SAND: Automated Time-Domain Modeling of NMR Spectra Applied to Metabolite Quantification.

Developments in untargeted nuclear magnetic resonance (NMR) metabolomics enable the profiling of thousands of biological samples. The exploitation of this rich source of information requires a detailed quantification of spectral features. However, the development of a consistent and automatic workflow has been challenging because of extensive signal overlap. To address this challenge, we introduce the software Spectral Automated NMR Decomposition (SAND). SAND follows on from the previous success of time-domain modeling and automatically quantifies entire spectra without manual interaction. The SAND approach uses hybrid optimization with Markov chain Monte Carlo methods, employing subsampling in both time and frequency domains. In particular, SAND randomly divides the time-domain data into training and validation sets to help avoid overfitting. We demonstrate the accuracy of SAND, which provides a correlation of ∼0.9 with ground truth on cases including highly overlapped simulated data sets, a two-compound mixture, and a urine sample spiked with different amounts of a four-compound mixture. We further demonstrate an automated annotation using correlation networks derived from SAND decomposed peaks, and on average, 74% of peaks for each compound can be recovered in single clusters. SAND is available in NMRbox, the cloud computing environment for NMR software hosted by the Network for Advanced NMR (NAN). Since the SAND method uses time-domain subsampling (i.e., random subset of time-domain points), it has the potential to be extended to a higher dimensionality and nonuniformly sampled data.

A simple approach for reconstruction of non-uniformly sampled pseudo-3D NMR data for accurate measurement of spin relaxation parameters.

We explain how to conduct a pseudo-3D relaxation series NUS measurement so that it can be reconstructed by existing 3D NUS reconstruction methods to give accurate relaxation values. We demonstrate using reconstruction algorithms IST and SMILE that this 3D approach allows lower sampling densities than for independent 2D reconstructions. This is in keeping with the common finding that higher dimensionality increases signal sparsity, enabling lower sampling density. The approach treats the relaxation series as ordinary 3D time-domain data whose imaginary part in the pseudo-dimension is zero, and applies any suitably linear 3D NUS reconstruction method accordingly. Best results on measured and simulated data were achieved using acquisitions with 9 to 12 planes and exponential spacing in the pseudo-dimension out to ~ 2 times the inverse decay time. Given these criteria, in typical cases where 2D reconstructions require 50% sampling, the new 3D approach generates spectra reliably at sampling densities of 25%.

Reducing the measurement time of exact NOEs by non-uniform sampling.

We have previously reported on the measurement of exact NOEs (eNOEs), which yield a wealth of additional information in comparison to conventional NOEs. We have used these eNOEs in a variety of applications, including calculating high-resolution structures of proteins and RNA molecules. The collection of eNOEs is challenging, however, due to the need to measure a NOESY buildup series consisting of typically four NOESY spectra with varying mixing times in a single measurement session. While the 2D version can be completed in a few days, a fully sampled 3D-NOESY buildup series can take 10 days or more to acquire. This can be both expensive as well as problematic in the case of samples that are not stable over such a long period of time. One potential method to significantly decrease the required measurement time of eNOEs is to use non-uniform sampling (NUS) to decrease the number of points measured in the indirect dimensions. The effect of NUS on the extremely tight distance restraints extracted from eNOEs may be very pronounced. Therefore, we investigated the fidelity of eNOEs measured from three test cases at decreasing NUS densities: the 18.4 kDa protein human Pin1, the 4.1 kDa WW domain of Pin1 (both in 3D), and a 4.6 kDa 14mer RNA UUCG tetraloop (2D). Our results show that NUS imparted negligible error on the eNOE distances derived from good quality data down to 10% sampling for all three cases, but there is a noticeable decrease in the eNOE yield that is dependent upon the underlying sparsity, and thus complexity, of the sample. For Pin1, this transition occurred at roughly 40% while for the WW domain and the UUCG tetraloop it occurred at lower NUS densities of 20% and 10%, respectively. We rationalized these numbers through reconstruction simulations under various conditions. The extent of this loss depends upon the number of scans taken as well as the number of peaks to be reconstructed. Based on these findings, we have created guidelines for choosing an optimal NUS density depending on the number of peaks needed to be reconstructed in the densest region of a 2D or 3D NOESY spectrum.

