A second look at canonical sampling of biomolecules using replica exchange simulation

The replica exchange approach, also called parallel tempering, is gaining popularity for biomolecular simulation. We ask whether the approach is likely to be efficient compared to standard simulation methods for fixed-temperature equilibrium sampling. To examine the issue, we make a number of straightforward observations on how "fast" high-temperature molecular simulations can be expected to run, as well as on how to characterize efficiency in replica exchange. Although our conclusions remain to be fully established, on the basis of a range of results in the literature and some of our own work with a 50-atom peptide, we are not optimistic for the efficiency of replica exchange for the canonical sampling of biomolecules.