Clustered sparsity and Poisson-gap sampling.

Non-uniform sampling (NUS) is a popular way of reducing the amount of time taken by multidimensional NMR experiments. Among the various non-uniform sampling schemes that exist, the Poisson-gap (PG) schedules are particularly popular, especially when combined with compressed-sensing (CS) reconstruction of missing data points. However, the use of PG is based mainly on practical experience and has not, as yet, been explained in terms of CS theory. Moreover, an apparent contradiction exists between the reported effectiveness of PG and CS theory, which states that a "flat" pseudo-random generator is the best way to generate sampling schedules in order to reconstruct sparse spectra. In this paper we explain how, and in what situations, PG reveals its superior features in NMR spectroscopy. We support our theoretical considerations with simulations and analyses of experimental data from the Biological Magnetic Resonance Bank (BMRB). Our analyses reveal a previously unnoticed feature of many NMR spectra that explains the success of "blue-noise" schedules, such as PG. We call this feature "clustered sparsity". This refers to the fact that the peaks in NMR spectra are not just sparse but often form clusters in the indirect dimension, and PG is particularly suited to deal with such situations. Additionally, we discuss why denser sampling in the initial and final parts of the clustered signal may be useful.

Temperature as an Extra Dimension in Multidimensional Protein NMR Spectroscopy.

NMR spectroscopy is a particularly informative method for studying protein structures and dynamics in solution; however, it is also one of the most time-consuming. Modern approaches to biomolecular NMR spectroscopy are based on lengthy multidimensional experiments, the duration of which grows exponentially with the number of dimensions. The experimental time may even be several days in the case of 3D and 4D spectra. Moreover, the experiment often has to be repeated under several different conditions, for example, to measure the temperature-dependent effects in a spectrum (temperature coefficients (TCs)). Herein, a new approach that involves joint sampling of indirect evolution times and temperature is proposed. This allows TCs to be measured through 3D spectra in even less time than that needed to acquire a single spectrum by using the conventional approach. Two signal processing methods that are complementary, in terms of sensitivity and resolution, 1) dividing data into overlapping subsets followed by compressed sensing reconstruction, and 2) treating the complete data set with a variant of the Radon transform, are proposed. The temperature-swept 3D HNCO spectra of two intrinsically disordered proteins, osteopontin and CD44 cytoplasmic tail, show that this new approach makes it possible to determine TCs and their non-linearities effectively. Non-linearities, which indicate the presence of a compact state, are particularly interesting. The complete package of data acquisition and processing software for this new approach are provided.

Enhancing Compression Level for More Efficient Compressed Sensing and Other Lessons from NMR Spectroscopy.

Modern nuclear magnetic resonance spectroscopy (NMR) is based on two- and higher-dimensional experiments that allow the solving of molecular structures, i.e., determine the relative positions of single atoms very precisely. However, rich chemical information comes at the price of long data acquisition times (up to several days). This problem can be alleviated by compressed sensing (CS)-a method that revolutionized many fields of technology. It is known that CS performs the most efficiently when measured objects feature a high level of compressibility, which in the case of NMR signal means that its frequency domain representation (spectrum) has a low number of significant points. However, many NMR spectroscopists are not aware of the fact that various well-known signal acquisition procedures enhance compressibility and thus should be used prior to CS reconstruction. In this study, we discuss such procedures and show to what extent they are complementary to CS approaches. We believe that the survey will be useful not only for NMR spectroscopists but also to inspire the broader signal processing community.

Nonstationary Two-Dimensional Nuclear Magnetic Resonance: A Method for Studying Reaction Mechanisms in Situ.

Nuclear magnetic resonance spectroscopy (NMR) is a versatile tool of chemical analysis allowing one to determine structures of molecules with atomic resolution. Particularly informative are two-dimensional (2D) experiments that directly identify atoms coupled by chemical bonds or a through-space interaction. Thus, NMR could potentially be powerful tool to study reactions in situ and explain their mechanisms. Unfortunately, 2D NMR is very time-consuming and thus often cannot serve as a "snapshot" technique for in situ reaction monitoring. Particularly difficult is the case of spectra, in which resonance frequencies vary in the course of reaction. This leads to resolution and sensitivity loss, often hindering the detection of transient products. In this paper we introduce a novel approach to correct such nonstationary 2D NMR signals and raise the detection limits over 10 times. We demonstrate success of its application for studying the mechanism of the reaction of AgSO4-induced synthesis of diphenylmethane-type compounds. Several reactions occur in the studied mixture of benzene and toluene, all with rather low yield and leading to compounds with similar chemical shifts. Nevertheless, with the use of a proposed 2D NMR approach we were able to describe complex mechanisms of diphenylmethane formation involving AgSO4-induced toluene deprotonation and formation of benzyl carbocation, followed by nucleophilic attacks.

