Protein folding with Charmm on LegionSan Diego 07 March 2001 At NPACI, a US research alliance, a large-scale molecular dynamics calculation has compressed a month's computing into just 36 hours, using a parallelized version of the Charmm molecular modelling and simulation package and the Grid operating system Legion on six supercomputing systems with five distinct architectures located in four different states in the USA.
Preliminary results of the calculation were announced at the annual All-Hands Meeting of the National Partnership for Advanced Computational Infrastructure (NPACI), held in San Diego, February 25-28, by leaders of the NPACI alpha project Protein Folding in a Distributed Computing Environment.
The calculation revealed the process by which a small protein (Protein L) folds up into its normal, three-dimensional configuration. The folding process occurs in nature every time a protein molecule is manufactured within a cell. "The biophysics of folding must be understood in detail," NPACI r Brooks said, to use the information in developing ways of interacting with proteins to cure such diseases as Alzheimer's or cystic fibrosis. Charmm was used to compute the folding free energy of Protein L, a small protein with 585 atoms, by using groups of 16 processors in a tightly coupled parallel calculation. Each gang controlled the exploration of one of approximately 200 distinct regions of conformational space (sampled in a loosely coupled parallel paradigm) to calculate the folding free energy surface. The supercomputers used in the calculation were all pulled together by Legion.
With the Legion metasystem all the scientist needs are compiled codes for each platform that may be used, a script for dispatching jobs, and another for keeping track of results. Legion does everything else, either through an easy-to-use Web interface or a more traditional Unix command-line interface. The system was developed for academic scientists by Grimshaw and his group at Virginia. Legion is now being commercialised by Grimshaw's company Applied MetaComputing, for use beyond the academic community. The protein-folding problem was run on Legion's NPACI-net, which links 11 major computer systems, including the two PACI partnership leading-edge sites (SDSC and NCSA). The calculation used six systems at SDSC, Caltech, and the universities of Michigan, Virginia, and Texas. Blue Horizon, the 1.7-teraflops IBM supercomputer at SDSC, did the majority of the calculation on 512 of its 1,152 processors. Also used was an HP V2500 system at Caltech (using all 128 processors), with additional calculation done on a Sun Enterprise 10000 system at SDSC (32 processors), IBM systems at Michigan and Texas (totaling 56 processors), and the Centurion Alpha cluster at the University of Virginia (32 processors).
Ad Emmen |
 |