Sampling and scoring in protein structure refinement

PHYS 7

Michael Feig, feig@msu.edu, Department of Biochemistry & Molecular Biology, Michigan State University, 218 Biochemistry Bldg, East Lansing, MI 48824
The refinement of template-based protein structures obtained from comparative modeling techniques to near-experimental accuracy remains a great challenge in current protein structure prediction efforts. An iterative protocol that combines an efficient sampling procedure with a reliable scoring method is explored. Structure refinement is possible with such a protocol if the sampling function is able to generate at least some more native-like conformations at each cycle and if the scoring function is able to identify these conformations at least most of the time. Different sampling strategies including coarse-grained models, normal-mode based sampling and molecular dynamics simulations are compared and tested in the context of structure refinement with a number of scoring functions.