical models, to construct simplified representations of biological processes and phenomena and to analyze and interpret them in biological fashion with predictions and hypothesis.
2.3 Introduction to Agent based Modeling
Multi agent system consists of number of agents, which interact with one-another. In general, agents will be acting on behalf of users with different goals and motivations. To interact successfully, they will require the ability to cooperate and coordinate with each other. The Agent based modeling specifically aims at searching for explanatory insight into the collective behavior of agents working on simple rules. Agent based modeling is a rule based, computational modeling methodology that focuses on rules and interactions among the individual components or the agents of the system. Decision making of agents mostly relies on optimizing, inquiring, imitation and repetition properties to the external environment. Agent based modeling is such a methodology to grow emergent phenomena using models to understand how simulated agents have needs and to implement those behavioral principles in simulated agents. It allows us to test the results against reality. i.e. classical experiments in the simulated environment. Agent based modeling paradigm consists of the observer which instantiates the world, the agents and the simulation environment including the phases setup, runtime loop and exit.
What is an agent?
There is no particular definition for the word ‘agent’. Some consider
independent components to be an agent. An individual component like software model’s agent behavior can be limited from simple reactive
branching decision rules. The agent label is preserved for the components that can study their model and changes of their behavior in reactions to their experiences.
In 1997, Casti demands that agent having both base level rules for behavior and also to the higher level set of rules to change the rules. The basic rules provide response to the environment along with the rules to change the rules. In 2000, Jennings provided a computer science point of view that an agent is having stamina to make its decision individually. This requires agents to be active responders and planners rather than purely passive components.
2.3.1 Characteristics
- Agent is a identifiable with a individual pair of characteristics and rules of their behavior and capability of decision taking
- An agent can function independently in its model and communicates with other agents for certain limit of situations
- An agent is a social communicator and responder with other agents by having protocols with other agents as far as they are in mutual communication
- Agents have ability to identify and differentiate the characteristics of other agents
- An agent is placed in an external environment in which the agent interacts in addition to other agents
- An agent having necessary objectives to achieve their defined goals with respect to its behavior. This gives an agent to compare the outcome of its behavior to the goal it is trying to achieve
- Agent is flexible and having capability to study and hold its behavior based on experience. This need certain form of memory. An agent may have rules that modify its behavior
Figure 1
2.3.2 Advantages
- Capture emergent phenomena
- Provides the natural environment for the study of certain systems.
- Is flexible particularly in relation to the development of geo-spatial models
Emergence is a phenomenon which is undetected and unpredicted behavior which is unknown to the computer science such as adoption, self organization.
Neural networks are the popular instances of complex structures that are manageable of organized behavior. As a result of parallel interaction of many interconnected neurons.
Interaction between agents is complicated nonlinear discontinuous or discrete by the other agents. These can be particularly defined discontinuity of individual behavior is difficult. For example, differential equations.
The ability to design the heterogeneous population of agents with an agent based model is significant. Agents can hold any type of unit from which naturally collection of individual units can be formed from the bottom-up.
2.3.3 Limitations
Adapting the agent based model approach for geo spatial modeling is reduced by some limitations. Though common to all modeling techniques, one issue relates to the purpose of the model, a model is only as useful as for which it was constructed and has to be developed at right level of description for every phenomena. For example, a system based on human beings will involve agents with potentially irrational behavior subject to complex and choice psychology. The factors are difficult to justify, quantify which complicates the implementation and development of a model. Agent based model having a limitation by constructing the large systems modeling.
Critics of complex theory point out that variety behavior exhibited exposure by mathematical and computational model are found in real world. In this, agent based model is very sensitive to initialize conditions in small variations in interaction rules.
2.4 Introduction to Mason
Mason is a multi agent simulation toolkit designed to support large number of agents effectively on a single machine. Mason stands for Multi-Agent Simulation of Neighborhoods and Networks. Mason is written in Java to make it easy to run on heterogeneous system environments. The models generated can be serialized to disk and we also can save the checkpoints on the simulation developed.
Mason is a Multi Agent Simulation of Neighbourhoods or network or something (MASON) tool kit . MASON is a quick custom event multi agent simulation library core in Java and developed for large custom purpose java simulation. It was developed by the George Mason University and designed by the Sean Luke, Gabriel Catalin Balan, Keith Sullivan, and Liviu Panait, with help from Claudio Cioffi-Revilla, Sean Paus, Keith Sullivan, Daniel Kuebrich, Joey Harrison, and Ankur Desai.
Mason has a particle domain specific feature. Mason contains model libraries and visualization libraries and can generate hexagonal 2D, 3D or network data. Mason user friendly you can run models out of visualization and with visualization and many types of data and you change between them.
