IGEM:IMPERIAL/2008/Prototype/Drylab/Data Analysis

=Model Fitting=

Alternative Models
Models of run velocity include a Gaussian Distribution, or the Maxwell Distribution which govern the velocities and energies of molecules. An exponential distribution may describe the memoryless characteristic of run duration. We will build up a database of models, for future model fitting.

Mechanical Model
We have created a simple mechanical model of motile bacteria, depicted in the figure below.

Fitting Models to Data
In this first level of inference, we apply Bayes' Theorem. We first assume a particular model, and go on to derive the parameters of our model which maximises the data obtained.

Mechanical Model
The following trajectories were fitted based on the above equations, and corresponding parameters determined:

In some cases, we can visually segment the cell's trajectory into two or more separate runs. In the case below, we see that if we fit the cell's trajectory to only one run and a single set of parameters, we do not obtain a satisfactory fit. However, if we introduce an orientation change at a user selected frame, we are able to obtain a better fit. This fit contains two runs, thus generating two sets of parameters for the two runs.

Assigning Preferences to Alternative Models
In this second level of inference, we use the evidence contributed by the data to compare fitted models. Using Occam's Razor, we are then able to deduce the best model which fits our data.