Simbios Talk by Vijay Pande, Stanford University, December 6, 2006

Title: New methods for long timescale dynamics

One of the primary hurdles in biomolecular simulation is the long timescales involved: experimental timescales are on milliseconds to seconds compared with simulations on the nanosecond timescale.  Today, I will present an update on our efforts to break this impasse.  In particular, I will discuss our new approach to use normal mode methods to differentiate low frequency from high frequency modes, and perform dynamics on the low frequency modes, while performing energy minimization on the high frequency modes.  This allows one to greatly push the timestep, without losing the high frequency degrees of freedom.  Thus, this has all the benefits of locking down degrees of freedom (i.e. leaving only slow degrees of freedom) while still allowing for relaxation in high frequency degrees of freedom.

I will also give an update on our work with using Graphics Processing Units (GPUs) for dynamics as well as the OpenMM API project.