Acquisition of a Multi-Processor Computer for Computational Physics and Structural Biology

点击翻译按钮获取中文摘要

Project Summary This proposal seeks funding for a new multi-processor computer for structural biology and computational physics at The Rockefeller University. Structural biologists need increasingly faster computers with memories in excess of 256 MB because of the size of data sets from X-ray crystallography, NMR spectroscopy, and electron microscopy that need to be processed and fitted by complicated models. In addition, the main subject of structural biology, understanding the relation between structure and function, frequently depends on large computer simulations such as molecular dynamics and protein structure modeling. Likewise, calculations in computational physics also require more computing power and memory than currently available in order to simulate larger systems at a higher resolution and to explore larger parameter spaces. The projects of the eleven major users include crystallographic, NMR, and electron microscopic studies of structure, dynamics and function of proteins. The proteins under study are involved in signal transduction, chromosomal replication, transcription, and transmission of signals through cell membranes. Additional work by the major users includes protein identification and characterization by mass spectrometry, simulations of protein, DNA, and chromosome structure and dynamics, and modeling of neuron function and visual cortex architecture. The researchers are currently using small workstations, many of which are more than three years old and have less than 64 MB of memory. Speed and memory of these workstations are increasingly restricting. The existing and future needs can best be met by a multi-processor computer such as the Silicon Graphics Power Challenge XL system with 12 R10000 processors and 1.25 GB of memory, which we propose to purchase. A multi-processor server computer is more suited to a small group of independent users than is a large number of client workstations because of the speed deriving from fast processors and from coarse parallelization, because memory and software can be shared among several users, and because of the savings in support and maintenance. The use of a locally located multi-processor computer is also more appropriate than the use of national supercomputer centers because of the faster and more flexible networking, sustained availability of computing power, smaller backlog, and access to local consultants. In comparison with the current equipment, the new server computer will provide a four-fold increase in computation speed and will allow effective execution of very large jobs by avoiding swapping to hard disks. The server will also free the current client workstations for their original purpose, interactive and graphical work, both of which are impeded by the large jobs running in the background. As a result, current computational bottlenecks in protein structure refinement in X-ray crystallography, NMR, and electron microscopy will be eliminated. The server will thus help in the structural determination of eukaryotic RNA polymerase system, DNA polymerase complex, E. coli RNA polymerase, voltage-gated potassium channel, Abl SH(32) dual domain, and other proteins. It will also allow many simulations that are currently out of reach. For example, NMR and molecular dynamics will be used to determine motions of protein domains and large loops on the nanosecond time scale; proteins will be identified rapidly by mass spectrometry and database searching, which will be useful in the analysis of sequences produced by genome projects; protein structure modeling by homology will be improved by simultaneous optimization of the model and the alignment, which will allow a significantly larger fraction of protein sequences determined by genome projects to be modeled with useful accuracy; molecular orbital calculations of retinal and its environment will elucidate the opsin shift mechanism in visual pigments; the experimental data on the condensation of eukaryotic chromosomes will be analyzed statistically and the models for this process will be proposed; and more realistic neural models will be solved that may result in better understanding of the function of the brain. The server will also provide graduate students and postdoctoral fellows at The Rockefeller University an opportunity to program and exploit the latest generation of mini supercomputers. This will be made easier and encouraged by the existing infrastructure for networking and consulting. In addition, the new server computer will facilitate planned recruitment of new faculty in the areas of structural biology and computational physics. And finally, it will stimulate the growing interest of all the faculty in structure-function studies in biology and in the problems at the interface between physics and biology.