StatPhysBio

Pressé Group

Latest Projects

Our group now works in three focus areas: 1) we experimentally investigate bacterial hunting dynamics and hydrodynamics; 2) we develop and adapt tools from statistics (most recently Bayesian nonparametrics) to infer models of dynamics from spectroscopy and imaging data (from the level of single molecules to cells); 3) we build mesoscopic statistical mechanical models to understand macromolecular and solvent dynamics in crowded environments as well as address questions regarding molecular motor and enzyme efficiency.

1. Hydrodynamic hunters

Bacterial predation presents an interesting conceptual challenge. Bacterial predators must at once detect moving prey on the basis of limited information and, if their bacterial prey move, they must forecast their prey's future position. Thus, the search for prey appears as a difficult problem. Indeed our experimental work (to appear soon) has revealed few statistical signatures of a targeted search for prey by our predator (Bdellovibrio bacteriovorus -- BV). Instead, hydrodynamics may play a role in BV's hunting strategy. We have a number of microscopy experiments underway to understand bacterial hunting dynamics both in vivo and in vitro. More on this soon!

2. Data worth a thousand words

Current biophysical imaging and spectroscopy methods can probe time (< 10-6 s), length (10-9 m) and force (10-12 N) scales relevant to the life cycle of a cell. These methods have revealed that all steps involved in molecular biology's central dogma (transcription, translation and DNA replication) are intrinsically stochastic. Despite the wealth of experimental data, the ability to gain meaningful insight from experiments on such small scales is severely limited by fundamental challenges common to all biological systems: current methods cannot capture complex processes in their full multi-dimensional detail. At best, current experiments provide a small slit through the curtains of the intricate cellular theatre on display by probing complex processes along just one or a few observable coordinates. Building models from such limited data is a central challenge in biophysics. Current modeling methods -- which we call forward methods -- are ill-equipped to tackle this challenge. Forward methods begin by positing models (such as reaction networks, coupled reaction-diffusion or multi-state models) and subsequently cross-check the model conclusions with experimental data. There are several disadvantages to this forward approach. First, forward models are rarely expressed using measurable quantities. Thus, model features and parameters must be fit to data. Second, the data's rich structure does not inform the model because the model's form is presupposed from the onset. Thus, in principle, other models could have worked equally well. To address these questions, we develop inverse modeling methods to directly infer stochastic models of complex biological systems from imaging and spectroscopy data with as few adjustable parameters as possible. Most recently our work has focused on developing the tools of Bayesian nonparametrics, an exciting rapidly evolving area of statistics, to tackle problems in biophysics. More in this soon!

Through our analysis, we hope to learn about the following key issues that lie beyond the predictive ability of current models: How variable is protein structure and function inside living cells? What is the mechanism behind intracellular anomalous diffusion of proteins? How variable are protein assembly sizes? How do cells integrate noisy signal? Here is a concrete example of such a project:

2.1 Tackling the 'single molecule counting problem'

Protein-protein interactions are the basis for most biological information processing and cellular control. A quantitative characterization of these interactions is an essential prerequisite for developing a mechanistic understanding of cell biology and the disease states associated with defective protein complexes. Nonetheless, characterizing protein complexes as they occur in their native cellular environment is a major challenge since complexes can involve up to many tens of proteins within approximately a 10nm range. Thus, there is currently no routine way to determine how many proteins of type X, say, are in a given complex in living cells. Our single PI NSF MCB grant awarded in 2014 is precisely focused on developing a method that will determine the stoichiometry of protein complexes in living cells using available data from a technique called PALM (a superresolution microscopy method). The methodology we propose here is closely inspired by our group's work in inverse methods that we are simultaneously applying to other forms of spectroscopy (including fluorescence correlation and photobleaching). Our work on this topic has been accepted in PNAS as a direct submission (selected as "Research Highlight" -- see publications tab).

3. Macromolecules stepping on landmines

Along with our experimental collaborators (C. Bustamante and S. Marqusee at UC Berkeley), we have begun investigating how enzymes that perform highly exothermic reactions may rapidly dissipate the catalytic reaction heat evolved in order to sustain their high turnover efficiency. For instance, we have investigated catalase which can perform up to a million reactions per second and each one of these million reactions releases the amount of heat that would be required to unfold a small protein. These types of results motivate our current work on energy dissipation and its impact of macromolecular function.