Principal component analysis for automated classification of 2D spectra and interferograms of protein therapeutics: influence of noise, reconstruction details, and data preparation.

Protein therapeutics have numerous critical quality attributes (CQA) that must be evaluated to ensure safety and efficacy, including the requirement to adopt and retain the correct three-dimensional fold without forming unintended aggregates. Therefore, the ability to monitor protein higher order structure (HOS) can be valuable throughout the lifecycle of a protein therapeutic, from development to manufacture. 2D NMR has been introduced as a robust and precise tool to assess the HOS of a protein biotherapeutic. A common use case is to decide whether two groups of spectra are substantially different, as an indicator of difference in HOS. We demonstrate a quantitative use of principal component analysis (PCA) scores to perform this decision-making, and demonstrate the effect of acquisition and processing details on class separation using samples of NISTmAb monoclonal antibody Reference Material subjected to two different oxidative stress protocols. The work introduces an approach to computing similarity from PCA scores based upon the technique of histogram intersection, a method originally developed for retrieval of images from large databases. Results show that class separation can be robust with respect to random noise, reconstruction method, and analysis region selection. By contrast, details such as baseline distortion can have a pronounced effect, and so must be controlled carefully. Since the classification approach can be performed without the need to identify peaks, results suggest that it is possible to use even more efficient measurement strategies that do not produce spectra that can be analyzed visually, but nevertheless allow useful decision-making that is objective and automated.

Comparative Analysis of One-Dimensional Protein Fingerprint by Line Shape Enhancement and Two-Dimensional 1H,13C Methyl NMR Methods for Characterization of the Higher Order Structure of IgG1 Monoclonal Antibodies.

The use of NMR spectroscopy has emerged as a premier tool to characterize the higher order structure of protein therapeutics and in particular IgG1 monoclonal antibodies (mAbs). Due to their large size, traditional 1H-15N correlation experiments have proven exceedingly difficult to implement on mAbs, and a number of alternative techniques have been proposed, including the one-dimensional (1D) 1H protein fingerprint by line shape enhancement (PROFILE) method and the two-dimensional (2D) 1H-13C methyl correlation-based approach. Both 1D and 2D approaches have relative strengths and weaknesses, related to the inherent sensitivity and resolution of the respective methods. To further increase the utility of NMR to the biopharmaceutical community, harmonized criteria for decision making in employing 1D and 2D approaches for mAb characterization are warranted. To this end, we have conducted an interlaboratory comparative study of the 1D PROFILE and 2D methyl methods on several mAbs samples to determine the degree to which each method is suited to detect spectral difference between the samples and the degree to which results from each correlate with one another. Results from the study demonstrate both methods provide statistical data highly comparable to one another and that each method is capable of complementing the limitations commonly associated with the other, thus providing a better overall picture of higher order structure.

Best Practices in Utilization of 2D-NMR Spectral Data as the Input for Chemometric Analysis in Biopharmaceutical Applications.

Quality attributes (QAs) are measureable parameters of a biologic that impact product safety and efficacy and are essential characteristics that are linked to positive patient health outcomes. One QA, higher order structure (HOS), is directly coupled to the function of protein biologics, and deviations in this QA may cause adverse effects. To address the critical need for HOS assessment, methods for analyzing structural fingerprints from 2D nuclear magnetic resonance spectroscopy (2D-NMR) spectra have been established for drug substances as large as monoclonal antibody therapeutics. Here, chemometric analyses have been applied to 2D 1H,13C-methyl NMR correlation spectra of the IgG1κ NIST monoclonal antibody (NISTmAb), recorded at natural isotopic abundance, to benchmark the performance and robustness of the methods. In particular, a variety of possible spectral input schemes (e.g., chemical shift, peak intensity, and total spectral matrix) into chemometric algorithms are examined using two case studies: (1) a large global 2D-NMR interlaboratory study and (2) a blended series of enzymatically glycan-remodeled NISTmAb isoforms. These case studies demonstrate that the performance of chemometric algorithms using either peak positions or total spectral matrix as the input will depend on the study design and likely be product-specific. In general, peak positions are found to be a more robust spectral parameter for input into chemometric algorithms, whereas the total spectral matrix approach lends itself to easier automation and requires less user intervention. Analysis with different input data also shows differences in sensitivity to certain changes in HOS, highlighting that product knowledge will further guide appropriate method selection based on the fit-for-purpose application in the context of biopharmaceutical development, production, and quality control.

