NMR Data Processing and Structure Calculations Using Parallel Computers
The determination of protein structures from NMR data involves a considerable amount of computer time, both for the calculation and interpretation of spectra. Parallel computers offer individual laboratories super-computer performance at reasonable cost.
Reconstruction of NMR spectra using the maximum entropy method involves a large number of Fourier transforms. We have ported a conventional 2-D maximum entropy program, written in FORTRAN 77, to our Meiko Computing Surface (a transputer-based parallel computer). The major modification necessary was the implementation of an efficient data communication scheme. This is particularly important for the 2-D Fourier transform, which involves each processor communicating with every other during the transpose stage. We have also implemented a stand-alone parallel 2-D Fourier transform program which minimizes communication time by overlapping this with the computation.
Obtaining the 3-D coordinates of a protein structure from 2-D spectra is commonly performed using a combination of distance-geometry and molecular dynamics calculations. We have implemented an efficient systolic loop algorithm for parallel molecular dynamics simulations on the Meiko. The program has been modified to include the distance constraints that are derived from the NMR data.
KeywordsCommunication Time Maximum Entropy Method Slave Processor Master Processor Slave Process
Unable to display preview. Download preview PDF.
- 2.W. H. Press, B. P. Flannery, S. A. Teuklosky, W. T. Vetterling, “Numerical Recipes,” Cambridge University Press, Cambridge (1986).Google Scholar
- 3.K. C. Bowler, R. D. Kenway, G. S. Pawley and D. Roweth, “An Introduction to Occam 2 Programming”, Chartwell-Bratt, Lund (1987).Google Scholar
- 4.G. Fox, M. Johnson, G. Lyzenga, S. Otto, J. Salmon, and D. Walker, “Solving Problems on Concurrent Processors, Vol. I: General Techniques and Regular Problems”, Prentice-Hall, New York (1988).Google Scholar