Mason GUI is somewhat different from regular GUI’s. When compile the code the mason doesn’t show the implementation and execution of all the line which boring part in other compilers. It display it GUI straight forward . In the above console there are some buttons at the bottom which paly, pause and stop beside there is number indicator which display the number cycle according user desired meaurment which can select right side bottom drop down list , the list contains Time, steps , rate and none.
This button is used to initialise a simulation by the function start() . This function is invoked on the SimState followed by a continuous calls to step the schedule
This button is used to pause or unpause the simulation. While in the paused state, the play button change to step button which gives you step through the simulation.
The Stop button is used to stop the running of simulation by calling the function finish(). This function is called on the SimState.
This drop down box helps to play the current simulation by time, step, rate or none is shown at right bottom of the console.
This tab used to display the information of the choice of an HTML file
This tab shows various widgets for manipulating the speed of the simulation, automatically pause or stop at a certain time, etc
This tab shows the list of all visualization displays.
This tab shows the list of all current inspectors of objects in the simulation. At present we have none, we’ll see them presently
The Console Menu, which allows to save and load simulations, choose new simulations, or quit.
This Menu displays 2D overlays each field portrayer on the top of one another. This layer also helps to display or hide various field portrayals.
This button used to record the motion of simulations. It creates a movie of simulation.
This button is used to make the picture of simulation and also save the captured image as a bitmap or pdf images
This option button gives various options in drawing.
This field helps to zoom in and zoom out of the display. This vary according to the value in the scale.
2.4.1 Features
- It is developed completely java to support the heterogeneous computer environment.
- Mason models are free from visualization. It can altered or changed easily.
- It can generate similar results across platform.
- It is compatible for 2D and 3D visualization.
- Mason is has top quality random number generators and it having feature of fast implementation.
There are many good simulations having the above top features but mason not in the Ultra light simulation category.
2.4.2 Advantages
- It is having separate 2D and 3D visualizations.
- It is platform independent
- It provides good support to million agents without visualization
- It can be embedded into larger existing libraries
- Mason is very speed and its control having its own inspection features
2.4.3 Disadvantages
- Mason is not multi-processor tool kit.
- Mason run and give results only in single processor.
- Mason can’t easily understand by the beginner java coders.
- There is not distribution between the multi processors and multi computers required single unified memory space.
- Even though it is very easy to use but not having plugins for the Eclipse or Netbeans
2.4.4 Applications
Mason is used by George Mason University for 2 years already it have been used for many simulation tasks from micro air vehicle coordination to models of collective behaviour in society.
Networks intrusion is an agent model to observe the computer network security issues, first developed in ASCAPE and then it is moved to Mason to test difficulty and speed of the porting system.
Urban Traffic simulation is also developed in Mason to study the flow of traffic from a multi agent perspective.
Recently, ant like robots swarm behaviours perform central place food foraging.
Generally we use laboratories and life studies to examine the interaction of pathogens and infected hosts. But due to deadly effects some diseases like anthrax are not possible by live studies. The anthrax model initially developed in Swarm and it is ported to Mason by having an advantage of Mason speed and its control.
2.5 Introduction to stem cells
Cell is the basic unit of structure in all organisms and also basic unit of reproduction. The cell is the basic unit of life. The structure of the cell is depends upon the membrane covered on the cell. A tissue has to be self-renewing: in order to guarantee a continuous replacement of cells that die. The populations of daughter cells then includes cells that remain undifferentiated, i.e., self-renew the identity of their parent, and cells that differentiate, i.e., change their properties. Cells with the potential for both self-renewal and differentiation are called stem cells.
A stem cell is an undifferentiated cell having capabilities:
(i) proliferate
(ii) self-renew, i.e., able to go through numerous cycles
(iii) Cell division should be maintaining the undifferentiated state,
(iv) Ability to produce a progeny of distinct cell types,
(v) Recover the tissue after injury or disease
3 Literature Review
An agent based model is an individual class of computational models for
simulating actions and interactions of autonomous agents [1]. Initially, agent based model was developed in 1940’s but it needs computation procedures.
It did not spread till 1990. When we go back to the past, Von Neumann
machine having a capability of reproduction. The concept was later developed by Neumann’s friend Stanislew Ulam, a mathematician proposed that this machine should initially put on paper as a collection of cells on grid.
This suggestion raised the curiosity to Nuemann who drew it up and created the first of the devices and later named ‘Cellular Automata’. Later in 1970s and 1980s, earliest agent based model concept was Thomas Schelling segregation model. In 1980s Robert Alexandro used agent based manner to finalize a winner in a tournament of prisoner’s dilemma. More recently, agent based simulations developed on human cognition known as cognitive social simulations.
Agent based modeling is used largely in biological applications like threat of bio warfare, include population dynamics, vegetation ecology and landscape diversity, 3D breast tissue formation model, spread of bacteria in human body, biochemical reactions and many more[1]. Agent based evolutionary search or algorithm is the current research topic in dealing complex optimization problems.