Chemometric Outlier Classification of 2D-NMR Spectra to Enable Higher Order Structure Characterization of Protein Therapeutics.

Protein therapeutics are vitally important clinically and commercially, with monoclonal antibody (mAb) therapeutic sales alone accounting for $115 billion in revenue for 2018.[1] In order for these therapeutics to be safe and efficacious, their protein components must maintain their high order structure (HOS), which includes retaining their three-dimensional fold and not forming aggregates. As demonstrated in the recent NISTmAb Interlaboratory nuclear magnetic resonance (NMR) Study[2], NMR spectroscopy is a robust and precise approach to address this HOS measurement need. Using the NISTmAb study data, we benchmark a procedure for automated outlier detection used to identify spectra that are not of sufficient quality for further automated analysis. When applied to a diverse collection of all 252 1H,13C gHSQC spectra from the study, a recursive version of the automated procedure performed comparably to visual analysis, and identified three outlier cases that were missed by the human analyst. In total, this method represents a distinct advance in chemometric detection of outliers due to variation in both measurement and sample.

Nonuniform sampling in multidimensional NMR for improving spectral sensitivity.

The development of multidimensional NMR spectroscopy enabled an explosion of structural and dynamical investigations on proteins and other biomacromolecules. Practical limitations on data sampling, based on the Jeener paradigm of parametric sampling of indirect time domains, have long placed limits on resolution in the corresponding frequency dimensions. The emergence of nonuniform sampling (NUS) in indirect time dimensions circumvents those limitations, affording high resolution spectra from short data records collected in practically realized measurement times. In addition to substantially improved resolution, NUS can also be exploited to improve sensitivity, with gains comparable to those obtained using cryogenically cooled probes. We describe a general approach for acquiring and processing multidimensional NUS NMR data for improving sensitivity.

Selective suppression of excipient signals in 2D 1H-13C methyl spectra of biopharmaceutical products.

While the use of 1H-13C methyl correlated NMR spectroscopy at natural isotopic abundance has been demonstrated as feasible on protein therapeutics as large as monoclonal antibodies, spectral interference from aliphatic excipients remains a significant obstacle to its widespread application. These signals can cause large baseline artifacts, obscure protein resonances, and cause dynamic range suppression of weak peaks in non-uniform sampling applications, thus hampering both traditional peak-based spectral analyses as well as emerging chemometric methods of analysis. Here we detail modifications to the 2D 1H-13C gradient-selected HSQC experiment that make use of selective pulsing techniques for targeted removal of interfering excipient signals in spectra of the NISTmAb prepared in several different formulations. This approach is demonstrated to selectively reduce interfering excipient signals while still yielding 2D spectra with only modest losses in protein signal. Furthermore, it is shown that spectral modeling based on the SMILE algorithm can be used to simulate and subtract any residual excipient signals and their attendant artifacts from the resulting 2D NMR spectra.

NMRbox: A Resource for Biomolecular NMR Computation.

