Research

Computations performed by chemistry

As evolved animals we use our brains to to make sense of our surroundings, but individual cells are also able to understand their environments without the use of any neural circuits.  Instead, they use networks of interacting molecules whose design looks, at first glance, disordered and non-modular. Our work asks how apparent molecular disorder gives rise to the emergent computational functionality cells need to thrive.  We develop theoretical frameworks that combine statistical mechanics, stochastic processes, and the language of machine learning to characterize the computational principles and thermodynamic constraints of biochemical systems.

Control and thermodynamics in soft active matter 

Cells continuously reorganize their cytoskeletal networks, which support cell shape and generate force, through a process of chemically guided non-equilibrium self-assembly.  Amazingly, they do this effectively using only locally accessible information, all while contending with molecular fluctuations and pressure to minimize their energy expenditure. Understanding how cells navigate these competing constraints to execute control is an open problem with significant physical depth. Our work develops theoretical and computational tools for this challenge, drawing on nonequilibrium physics, control theory, and reinforcement learning.

Biophysical modeling across scales

Many cellular processes span spatial and temporal scales, linking molecular interactions to collective cellular behavior through complex, nonlinear dynamics that often do not fall neatly into existing frameworks. Our work combines first-principles physics, stochastic modeling, and modern computational methods to build interpretable simulations that connect observations to underlying biology, often in close collaboration with experimental groups.