Development of new approaches to NMR data collection for protein structure determination

by Coggins, Brian E.

Abstract (Summary)
Multidimensional nuclear magnetic resonance (NMR) spectroscopy has become

one of the most important techniques available for studying the structure and function of

biological macromolecules at atomic resolution. The conventional approach to

multidimensional NMR involves the sampling of the time domain on a Cartesian grid

followed by a multidimensional Fourier transform (FT). While this approach yields high

quality spectra, as the number of dimensions is increased the time needed for sampling on

a Cartesian grid increases exponentially, making it impractical to record 4-D spectra at

high resolution and impossible to record 5-D spectra at all.

This thesis describes new approaches to data collection and processing that make

it possible to obtain spectra at higher resolution and/or with a higher dimensionality than

was previously feasible with the conventional method. The central focus of this work has

been the sampling of the time domain along radial spokes, which was recently introduced

into the NMR community. If each radial spoke is processed by an FT with respect to

radius, a set of projections of the higher-dimensional spectrum are obtained. Full spectra

at high resolution can be generated from these projections via tomographic

reconstruction. We generalized the lower-value reconstruction algorithm from the

literature, and later integrated it with the backprojection algorithm in a hybrid

reconstruction method. These methods permit the reconstruction of accurate 4-D and 5-

D spectra at very high resolution, from only a small number of projections, as we

demonstrated in the reconstruction of 4-D and 5-D sequential assignment spectra on

small and large proteins. For nuclear Overhauser spectroscopy (NOESY), used to

measure interproton distances in proteins, one requires quantitative reconstructions. We

have successfully obtained these using filtered backprojection, which we found was

equivalent to processing the radially sampled data by a polar FT. All of these methods

represent significant gains in data collection efficiency over conventional approaches.

The polar FT interpretation suggested that the problem could be analyzed using

FT theory, to design even more efficient methods. We have developed a new approach to

sampling, using concentric rings of sampling points, which represents a further

improvement in efficiency and sensitivity over radial sampling.

Bibliographical Information:

Advisor:Zhou, Pei; Lee, Andrew L.; Raetz, Christian R.H.; Richardson, Jane S.; Spicer, Leonard D.

School:Duke University

School Location:USA - North Carolina

Source Type:Master's Thesis

Keywords:nuclear magnetic resonance


Date of Publication:05/10/2007

© 2009 All Rights Reserved.