Advances in computation have been enabling many recent advances in biomolecular applications of NMR. Due to the wide diversity of applications of NMR, the number and variety of software packages for processing and analyzing NMR data is quite large, with labs relying on dozens, if not hundreds of software packages. Discovery, acquisition, installation, and maintenance of all these packages is a burdensome task. Because the majority of software packages originate in academic labs, persistence of the software is compromised when developers graduate, funding ceases, or investigators turn to other projects. To simplify access to and use of biomolecular NMR software, foster persistence, and enhance reproducibility of computational workflows, we have developed NMRbox, a shared resource for NMR software and computation. NMRbox employs virtualization to provide a comprehensive software environment preconfigured with hundreds of software packages, available as a downloadable virtual machine or as a Platform-as-a-Service supported by a dedicated compute cloud. Ongoing development includes a metadata harvester to regularize, annotate, and preserve workflows and facilitate and enhance data depositions to BioMagResBank, and tools for Bayesian inference to enhance the robustness and extensibility of computational analyses. In addition to facilitating use and preservation of the rich and dynamic software environment for biomolecular NMR, NMRbox fosters the development and deployment of a new class of metasoftware packages. NMRbox is freely available to not-for-profit users.

Sparse multidimensional iterative lineshape-enhanced (SMILE) reconstruction of both non-uniformly sampled and conventional NMR data.

Implementation of a new algorithm, SMILE, is described for reconstruction of non-uniformly sampled two-, three- and four-dimensional NMR data, which takes advantage of the known phases of the NMR spectrum and the exponential decay of underlying time domain signals. The method is very robust with respect to the chosen sampling protocol and, in its default mode, also extends the truncated time domain signals by a modest amount of non-sampled zeros. SMILE can likewise be used to extend conventional uniformly sampled data, as an effective multidimensional alternative to linear prediction. The program is provided as a plug-in to the widely used NMRPipe software suite, and can be used with default parameters for mainstream application, or with user control over the iterative process to possibly further improve reconstruction quality and to lower the demand on computational resources. For large data sets, the method is robust and demonstrated for sparsities down to ca 1%, and final all-real spectral sizes as large as 300 Gb. Comparison between fully sampled, conventionally processed spectra and randomly selected NUS subsets of this data shows that the reconstruction quality approaches the theoretical limit in terms of peak position fidelity and intensity. SMILE essentially removes the noise-like appearance associated with the point-spread function of signals that are a default of five-fold above the noise level, but impacts the actual thermal noise in the NMR spectra only minimally. Therefore, the appearance and interpretation of SMILE-reconstructed spectra is very similar to that of fully sampled spectra generated by Fourier transformation.

Multivariate Analysis of Two-Dimensional 1H, 13C Methyl NMR Spectra of Monoclonal Antibody Therapeutics To Facilitate Assessment of Higher Order Structure.

Two-dimensional (2D) 1H-13C methyl NMR provides a powerful tool to probe the higher order structure (HOS) of monoclonal antibodies (mAbs), since spectra can readily be acquired on intact mAbs at natural isotopic abundance, and small changes in chemical environment and structure give rise to observable changes in corresponding spectra, which can be interpreted at atomic resolution. This makes it possible to apply 2D NMR spectral fingerprinting approaches directly to drug products in order to systematically characterize structure and excipient effects. Systematic collections of NMR spectra are often analyzed in terms of the changes in specifically identified peak positions, as well as changes in peak height and line widths. A complementary approach is to apply principal component analysis (PCA) directly to the matrix of spectral data, correlating spectra according to similarities and differences in their overall shapes, rather than according to parameters of individually identified peaks. This is particularly well-suited for spectra of mAbs, where some of the individual peaks might not be well resolved. Here we demonstrate the performance of the PCA method for discriminating structural variation among systematic sets of 2D NMR fingerprint spectra using the NISTmAb and illustrate how spectral variability identified by PCA may be correlated to structure.

Probing Invisible, Excited Protein States by Non-Uniformly Sampled Pseudo-4D CEST Spectroscopy.

Chemical exchange saturation transfer (CEST) NMR spectroscopy is a powerful tool for studies of slow timescale protein dynamics. Typical experiments are based on recording a large number of 2D data sets and quantifying peak intensities in each of the resulting planes. A weakness of the method is that peaks must be resolved in 2D spectra, limiting applications to relatively small proteins. Resolution is significantly improved in 3D spectra but recording uniformly sampled data is time-prohibitive. Here we describe non-uniformly sampled HNCO-based pseudo-4D CEST that provides excellent resolution in reasonable measurement times. Data analysis is done through fitting in the time domain, without the need of reconstructing the frequency dimensions, exploiting previously measured accurate peak positions in reference spectra. The methodology is demonstrated on several protein systems, including a nascent form of superoxide dismutase that is implicated in neurodegenerative disease.