In 2016, Niazi in Complex Adaptive System Modeling says “complex systems are networks made of number of components that interact with each other in a non-linear fashion. Complex System may raise and evolve through self organization such that they are not completely regular nor completely random gives the development of emergent behavior at microscopic scale”[2].
3.1 Approaches
Agent based modeling is a different way to view your organization. Traditional model approaches can deal with organization employees, customers, projects and products and defines the behavior like memory, place, actions and puts them in a accurate environment with established connections and executes the simulations.
In recent years, in biology they are mostly focusing on relationship between stochastic nature of molecular interactions and variability of cellular behavior. Describing this relationship, it is necessary to develop a computational approach at the single level cell. We took bacterial chemo-taxis as a test-bed to our approach. It is the one of the prominent characteristic in biological system. Each bacterium is an agent equipped with chemo-taxis network. In a 3D environment, swimming cells are free to move.
However, the agent based model as a method applicable only to a large number of agents. Agent based model works perfectly than any other for the models of problems in manufacturing logistics and business process.
Agent based method is also well suitable for the field epidemiology. In this model the agents are susceptible recovered to diseases. Here, the agents are the people. So, the agent based methodology catches the contacts between people explicitly which is a better way to forecast the disease to spread all over the population.
3.2 Agent Based Modeling and Simulation
Agent based simulation is a fast developing method in both in the social and science as well as in the engineering sciences.
Build an agent based simulation model
Agent based modeling brings unique aspects when access into the fact that ABMS takes foremost the perspective of agent first. But in contrast the process perspective in the traditional simulation modeling. ABM needs knowing the agent behavior theory and it identify relationship between the agent and agent theory interaction. It needs the necessary agent related data. As a whole model, it validates the agent behavior model in addition to the model. It analyze the output from the standard point.
Agent based modeling simulation and modeling life cycle
It is as the more common model software development process. There are several highly leaved stages in development of the ABMS life cycle. The part development and articulation stage defines the project goals. Function and the model structure are defined in the design stage. The design stage is helpful to build the implementation stage. The operationalization puts the model in use. The ABMS project complexly revolves in these stages many times with more detailed models for every iteration.
The agent modeling will be done in minor or small on the desktop or in large cluster computers. Successful projects first initially starts with small desktop ABMS tool and later developed into large scale ABMS.
Desktop agent model can be developed in few number of days by a single computer. It can be used to learn how to develop agent modeling, testing and designing concepts.
Large scale ABMS is an extension of extending small desktop agent model. It is modeling beyond the properties of the simple desktop environment and allow large number of agents to put them in highly developed behaviors. Large scale agent model need more master level skills and development resource than the desktop environment.
3.3 ABMS Modeling toolkits
Now a days, there are many ABMS software environments which are freely available. In 1944, Sanata FE institute developed the Swarm as the first ABMS software development. By following these swarm innovation, the Recursive Porus ABS tool kit is developed. It is widely used in the social simulation applications. This is the leading open large scale ABM and simulation library. It converts their library components into their own programs or by visual scripting environments. There are three versions: repast for python (repastpy), for java repast, for Microsoft .Net framework repast .Net. In the above, the repast for python is cross platform and allow users to develop models by using a graphical user interface and python scripting helps to develop agent behaviors. Repast Java is 100% java modeling environment. It supports for the large scale agent model development. It includes many different features and it is discrete event scheduler. Repast .Net is 100% C# modeling environment which imports all the features of repast Java to the Microsoft .net frameworks. There are many other tool kits which are swarm (SDG 2006), Netlogo, Mason are the many others properties tool kits developed due to public research and development investments helps to develop many tools for ABMS tool kit.
3.4 Technologies
In order to access the web service oriented applications through complete phase of product development. The next generation need intention to draw heavily on agent based and emerging technologies. The technologies and technique in ABM are very important to understand there is nothing special or static language for agent based modeling. ABM are having very large variety in architecture and type of behavior and agent number and so on. However, other modeling methods are used frequently inside and outside the agents. The design patterns of many ABM are common.
• Agent based modeling and Object oriented design
There are lot of similarities between object oriented software design and agent based simulation modeling. In programming the developer thinks program is a sequence of steps and that motivation takes him to the goal. In the same way, discrete event modeling represents the system and perform over entities by the sequence of operations. So, unlike the procedural programming, almost replaces the object oriented programming.
• Time model – asynchronous and synchronous
While considering the agent based models, we have to differentiate the asynchronous and synchronous time models. Asynchronous time means there is no grid on the time axis and arbitrary movements of events may occur exactly when they are to initialize. Synchronous time assumes that things are only happened during discrete time steps. Nothing happened in between.
Agent based models have either of asynchronous time and time steps.
• Space in agent based models
Agent based models uses more space. Space in agent based models are used to visualize the agents. Even in the models the location and movement of the agents are not important in the model logic view.
2016-11-25-1480034969