Learning the Non-Equilibrium Dynamics of Brownian Movies

Soft living systems such as cytoskeletal networks, membranes, and tissues are driven out of thermodynamic equilibrium by internal enzymatic activity. Measuring and characterizing the non-equilibrium properties in these systems is a major challenge, owing to the large number of interacting degrees of freedom. Typically, the experimental characterizations of such systems rely on tracking the trajectories of embedded or endogenous probes. However, it is not clear how to select appropriate tracer probes and how this choice affects the resulting characterization; in general, it is unknown a priori which degrees of freedom are most informative about non-equilibrium activity in the system. In this talk, we present a new approach that does not rely on the tracking of probes in the system. Instead, we directly learn the non-equilibrium dynamics from a Brownian movie of a fluctuating soft assembly, yielding force fields and entropy production rates. Our approach is based on a principled analysis that reduces the dimensionality of the system by identifying the most dissipative components. We will discuss how this approach performs in different scenarios inspired by cytoskeletal networks.