Correlated analytical and functional evaluation of higher order structure perturbations from oxidation of NISTmAb.

The clinical efficacy and safety of protein-based drugs such as monoclonal antibodies (mAbs) rely on the integrity of the protein higher order structure (HOS) during product development, manufacturing, storage, and patient administration. As mAb-based drugs are becoming more prevalent in the treatment of many illnesses, the need to establish metrics for quality attributes of mAb therapeutics through high-resolution techniques is also becoming evident. To this end, here we used a forced degradation method, time-dependent oxidation by hydrogen peroxide, on the model biotherapeutic NISTmAb and evaluated the effects on HOS with orthogonal analytical methods and a functional assay. To monitor the oxidation process, the experimental workflow involved incubation of NISTmAb with hydrogen peroxide in a benchtop nuclear magnetic resonance spectrometer (NMR) that followed the reaction kinetics, in real-time through the water proton transverse relaxation rate R2(1H2O). Aliquots taken at defined time points were further analyzed by high-field 2D 1H-13C methyl correlation fingerprint spectra in parallel with other analytical techniques, including thermal unfolding, size-exclusion chromatography, and surface plasmon resonance, to assess changes in stability, heterogeneity, and binding affinities. The complementary measurement outputs from the different techniques demonstrate the utility of combining NMR with other analytical tools to monitor oxidation kinetics and extract the resulting structural changes in mAbs that are functionally relevant, allowing rigorous assessment of HOS attributes relevant to the efficacy and safety of mAb-based drug products.

Interlaboratory Studies Using the NISTmAb to Advance Biopharmaceutical Structural Analytics.

Biopharmaceuticals such as monoclonal antibodies are required to be rigorously characterized using a wide range of analytical methods. Various material properties must be characterized and well controlled to assure that clinically relevant features and critical quality attributes are maintained. A thorough understanding of analytical method performance metrics, particularly emerging methods designed to address measurement gaps, is required to assure methods are appropriate for their intended use in assuring drug safety, stability, and functional activity. To this end, a series of interlaboratory studies have been conducted using NISTmAb, a biopharmaceutical-representative and publicly available monoclonal antibody test material, to report on state-of-the-art method performance, harmonize best practices, and inform on potential gaps in the analytical measurement infrastructure. Reported here is a summary of the study designs, results, and future perspectives revealed from these interlaboratory studies which focused on primary structure, post-translational modifications, and higher order structure measurements currently employed during biopharmaceutical development.

Consistent blind protein structure generation from NMR chemical shift data.

Protein NMR chemical shifts are highly sensitive to local structure. A robust protocol is described that exploits this relation for de novo protein structure generation, using as input experimental parameters the (13)C(alpha), (13)C(beta), (13)C', (15)N, (1)H(alpha) and (1)H(N) NMR chemical shifts. These shifts are generally available at the early stage of the traditional NMR structure determination process, before the collection and analysis of structural restraints. The chemical shift based structure determination protocol uses an empirically optimized procedure to select protein fragments from the Protein Data Bank, in conjunction with the standard ROSETTA Monte Carlo assembly and relaxation methods. Evaluation of 16 proteins, varying in size from 56 to 129 residues, yielded full-atom models that have 0.7-1.8 A root mean square deviations for the backbone atoms relative to the experimentally determined x-ray or NMR structures. The strategy also has been successfully applied in a blind manner to nine protein targets with molecular masses up to 15.4 kDa, whose conventional NMR structure determination was conducted in parallel by the Northeast Structural Genomics Consortium. This protocol potentially provides a new direction for high-throughput NMR structure determination.

Principal Component Analysis of 1D 1H Diffusion Edited NMR Spectra of Protein Therapeutics.

