Team:Dundee/Software

From the onset of the project, the modelling team had to carry out investigative research to understand the biological processes that were involved in the project to ascertain appropriate modelling methodologies to adopt. Due to the unique combination of factors involved in the project, it became evident that there were considerable quantities of mathematical variables that defined the behaviour of the cellular interaction. From a software development perspective, it was attractive to take the mathematical axioms of behaviour, and to wrap them within a bespoke modelling solution to provide a visual representation of the project.

CellSim is a cross-platform, 2D, cellular automata [4] (CA) modelling tool that was developed to demonstrate, in a graphical manner, how the synthetically modified E-coli and C.diff cells could interact in a simulated environment. The application was designed to simulate the cellular interactions whilst also providing a customisable simulation environment that could enable the tool to be used in other, similar projects. CellSim was developed to enable it to be executed on either Linux or Windows based systems, and is released under the GNU GPL 3 license that permits copying, modification and redistribution of the program. The source code and full class diagram is provided within the download archives for educational purposes as well as encouraging others to build upon the initial release.
The current development version (1.0) provides an environment in which the following aspects can be simulated and defined by the user:

• Cellular Mitosis Rates
• Strains
• Colony Size, Initial Positioning & Colouration
• Medium Flow Effects
• Mucus Wall Effects
• Time Calibrated Rendering
• Nutrient Availability, Supply & Absorption
• Low Nutrient Mitosis Suppression

The program was developed around an MVC structural design pattern [3] with a subtle modification. The initial design adopted a classic MVC [3] architecture, whereby for each generation, each cell was updated within the model classes, and subsequently rendered within the view classes. Due to the large number of cells that the program is required to simulate, code profiling highlighted the inefficiency if iterating through all cells twice. Furthermore, the status of each cell is partly determined by neighbouring cells, and therefore a method was required to detect the presence of neighbouring cells and their status. To accomplish this through data provided solely within the model, for each cell, all other cells would require iteration and interrogation leading to a time complexity of at least O(n2). Due to the mitosis behaviour of cells, the source number of cells also increases in a quadratic manner. The quadratic mitosis behaviour cannot be alleviated, as it represents the real-world scenario, so a solution was sought to alleviate the compound quadratic time complexity that arose from the classic MVC [3] implementation.

The solution adopted was to carry out cell interrogation, cell update and rendering in a single pass, and establishing the presence of neighbouring cells by direct interrogation of the graphics memory as opposed to model data interrogation. By calculating the base address of the graphics memory in conjunction with the co-ordinates of the current cell, a direct memory address could be calculated and accessed to establish the presence of neighbouring cells, alleviating the need to iterate through all cells. By adopting these measures, the time complexity of the ‘cell evolution’ section of code was reduced to a constant, greatly improving the speed and number of cells that the program could comfortably emulate.

Cellular movement emulates fluid flow by imposing a circular pressure front that is circular in nature. This provides a simple model that affects each cell’s horizontal position by exerting a greater force on those nearest to the centre of fluid flow. To enable variable flow rates to be simulated, a normalised force was calculated from the cell’s co-ordinates (x, y) relative to the display screens height (Ymax), multiplied by a user adjustable force factor (f) and utilised to modify the horizontal position of the cell. By carrying out this calculation on each cell, a realistic, circular flow pressure is simulated. The new horizontal cell position is calculated from the following formula:

To provide a realistic and accurate movement of cells within CA environment, each cell contains co-ordinate members for the respective ‘x’ and ‘y’ position. These class members are defined as float data types as opposed to integers to permit gradual movements to be achieved. To render each cell, the co-ordinates are simply cast to integer types before being passed to the graphics interface. This permits cell movements to be calculable to fractions of a pixel, and provides an accurate representation of the behaviour required.

The application adopts the SDL (Simple Directmedia Layer) graphics API that was chosen primarily due to its flexibility, and is licensed under GNU LGPL version 2 [2]. Furthermore, the library provides binary compilations for a wide variety of operating systems including Windows, Linux and MacOS. To encourage development of the program, pre-developed Eclipse and Visual Studio 11 project files have been created, and are free to download and explore. Alternatively, the packages also include the pre-compiled binary versions of the application within the ‘Release’ folder of the download archive (ZIP) files.