Quick temperature-sweep pure-shift NMR: the case of solvent effects in atorvastatin.

Pure-shift NMR experiments provide highly resolved spectra, which could be perfect for precise monitoring of chemical shift variations under different conditions, such as temperature or concentration. However, their sensitivity is relatively low and signal sampling is time-consuming, which leads to long experimental times, making such serial acquisition problematic. In this paper we present a new method of NMR spectroscopy which improves the speed and sensitivity of serial pseudo-two-dimensional pure-shift experiments. The example of variable-temperature study of atorvastatin reveals the potential of the method in verifying the theoretical predictions of solvent-dependent spectral effects.

Pitfalls in compressed sensing reconstruction and how to avoid them.

Multidimensional NMR can provide unmatched spectral resolution, which is crucial when dealing with samples of biological macromolecules. The resolution, however, comes at the high price of long experimental time. Non-uniform sampling (NUS) of the evolution time domain allows to suppress this limitation by sampling only a small fraction of the data, but requires sophisticated algorithms to reconstruct omitted data points. A significant group of such algorithms known as compressed sensing (CS) is based on the assumption of sparsity of a reconstructed spectrum. Several papers on the application of CS in multidimensional NMR have been published in the last years, and the developed methods have been implemented in most spectral processing software. However, the publications rarely show the cases when NUS reconstruction does not work perfectly or explain how to solve the problem. On the other hand, every-day users of NUS develop their rules-of-thumb, which help to set up the processing in an optimal way, but often without a deeper insight. In this paper, we discuss several sources of problems faced in CS reconstructions: low sampling level, missassumption of spectral sparsity, wrong stopping criterion and attempts to extrapolate the signal too much. As an appendix, we provide MATLAB codes of several CS algorithms used in NMR. We hope that this work will explain the mechanism of NUS reconstructions and help readers to set up acquisition and processing parameters. Also, we believe that it might be helpful for algorithm developers.

Modified OMP algorithm for exponentially decaying signals.

A group of signal reconstruction methods, referred to as compressed sensing (CS), has recently found a variety of applications in numerous branches of science and technology. However, the condition of the applicability of standard CS algorithms (e.g., orthogonal matching pursuit, OMP), i.e., the existence of the strictly sparse representation of a signal, is rarely met. Thus, dedicated algorithms for solving particular problems have to be developed. In this paper, we introduce a modification of OMP motivated by nuclear magnetic resonance (NMR) application of CS. The algorithm is based on the fact that the NMR spectrum consists of Lorentzian peaks and matches a single Lorentzian peak in each of its iterations. Thus, we propose the name Lorentzian peak matching pursuit (LPMP). We also consider certain modification of the algorithm by introducing the allowed positions of the Lorentzian peaks' centers. Our results show that the LPMP algorithm outperforms other CS algorithms when applied to exponentially decaying signals.

Optimizing measurements of linear changes of NMR signal parameters.

Serial NMR experiments are commonly applied in variable-temperature studies, reaction monitoring, and other tasks. The resonance frequencies often shift linearly over the series, and the shift rates help to characterize the studied system. They can be determined using a classical fitting of peak positions or a more advanced method of Radon transform. However, the optimal procedure for data collection remains to be determined. In this paper, we discuss how to invest experimental time, i.e., whether to measure more scans at the expense of the number of spectra or vice versa. The results indicate that classical fitting provides slightly less error than the Radon transform, although the latter can be the method of choice for a low signal-to-noise ratio. We demonstrate this fact through theoretical consideration, simulations, and an experiment. Finally, we extend our considerations to the linear fitting of peak amplitudes. Interestingly, the optimal setup for measuring peak height changes differs from the one for resonance frequency changes - fewer spectra with more scans provide better results.

Fast time-resolved NMR with non-uniform sampling.

