Exploring Algorithm Space: Resolution Exchange Simulation and Other Exchange Variants

Daniel Zuckerman
University of Pittsburgh
Dept of Computational Biology

It is possible to pursue biomolecular equilibrium sampling in a number of new ways. The "resolution exchange" approach, analogous to replica exchange, enables configuration-swaps between independent simulations differing in *resolution* -- i.e., differing in the number of degrees of freedom. Canonical sampling is preserved at all resolution levels. Tests in molecular systems indicate that the approach is capable of crossing barriers in the inexpensive, low-resolution models. Substantial reductions in total-CPU usage are obtained. We also introduce a generally applicable "top-down" implementation of exchange simulation, which can be applied to conventional (temperature-based) replica exchange. Top-down simulation seems to guarantee a reduction in overall CPU usage. A related annealing-based algorithm will be discussed, as will a simple, rigorous and "structurally concrete" approach for assessing convergence.

Presentation (PowerPoint File)

Back to Workshop I: Multiscale Modeling in Soft Matter and Bio-Physics