Since mechanical forces acting between bacterial cells in a dense colony are notoriously difficult to measure directly, we will use an indirect approach. We will first generate videos of growing colonies with high spatial- and temporal resolution. Next, we will compare the experimental data to computer simulations with interaction forces treated as “parameters” of the model, and select the model which fits the data best. This will allow us to determine interactions such as adhesion and friction between bacterial cells in the environment relevant to what bacteria experience in colonies and biofilms. Our approach has an advantage over earlier studies, which focused on measuring mechanical properties of isolated cells under non-physiological conditions, in which such interactions may be significantly altered. The first part of the project will benefit from the Dioscuri center experimentalists’ expertise to generate a representative dataset of bacterial colonies in different experimental conditions (grown on agar pad, in catheters, etc.) and different imaging techniques (fluorescence, bright-field and phase contrast). The dataset will be fed into a convolutional neural network in order to obtain high-quality segmentation of the cells within colonies. From those segmentations various properties can be extracted: evolution of the colony roughness, position and speed of the individual bacteria and even division time distribution (which require to follow bacteria and their offspring through several division cycles). This data will be used to fit a computer model where different types of mechanical interactions - in between cells and with their substrate - are accounted for. The forces will be modelled using functions whose shape can be described by a few parameters. A large number of simulations of this model will be run and the model will be iteratively fitted to the data by minimizing the difference between the simulated and observed colony. The outcome of the project will be a computer model of bacterial colonies capable of making quantitative predictions of how such colonies grow on different substrates. Longer-term, we hope the model may help to develop new strategies of reducing bacterial growth of surfaces.