Modeling the responses of heterogeneous cell populations to molecular perturbations such as drug therapies offers the chance to decipher molecular processes of cells and tissue, facilitate the discovery of new drugs and tailor treatments to patients’ individual molecular profiles.
Recent approaches leverage for instance optimal transport or diffusion-based models to predict the trajectories of individual cells in isolation. However, cell responses do not only depend on the internal states of the diseased cells but also on their interaction with their cellular environment. Therefore, using patient samples collected at the university hospitals of Zurich and Hamburg, we will develop new flow matching and diffusion models to predict the trajectories of cells in the context of their tissue environment. More broadly, we aim to create a general framework and algorithm to learn the temporal behavior of interaction particle systems – with potential future applications in meteorology or fluid simulations.