Conformation-Family Monte Carlo (CFMC): An efficient computational
method for identifying the low-energy states of a macromolecule
Pillardy J, Czaplewski C, Wedemeyer WJ, Scheraga HA
HELVETICA CHIMICA ACTA 83 (9): 2214-2230 2000
Document type: Article |
Language: English |
|
|
Abstract:
A highly efficient method, Conformation-Family Monte Carlo (CFMC), has been
developed for searching the conformational space of a macromolecule and
identifying its low-energy conformations. This method maintains a database of
low-energy conformations that are clustered into families. The conformations in
this database are improved iteratively by a Metropolis-type Monte Carlo
procedure, together with energy minimization, in which the search is biased
towards investigating the regions of the lowest-energy families. The CFMC method
has the advantages of our earlier potential-smoothing methods tin that it
'coarse-grains' the conformational space and exploits information about nearby
low-energy states), but avoids their disadvantages (such as the displacement of
the global minimum at large smoothings). The CFMC method is applied to a test
protein, domain B of Staphylococcal protein A. Independent CFMC runs yielded the
same low-energy families of conformations from random starts indicating that the
thermodynamically relevant conformational space of this protein has been
explored thoroughly. The CFMC method is highly efficient, performing as well as
or better than competing methods, such as Monte Carlo with minimization,
conformational-space annealing, and the self-consistent basin-to-deformed-basin
method.