Mike Holcombe - MSc Projects 2008-9

WMLH-1. Modelling the innate immune system – collaboration with School of Medicine (Cell Biology Unit and Cardiovascular Research Unit)

New insights into the molecular mechanisms by which the basic immune system works are being uncovered by cell biologists.

This project will develop an agent-based model of part of these pathways in conjunction with new experimental work. We will use the FLAME system to build the model. (http://www.flame.ac.uk)

There are 2 projects. One is with Prof. Eva Qwarnstrom and will look at new results from research into a new receptor, recently discovered by the group and the other will investigate the tribbles protein interactions with the MapKinase pathways with Dr. Endre Kiss-Toth

Further information can be found in these papers:

1. Formal agent-based modelling of intracellular chemical interactions, Mark Pogson, Rod Smallwood, Eva Qwarnstrom, Mike Holcombe, BioSystems 85 (2006) 37–45

2. Introducing Spatial Information into Predictive NF-kB Modelling – An Agent-Based Approach, Mark Pogson, Mike Holcombe, Rod Smallwood, Eva Qwarnstrom, PLoS ONE 3(6): e2367. doi:10.1371/journal.pone.0002367

A knowledge of C programming is required.

Back to top 


Project description.

WMLH-2: Simulations of agent-based systems of bacterial systems and analysis of their results

A number of research projects in the Department involve the simulation of massive multi-agent systems – applications include biological models – tissue growth, social insect networks, molecular biology of bacteria, particularly E. coli. The E. coli work is a collaboration between a number of microbiology labs (Sheffield, Amsterdam, Max Planck Institute Magdeburg) and ourselves. It is studying the detailed genetic mechanisms that come into play as the bacteria experiences conditions moving from aerobic (plenty if Oxygen available) to anaerobic (no Oxygen).
We have developed a detailed agent-based model of these conditions – representing each key molecule in the cell as an agent. 
These massive simulations – maybe involving hundreds of thousands of agents generate a lot of data. Trying to extract useful information from this data is a challenge.
 
We will evaluate a number of techniques for analysing stochastic simulation data and investigate one of these models in detail. The main purpose of the project is to provide an environment where this analysis can be done and the data visualised in suitable ways, including accurate graphical images of the cell’s molecular activity.
 
Resources.
FLAME - Flexible Large scale Agent-based Modelling Environment
 
Information provided by supervisor.

Back to top


Project description.

WMLH-3 Simulations of agent-based systems of economic and social systems and analysis of their results

A number of research projects in the Department involve the simulation of massive multi-agent systems – applications include biological models – tissue growth, social insect networks, molecular biology of bacteria - and economic and social systems – economic markets of various types.
The EURACE project is an EU funded project involving economics researchers in Italy, France and Germany. The model is an integration of 5 different models of markets – retail, labour, credit, financial and capital markets and their interactions.
These massive simulations – maybe involving hundreds of thousands of agents generate a lot of data. Trying to extract useful information from this data is a challenge.
 
We will evaluate a number of techniques for analysing stochastic simulation data and investigate one of these models in detail. The main purpose of the project is to provide an environment where this analysis can be done and the data visualised in suitable ways, including graphical images of the different markets activity and to demonstrate the outcomes from policy experiments – e.g. what a change in interest rates, taxation rates etc. would cause.
 
Resources.
FLAME - Flexible Large scale Agent-based Modelling Environment
 
Information provided by supervisor.


Back to top 

Project description.

WMLH-4 FLAME for beginners

This project will look at the issue of providing a simple entry into using the FLAME environment, providing inexperienced users – often biologists -with all the support they need in order to get started with simple model building.

It will involve creating a set of resources and mechanisms for installing as automatically as possible the FLAME environment, together with the development of simple interfaces to build and run models. Excellent documentation will be needed if the users are to be able to user the environment to its full potential

Resources.
FLAME - Flexible Large scale Agent-based Modelling Environment
 
Information provided by supervisor.

Back to top


Project description.

WMLH-5 Generalisations of Hidden Markov Models

A Hidden Markov Model (HMM) is a statistical method used frequently in areas such as speech technology. It is based on the idea of a state machine where the transitions are labelled by probabilities. There are a number of hidden states and there are transitions between theses states, different transitions are triggered with different probabilities.

From each state there are other transitions to outputs, again these fire with different probabilities. The probabilities are trained in order to generate the desired results. Such systems are used in speech recognition, handwriting recognition etc.

X-machines are generalisations of state machines that contain an internal memory – this greatly enhances their capabilities and it is possible to construct efficient machines since the number of states needed is greatly reduced. If we can figure out a way to utilise X-machines here then the efficiency gains could be very great.

The project will investigate whether it is possible to make the same gains with HMMs possessing an internal memory. This is an ambitious project that is not guaranteed to work – however we should learn a lot about the possibilities and the limitations of the approach.

Back to top

Project description.

WMLH-6 Using X-machines to generate Celerity features – suitable for Genesys testing specialist

Celerity is a Ruby library for easy and fast automation of web application testing. It is currently used in Genesys.
 
http://celerity.rubyforge.org/

Celerity Features

  • Fast: No time-consuming GUI rendering or unessential downloads
  • Scalable: Java threads lets you run tests in parallel
  • Easy to use: Simple API
  • Portable: Cross-platform
  • Unintrusive: No browser window interrupting your workflow (runs in background)

This project will investigate how information derived from an X-machine description of a web-based application can be used to automate the setting up of the Celerity scripts.

Back to top