Fast reconstruction of non-uniform sampling multidimensional NMR spectroscopy via a deep neural network

Restore High-Resolution Nuclear Magnetic Resonance Spectra from Inhomogeneous Magnetic Fields Using a Neural Network

High-resolution nuclear magnetic resonance (NMR) spectroscopy is a powerful analytical tool with wide applications. However, the conventional shim technique may not guarantee the homogeneity of the magnetic field when the experimental conditions are unfavorable. In this study, we proposed a data postprocessing method called Restore High-resolution Unet (RH-Unet), which uses a convolutional neural network to restore distorted NMR spectra that have been acquired in inhomogeneous magnetic fields. The method generates feature-label pairs from singlet peak regions and ideal Lorentzian line shapes and trains a RH-Unet model to map low-resolution spectra to high-resolution spectra. The method was applied to different samples and showed superior performance than the reference deconvolution method incorporated in Bruker Topspin software. The proposed method provides a simple and fast way to obtain high-resolution NMR spectra in inhomogeneous fields that can facilitate the application of NMR spectroscopy in various fields.

Biomolecular NMR spectroscopy in the era of artificial intelligence

Biomolecular nuclear magnetic resonance (NMR) spectroscopy and artificial intelligence (AI) have a burgeoning synergy. Deep learning-based structural predictors have forever changed structural biology, yet these tools currently face limitations in accurately characterizing protein dynamics, allostery, and conformational heterogeneity. We begin by highlighting the unique abilities of biomolecular NMR spectroscopy to complement AI-based structural predictions toward addressing these knowledge gaps. We then highlight the direct integration of deep learning approaches into biomolecular NMR methods. AI-based tools can dramatically improve the acquisition and analysis of NMR spectra, enhancing the accuracy and reliability of NMR measurements, thus streamlining experimental processes. Additionally, deep learning enables the development of novel types of NMR experiments that were previously unattainable, expanding the scope and potential of biomolecular NMR spectroscopy. Ultimately, a combination of AI and NMR promises to further revolutionize structural biology on several levels, advance our understanding of complex biomolecular systems, and accelerate drug discovery efforts.

Denoising single MR spectra by deep learning: Miracle or mirage?

PURPOSE

The inherently poor SNR of MRS measurements presents a significant hurdle to its clinical application. Denoising by machine or deep learning (DL) was proposed as a remedy. It is investigated whether such denoising leads to lower estimate uncertainties or whether it essentially reduces noise in signal-free areas only.

METHODS

Noise removal based on supervised DL with U-nets was implemented using simulated 1 H MR spectra of human brain in two approaches: (1) via time-frequency domain spectrograms and (2) using 1D spectra as input. Quality of denoising was evaluated in three ways: (1) by an adapted fit quality score, (2) by traditional model fitting, and (3) by quantification via neural networks.

RESULTS

Visually appealing spectra were obtained; hinting that denoising is well-suited for MRS. However, an adapted denoising score showed that noise removal is inhomogeneous and more efficient for signal-free areas. This was confirmed by quantitative analysis of traditional fit results as well as DL quantitation following DL denoising. DL denoising, although apparently successful as judged by mean squared errors, led to substantially biased estimates in both implementations.

CONCLUSION

The implemented DL-based denoising techniques may be useful for display purposes, but do not help quantitative evaluations, confirming expectations based on estimation theory: Cramér Rao lower bounds defined by the original data and the appropriate fitting model cannot be circumvented in an unbiased way for single data sets, unless additional prior knowledge can be incurred in the form of parameter restrictions/relations or applicable substates.

