We study a theoretical model of the growth of sessile Escherichia coli colonies in the exponential growth phase (feasting), by performing Discrete Element Simulations in two dimensions. We find that mechanical interactions are sufficient for the formation of highly ordered mesoscopic structures, vulgo microdomains. Basic tools of analysis, such as the contact angle distribution or the radial distribution function do not indicate the formation of these microdomains. For this purpose, we employ a community detection algorithm on contact networks representing the colonies. We compare three different variants with a range of threshold angles, and evaluate their overall performance and correlation with more conventional measures. We find that a good threshold angle to discriminate “cohesive” from “repulsive” contacts between particles i and j, to be ct = |ui uj| 0.96. For high threshold angles, all three variants performed comparably well, whereas the naive dichotomization variant was clearly outperformed for lower threshold angles. We believe that this method opens new avenues to study morphogenesis and can be equally beneficially applied in related fields such as systems of anisometric particles.