PAW - the Protein Analysis Workshop for 2D nuclear magnetic resonance spectroscopy : a thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Physics at Massey University, New Zealand

Abstract

An X Window-based software package for SGI workstations has been developed to process and assign NMR spectra. Special consideration has been given to the assignment of two-dimensional 1H NMR spectra of proteins. The program combines features from the packages PROSPA [Eccles 1995], EASY [Eccles 1991] and FELIX [Biosym 1995] as well as having its own capabilities. It allows simultaneous display of multiple toolboxes and spectra, which can be flexibly manipulated by mouse operations, command entries, and user-editable macros. NMR spectra can be processed either interactively or with macros containing commands with parameters. A unique filter that combines the exponential and sine-bell functions has been frequently used. A water suppression technique based on fitting averaged time-domain data, as well as an efficient algorithm for calculating fast Fourier transform and Hilbert transform [Eccles 1995] are discussed and implemented. NMR spectral assignment is done interactively in three steps: peak picking, spin-system identification, and sequence-specific assignment. The process utilises three peak lists: a raw-peak list that contains records of all possible peaks in a NOESY spectrum, a diagonal peak list that contains records of peaks that define a curve about which the spectrum is symmetric, and a cross-peak list that contains records of peaks that are assigned. Details of the peak-picking methods are discussed. By reference to a list of diagonal peaks, a common calibration problem caused by Bloch-Siegert shifts [Bloch and Siegert 1940, Ernst 1987] has been minimised. Automatically produced NOE summaries allow a quick identification of peaks that are unassigned or incorrectly assigned. The peak position and integration parameters can be calculated through non-linear curve fitting with Gaussians. NMR data processing and spectral assignment using the package has been completed for Caerin 4.1, a 23-residue protein. Linear-prediction has been applied to increase the spectral resolution. Detailed results for this protein are presented. The NOE summary of the sequential assignments indicates a well-defined secondary structure that is different from Caerin 1.1 [Wong 1996, 1997].