The goal of this project is to develop a method for using an agent-based approach for modelling bacterial behaviour without the need for accurate parameters. The parameters, when available, are optimised using a genetic algorithm and are tested using data from experimental biological results. The parameter optimisation step, through genetic algorithms coupled to an agent-based model is new and adds an extra level of testing to biological hypotheses. To begin with a differential equation model was constructed to simulate basic interaction of the bacterial cells with a host. This was compared with a simple agent-based model. It was established that even though the two systems could produce a similar outcome, they are fundamentally different. The discrete aspect of agent-based modelling was further explored in testing a basic biological experiment with a goal to measure the frequency of the cells to move from not developing fimbriae (OFF) to developing fimbriae (ON). In experimental data an anomaly was found at roughly one in seven of the experiments. Computer modelling established that the cause was related to how many cells were present at the beginning of the experiment (assumption was a maximum of one per flask) and how many of those were fimbriate at the start (assumption was none). A combination of statistics and modelling showed that one in seven flasks contained two cells or more and that a higher percentage of the cells was fimbriate at the start of the experiment. The computer model was further enhanced to establish the best hypothesis on how regulatory protein and their antagonist are able to interact and bind to the regulatory sites on the DNA. In this case the focus was on fimB expression, responsible for turning on the switch to enable the production of fimbriae. From the modelling it was established that H·NS binds to two sites on the DNA, but its effect is curilUlative rather than cooperative. SlyA can bind to two other sites to partially remove the repression of H·NS on fimB expression. Other sites and further hypotheses were tested, among them the assumption that H-NS and SlyA are involved in temperature regulation of limB expression. The assumptions made on how the temperature regulation functions were proven to be invalid by both biological experiments and by computer modelling. A working hypothesis was not established. Finally to test if the agent-based model with parameter optimisation could be applied for different systems it was attempted to model chemotaxis. Reaction rates for the processes within chemotaxis are known and the computer model was adapted so the parameter optimisation would lead the bacteria to a location with an optimum amount of nutrients, providing US with panmeters we can compare with literature. Not all parameters matched up with those provided, but the bacteria were able to find an optimum and a potential for using the parameters absolute values was established. All in all an agent-based computer model that is capable of functioning with- out (accurate) parameters has been established and the model can be used in a wide variety of applications to test hypotheses and potentially predict biological parameters.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:595789 |
| Date | January 2011 |
| Creators | De Vries, Patrick |
| Publisher | University of Kent |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
Page generated in 0.0016 seconds