One of the computational grand challenge problems is to develop methodology
capable of sampling conformational equilibria in systems with rough
energy landscapes. If met, many important problems, most notably
protein folding, could be significantly impacted. In this work,
a new approach for sampling the conformational space of complex systems
will be presented. The new method builds on the variable transformation
or REPSWA (Reference Potential Spatial Warping Algorithm)
approach previously introduced by us
[ Z. Zhu, et al.,
Phys. Rev. Lett. 88, 100201 (2002) ],
in which a change of variables in the canonical partition is made that
effects a shrinking of barrier regions and an expansion of attractive
basins in the phase space. The approach was shown to yield efficient
sampling of very long alkane chains. However, in order to treat biological
systems, several new developments are needed. In particular, new extensions
for handling branching and strong short-range nonbonded interactions are
discussed. The latter is achieved by the introduction of a dynamic
transformation scheme that senses the presence of barriers due to
the close proximity of neighboring atoms. The performance of these new
developments is tested on branched alkanes, solvated polymers, and
a poly-glycine chain.
