next up previous
Next: Appendices Up: FAMUSAMM: An Algorithm for Previous: Computational Performance

Summary and Conclusion


In this paper we have combined the fast SAMM method for approximate evaluation of long range Coulomb interactions in MD simulations of protein dynamics with the DC-1d procedure designed for the same purpose into a new and highly efficient MD method. Upon extended test simulations the resulting FAMUSAMM algorithm has been demonstrated to preserve and combine the favorable properties of its parent methods concerning accuracy, lack of sizable algorithmic artifacts, and last, but not least, computational efficiency. Thus, FAMUSAMM opens the way for extended MD simulations which properly account for the long range Coulomb interactions of large protein-solvent systems comprising several 10,000 atoms.

The algorithm has been implemented in the MD program EGO_VIII, which is based on the CHARMM force field, in a sequential and parallelized version and is available via Internet [51]. Whereas the current paper is exclusively devoted to the presentation and evaluation of our method, a previously published application [62] has already demonstrated its suitability for the quantitative analysis of experimental findings; there, the method has enabled an atomic interpretation of atomic force microscopy experiments on ligand-receptor binding [63]. As a check of the FAMUSAMM MD simulation performed in ref. [62] the rms deviation from the x-ray structure of a solvated protein (Streptavidin in water, 10969 atoms) has been monitored. After 0,5ns simulation an rms deviation as small as 1,5Å was found showing that, indeed, FAMUSAMM provides realistic descriptions of solvated proteins.

Helmut Grubmueller
Wed Apr 30 15:40:09 MET DST 1997