Skip the Sweep – Learning what matters in models of living systems
For a physicist, living systems are notoriously difficult to understand: typically, these systems are out-of-equilibrium, non-linear, spatially coupled, and involve many components. Yet, the standard methods to study these systems are derived from equilibrium physics that work best for linear(ized) and local systems involving either few or infinitely many components. These methods work surprisingly well, but not nearly good enough to fully understand non-equilibrium systems. In addition, they often rely on guessing a set of observables that should be tuned towards their experimental counterparts and/or large-scale numerical parameter sweeps of the model.
Over the last decades, there has been only incremental progress on modeling living systems using purely theoretical approaches, though in the same period experimental advances have provided us with vast data of a wide range of biological systems. I believe that we should make use of this data to address two central issues: what are the relevant observables for biological systems (can we do better than guessing)?, and can we find relations between model parameters and the observables when analytical approaches fail? I will present some ideas for implementing this approach and preliminary results using a Contrastive Embedding approach on a toy model.
This colloquium is sponsored by the Dean’s fund for Colloquia and Symposia
Host: XJ Xu
