Modeling cell population dynamics with Schrödinger bridges

Sammanfattning

This thesis investigates the use of optimal transport theory for modelling cell population dynamics in simulated cancer cell imaging data. The work focuses on capturing biological processes such as cell movement, birth and death using stochastic simulation, building on the Gillespie algorithm. We propose a partial Schrödinger bridge formulation that combines entropy-regularised optimal transport with Markov chain transitions and dummy states used for the mass creation and disappearance. We generated simulated datasets with known parameters using a Gillespie-based simulator in order to evaluate our proposed methodology and test whether it can recover the underlying process rates. The results show that the proposed methodology can successfully capture population-level dynamics and estimate birth and death processes robustly, while movement estimation is more sensitive to prior parameters. They further suggest that the proposed methodology may provide a useful framework for inferring cell population dynamics from imaging data and for identifying conditions under which the model performs less accurately.