Stochastic Models of Complex Systems


In many situations of physical interest one would like to concentrate on slow dynamics occuring on a long time scale and neglect any rapid motions which play little role in large global changes in the system. We are currently examining the use of generalized Langevin equations constructed by eliminating fast degrees of freedom to describe the long time structural changes in biomolecular systems. This approach can be implemented in a biomolecular system by modeling flexible regions of a generic protein as hinges and more robust secondary structures, such as helices, quasi-harmonically. By eliminating short-wavelength modes of the secondary structural elements, one can obtain generalized Langevin equations for the angular motions and other slowly varying modes of the system. A natural question is therefore to what extent can the dynamics be simplified by making simple approximations for the memory kernel? Initial analysis of a harmonic rotor has been conducted to explore this issue. These studies indicate that memory effects are extremely important in the dynamics of the rotor, which suggests that simple stochastic dynamical methods, such as Brownian MD, may be inappropriate for simulations of in vacuo biomolecules.

Of course, the nature of the dynamics in solution is often qualitatively different than that in vacuo, and interactions with the solvent can lead to rapid loss of memory depending on the strength and time scale of system-bath interactions. We are currently attempting to extend our analysis of the harmonic rotor to include solvent interactions. The major challenge that must be addressed is how one can utilize stochastic models when there is a clear lack of separation of time scale between the system-bath interactions and the motions of the system. Similar problems exist in many models of polymer dynamics, such as the Rouse or Rouse-Zimm models which attempt to incorporate the effect of the solvent in an approximate fashion. A number of fundamental questions concerning the effects and appropriateness of eliminating fast variables are targeted. These include: What slow modes of the secondary structural elements of the protein couple significantly to the hinge dynamics? What is the nature of the ``noise'' and memory in the angular dynamics? Does a hydrodynamic-like mode-coupling theory of the memory functions make sense? These issues should help clarify the feasibility of constructing simple analytical models of protein dynamics.

Created by Jeremy Schofield
Last modified: Tu Dec 3 17:53:20 EST 2001