Microbes employ a vast arsenal of tools to manipulate the environments in which they live. These manipulations affect the survival of other microbes and therefore microbial populations evolve in ways that reflect these social interactions. Interactions between microbes are particularly important in structured communities called biofilms. In this thesis, we study biofilm evolution through the lens of social interaction.

Microbial biofilms are heterogeneous assemblages that develop on many scales of time and space simultaneously. A major challenge in understanding biofilms, therefore, is developing an informative and tractable model. The thesis begins with a simulation based on an established paradigm, namely, a death-birth agent-based model (ABM) on a regular lattice. In addition to reproducing by replacing neighbors, all of the individuals utilize a shared resource that diffuses through the environment, though only some of them produce it. The realistic treatment of diffusion in this model leads to loss of the coexistence previously observed in similar game-theoretic models involving nearest-neighbor interactions.

In the non-local interactions model, the introduction of long-range interactions into a game theoretic model leads to the loss of biologically relevant emergent dynamics, arguing against the generality of game-theoretic lattice models of social interaction. In studying the effects of intermicrobial warfare on community structure, therefore, we instead take a mechanistic approach. Approximately 25% of Gram-negative bacteria possess at least one Type VI Secretion (T6S) system, which can be used to kill other microbes. Using an ABM that describes the local interactions between cells during T6S attack, we predict that the system can only be used to displace small or diffuse populations. We then use in vivo experiments to verify that the same phenomenon occurs in real microbial colonies.

These studies both required the development of ad-hoc ABMs. The process of creating and exploring such a model requires computer skills that are wholly independent from expertise in the biological problems at hand. We therefore conclude with a method for designing ABMs that requires minimal programming knowledge. The technique, which draws on the artificial intelligence field known as constraint programming, replaces step-by-step computer instructions with a simple list of user generated requirements.