HIGH-PERFORMANCE MOLECULAR MODELING AND GRAPHICS SYSTEM

点击翻译按钮获取中文摘要

Molecular modeling and graphics needs of NIH-supported researchers at the University of Kansas have overwhelmed the limited capabilities of the micro VAX II/Evans & Sutherland PS330 modeling system now in place. Research plans for the investigators listed below require substantially more computing power to exploit recent advances in modeling software and perform needed tasks in the areas of receptor modeling, solvated dynamics, linear density function (LDF) modeling of metallo-organic complexes, quantum- mechanical analysis of transition states, and related techniques. This request is for a Silicon Graphics IRIS (SGI) cluster comprising an eight- CPU server (with graphics capability) and a satellite high-performance graphics workstation. The cluster would provide a best-case performance of about 160 times the existing muVAX [weak case ~ 31 times] at a price of less than four times what the muV/PS330 system cost originally. The system proposed includes nine 25 MHz MIPS R/3000 processors (rated individually at about 4.0 Mflops), 256Mb main memory, two high-resolution graphics displays, and appropriate peripherals. A second satellite display is proposed as part of the local match. SGI models have recently emerged as the platform of choice in the molecular modeling area, with robust support from software vendors including Tripos, Biosym, BioDesign, Hare Research, and Polygen, as well as numerous academic software producers. Some products (e.g., Tripos Sybyl conformational "Search") can in their current releases exploit parallelism on multi-CPU SGI systems. We propose immediately to implement IRIS Sybyl and MacroModel, QCPE's AMPAC/MOPAC, Yeti/Yak, Amber, FTNMR, GAMESS, and GRID for seamless transition to the new environment, and acquire Biosym DMol for LDF analyses and Hare FTNMR/Dspace for NMR structure determination. The resulting configuration will provide dedicated computing capabilities comparable to that of a low-end Cray in a compact, modest-cost, easily-maintained configuration suitable for the limited number of investigations indicated herein.