| Project Detail |
The dynamics of biological systems, from proteins to cells to organisms, is complex and stochastic. To decipher their physical laws, we need to bridge between experimental observations and theoretical modeling. Thanks to progress in microscopy and tracking, there is today an abundance of experimental trajectories reflecting these dynamical laws. Inferring physical models from noisy and imperfect experimental data, however, is challenging. Because there are no inference methods that are robust and efficient, model reconstruction from experimental trajectories is a bottleneck to data-driven biophysics.
I will bridge this gap by developing practical algorithms that permit robust and universal inference of stochastic dynamical models from experimental trajectories. To this aim, I will build data-efficient tools to learn stochastic differential equations and discover physical models, employing methods from statistical physics and machine learning. The main focus of SuperStoc will be in resolving models with high precision from limited trajectories. To assess the efficiency of the methods I develop, I will design information-theoretical frameworks to quantify how much can be inferred from trajectories that are short, partial and noisy. The convergence of the resulting algorithms will be backed by mathematical proofs and numerical simulations in realistic conditions.
I will apply these new tools to several key open biophysical problems where existing methods are failing: condensate-mediated interactions between genomic loci, cellular mechanosensing in confined environments, pattern formation in embryo development, and visual interaction between fish leading to collective motion.
The resulting algorithms will be implemented into a software designed to be useful for the broad soft biological matter community. By proving that one can do more with the same data and providing tools to do so, SuperStoc will help bridge the inference gap towards data-driven biophysics. |