NMR spectroscopy is a versatile tool for studying time-dependent processes: chemical reactions, phase transitions or macromolecular structure changes. However, time-resolved NMR is usually based on the simplest among available techniques - one-dimensional spectra serving as "snapshots" of the studied process. One of the reasons is that multidimensional experiments are very time-expensive due to costly sampling of evolution time space. In this review we summarize efforts to alleviate the problem of limited applicability of multidimensional NMR in time-resolved studies. We focus on techniques based on sparse or non-uniform sampling (NUS), which lead to experimental time reduction by omitting a significant part of the data during measurement and reconstructing it mathematically, adopting certain assumptions about the spectrum. NUS spectra are faster to acquire than conventional ones and thus better suited to the role of "snapshots", but still suffer from non-stationarity of the signal i.e. amplitude and frequency variations within a dataset. We discuss in detail how these instabilities affect the spectra, and what are the optimal ways of sampling the non-stationary FID signal. Finally, we discuss related areas of NMR where serial experiments are exploited and how they can benefit from the same NUS-based approaches.

Resolution enhancement in NMR spectra by deconvolution with compressed sensing reconstruction.

NMR spectroscopy is one of the basic tools for molecular structure elucidation. Unfortunately, the resolution of the spectra is often limited by inter-nuclear couplings. The existing workarounds often alleviate the problem by trading it for another deficiency, such as spectral artefacts or difficult sample preparation and, thus, are rarely used. We suggest an approach using the coupling deconvolution in the framework of compressed sensing (CS) spectra processing that leads to a major increase in resolution, sensitivity, and overall quality of NUS reconstruction. A new mathematical description of the decoupling by deconvolution explains the effects of thermal noise and reveals a relation with the underlying assumption of the CS. The gain in resolution and sensitivity for challenging molecular systems is demonstrated for the key HNCA experiment used for protein backbone assignment applied to two large proteins: intrinsically disordered 441-residue Tau and a 509-residue globular bacteriophytochrome fragment. The approach will be valuable in a multitude of chemistry applications, where NMR experiments are compromised by the homonuclear scalar coupling.

Accelerated acquisition in pure-shift spectra based on prior knowledge from 1H NMR.

Pure-shift NMR enhances spectral resolution, but the optimal resolution can only be obtained at the cost of acquisition time. We propose to accelerate pure-shift acquisition using optimised 'burst' non-uniform sampling schemes [I. E. Ndukwe, A. Shchukina, K. Kazimierczuk and C. P. Butts, Chem. Commun., 2016, 52, 12769] and then reconstruct the undersampled signal mathematically. Here, we focus on the reliability of this reconstruction depending on the sampling scheme and present a workflow for the sampling optimization. It is ready to be implemented in routine measurements and yields a great improvement in reconstruction of challenging cases.

Quick, sensitive serial NMR experiments with Radon transform.

The Radon transform is a potentially powerful tool for processing the data from serial spectroscopic experiments. It makes it possible to decode the rate at which frequencies of spectral peaks shift under the effect of changing conditions, such as temperature, pH, or solvent. In this paper we show how it also improves speed and sensitivity, especially in multidimensional experiments. This is particularly important in the case of low-sensitivity techniques, such as NMR spectroscopy. As an example, we demonstrate how Radon transform processing allows serial measurements of 15N-HSQC spectra of unlabelled peptides that would otherwise be infeasible.

Artifacts in time-resolved NUS: A case study of NOE build-up curves from 2D NOESY.

Multidimensional NMR spectroscopy requires time-consuming sampling of indirect dimensions and so is usually used to study stable samples. However, dynamically changing compounds or their mixtures commonly occur in problems of natural science. Monitoring them requires the use multidimensional NMR in a time-resolved manner - in other words, a series of quick spectra must be acquired at different points in time. Among the many solutions that have been proposed to achieve this goal, time-resolved non-uniform sampling (TR-NUS) is one of the simplest. In a TR-NUS experiment, the signal is sampled using a shuffled random schedule and then divided into overlapping subsets. These subsets are then processed using one of the NUS reconstruction methods, for example compressed sensing (CS). The resulting stack of spectra forms a temporal "pseudo-dimension" that shows the changes caused by the process occurring in the sample. CS enables the use of small subsets of data, which minimizes the averaging of the effects studied. Yet, even within these limited timeframes, the sample undergoes certain changes. In this paper we discuss the effect of varying signal amplitude in a TR-NUS experiment. Our theoretical calculations show that the variations within the subsets lead to t1-noise, which is dependent on the rate of change of the signal amplitude. We verify these predictions experimentally. As a model case we choose a novel 2D TR-NOESY experiment in which mixing time is varied in parallel with shuffled NUS in the indirect dimension. The experiment, performed on a sample of strychnine, provides a near-continuous NOE build-up curve, whose shape closely reflects the t1-noise level. 2D TR-NOESY reduces the measurement time compared to the conventional approach and makes it possible to verify the theoretical predictions about signal variations during TR-NUS.