Computational Approaches for Assessing Spectral Quality in NMR Spectroscopy

Nuclear magnetic resonance spectroscopy is a powerful biophysical technique for characterizing biological macromolecules including determination of three-dimensional structure, dynamics, and ligand interactions. The advent of multidimensional NMR spectroscopy facilitated a surge of structural and dynamical investigations of biological macromolecules by yielding unmatched gains in resolution. Indeed, the biological applications of NMR depend on acquiring the best possible spectra with desirable features such as high signal-to-noise ratio and high resolution. Because such spectra must be obtainable in reasonable time frames, practical limitations, in particular prohibitive multidimensional experiment times, have restricted implementation of NMR spectroscopy for certain biological problems. To address this problem, numerous data acquisition and signal processing strategies have been developed. To reduce the burden of experiment duration, nonuniform sampling can be used for collection of higher dimensionality experiments in shorter time frames and also permits acquisition of longer evolution times along indirect dimensions to achieve higher resolution spectra. However, data acquired according to nonuniform sampling strategies is not amenable to conventional data processing techniques, namely the Fourier Transform and therefore non-Fourier methods are increasingly relied upon due to their ability to handle such data. The relative prowess of these novel techniques and data processing algorithms have yet to be compared in a systematic fashion. Part of the difficulty is that non-Fourier methods present unique challenges due to their nonlinearity, which can produce nonrandom noise and render conventional metrics for spectral quality such as signal-to-noise ratio unreliable. The in situ receiver operating characteristic analysis (IROC) is a workflow for making comparisons between NMR data acquisition strategies and processing algorithms that circumvents the traditional difficulties of spectral comparison. IROC analysis is based on the Receiver Operating Characteristic curve and utilizes synthetic signals added to empirical data and yields several robust quantitative metrics for spectral quality. In this work, the theoretical development, underlying algorithm, and practical potential of IROC analysis are first presented to show its ability to make quantitative comparisons of spectral quality in situations were other metrics fail. The IROC method is subsequently applied to experimental data to quantify the sensitivity and resolution that can be achieved through various nonuniform sampling schemes that each have different properties