The one-dimensional (1D) diffusion edited proton NMR method, Protein Fingerprint by Lineshape Enhancement (PROFILE) has been demonstrated to be suitable for higher order structure (HOS) characterization of protein therapeutics including monoclonal antibodies. Recent reports in the literature have demonstrated its advantages for HOS characterization over traditional methods such as circular dichroism and Fourier-transform infrared spectroscopy. Previously, we have demonstrated that the PROFILE method is complementary with high resolution 2D methyl correlated NMR methods and how both may be deployed as a multi-modal platform to further the utility of NMR for HOS characterization. A major limitation of the PROFILE method remains its need for high signal to noise data due to its reliance on convolution difference processing and linear correlation metrics to assess spectral similarity. Here we present an alternative method for analyzing 1D diffusion edited spectra, which overcomes this limitation by using nonlinear iterative partial least squares (NIPALS) principal component analysis, and which we dub PROtein Fingerprint Observed Using NIPALS Decomposition (PROFOUND). We demonstrate that results from the PROFOUND method are robust with respect to instrument, operator and in the presence of high experimental noise and how it may be employed to provide quantitative assessment of spectral similarity.

NUScon: a community-driven platform for quantitative evaluation of nonuniform sampling in NMR.

Although the concepts of nonuniform sampling (NUS​​​​​​​) and non-Fourier spectral reconstruction in multidimensional NMR began to emerge 4 decades ago , it is only relatively recently that NUS has become more commonplace. Advantages of NUS include the ability to tailor experiments to reduce data collection time and to improve spectral quality, whether through detection of closely spaced peaks (i.e., "resolution") or peaks of weak intensity (i.e., "sensitivity"). Wider adoption of these methods is the result of improvements in computational performance, a growing abundance and flexibility of software, support from NMR spectrometer vendors, and the increased data sampling demands imposed by higher magnetic fields. However, the identification of best practices still remains a significant and unmet challenge. Unlike the discrete Fourier transform, non-Fourier methods used to reconstruct spectra from NUS data are nonlinear, depend on the complexity and nature of the signals, and lack quantitative or formal theory describing their performance. Seemingly subtle algorithmic differences may lead to significant variabilities in spectral qualities and artifacts. A community-based critical assessment of NUS challenge problems has been initiated, called the "Nonuniform Sampling Contest" (NUScon), with the objective of determining best practices for processing and analyzing NUS experiments. We address this objective by constructing challenges from NMR experiments that we inject with synthetic signals, and we process these challenges using workflows submitted by the community. In the initial rounds of NUScon our aim is to establish objective criteria for evaluating the quality of spectral reconstructions. We present here a software package for performing the quantitative analyses, and we present the results from the first two rounds of NUScon. We discuss the challenges that remain and present a roadmap for continued community-driven development with the ultimate aim of providing best practices in this rapidly evolving field. The NUScon software package and all data from evaluating the challenge problems are hosted on the NMRbox platform.

Assessment of the Higher-Order Structure of Formulated Monoclonal Antibody Therapeutics by 2D Methyl Correlated NMR and Principal Component Analysis.

Characterization of the higher-order structure (HOS) of protein therapeutics, and in particular of monoclonal antibodies, by 2D 1 H-13 C methyl correlated NMR has been demonstrated as precise and robust. Such characterization can be greatly enhanced when collections of spectra are analyzed using multivariate approaches such as principal component analysis (PCA), allowing for the detection and identification of small structural differences in drug substance that may otherwise fall below the limit of detection of conventional spectral analysis. A major limitation to this approach is the presence of aliphatic signals from formulation or excipient components, which result in spectral interference with the protein signal of interest; however, the recently described Selective Excipient Reduction and Removal (SIERRA) filter greatly reduces this issue. Here we will outline how basic 2D 1 H-13 C methyl-correlated NMR may be combined with the SIERRA approach to collect 'clean' NMR spectra of formulated monoclonal antibody therapeutics (i.e., drug substance spectra free of interfering component signals), and how series of such spectra may be used for HOS characterization by direct PCA of the series spectral matrix. © 2020 U.S. Government. Basic Protocol 1: NMR data acquisition Basic Protocol 2: Full spectral matrix data processing and analysis Support Protocol: Data visualization and cluster analysis.