A review of machine learning applications for the proton MR spectroscopy workflow

This literature review presents a comprehensive overview of machine learning (ML) applications in proton MR spectroscopy (MRS). As the use of ML techniques in MRS continues to grow, this review aims to provide the MRS community with a structured overview of the state-of-the-art methods. Specifically, we examine and summarize studies published between 2017 and 2023 from major journals in the MR field. We categorize these studies based on a typical MRS workflow, including data acquisition, processing, analysis, and artificial data generation. Our review reveals that ML in MRS is still in its early stages, with a primary focus on processing and analysis techniques, and less attention given to data acquisition. We also found that many studies use similar model architectures, with little comparison to alternative architectures. Additionally, the generation of artificial data is a crucial topic, with no consistent method for its generation. Furthermore, many studies demonstrate that artificial data suffers from generalization issues when tested on in vivo data. We also conclude that risks related to ML models should be addressed, particularly for clinical applications. Therefore, output uncertainty measures and model biases are critical to investigate. Nonetheless, the rapid development of ML in MRS and the promising results from the reviewed studies justify further research in this field.

Deep Learning Methodology for Obtaining Ultraclean Pure Shift Proton Nuclear Magnetic Resonance Spectra

Nuclear magnetic resonance (NMR) is one of the most powerful analytical techniques. In order to obtain high-quality NMR spectra, a real-time Zangger-Sterk (ZS) pulse sequence is employed to collect low-quality pure shift NMR data with high efficiency. Then, a neural network named AC-ResNet and a loss function named SM-CDMANE are developed to train a network model. The model with excellent abilities of suppressing noise, reducing line widths, discerning peaks, and removing artifacts is utilized to process the acquired NMR data. The processed spectra with noise and artifact suppression and small line widths are ultraclean and high-resolution. Peaks overlapped heavily can be resolved. Weak peaks, even hidden in the noise, can be discerned from noise. Artifacts, even as high as spectral peaks, can be removed completely while not suppressing peaks. Eliminating perfectly noise and artifacts and smoothing baseline make spectra ultraclean. The proposed methodology would greatly promote various NMR applications.

Pure Isotropic Proton NMR Spectra in Solids using Deep Learning

The resolution of proton solid-state NMR spectra is usually limited by broadening arising from dipolar interactions between spins. Magic-angle spinning alleviates this broadening by inducing coherent averaging. However, even the highest spinning rates experimentally accessible today are not able to completely remove dipolar interactions. Here, we introduce a deep learning approach to determine pure isotropic proton spectra from a two-dimensional set of magic-angle spinning spectra acquired at different spinning rates. Applying the model to 8 organic solids yields high-resolution ¹ H solid-state NMR spectra with isotropic linewidths in the 50-400 Hz range.

NMR spectrum reconstruction as a pattern recognition problem

A new deep neural network based on the WaveNet architecture (WNN) is presented, which is designed to grasp specific patterns in the NMR spectra. When trained at a fixed non-uniform sampling (NUS) schedule, the WNN benefits from pattern recognition of the corresponding point spread function (PSF) pattern produced by each spectral peak resulting in the highest quality and robust reconstruction of the NUS spectra as demonstrated in simulations and exemplified in this work on 2D ¹H-¹⁵N correlation spectra of three representative globular proteins with different sizes: Ubiquitin (8.6 kDa), Azurin (14 kDa), and Malt1 (44 kDa). The pattern recognition by WNN is also demonstrated for successful virtual homo-decoupling in a 2D methyl ¹H-¹³C - HMQC spectrum of MALT1. We demonstrate using WNN that prior knowledge about the NUS schedule, which so far was not been fully exploited, can be used for designing new powerful NMR processing techniques that surpass the existing algorithmic methods.

The rise of the machines in chemistry

The use of artificial intelligence and, more specifically, deep learning methods in chemistry is becoming increasingly common. Applications in informatics fields, such as cheminformatics and proteomics, structural biology, and spectroscopy, including NMR, are on the rise. Recent developments in model architectures, such as graph convolutional neural networks and transformers, have been enabled by advancements in computational hardware and software. However, model architectures with more predictive power often require larger amounts of training data, which can be challenging to acquire, but this requirement can be mitigated through techniques like pretraining and fine-tuning. In spite of these successes, challenges remain, such as normalization and scaling of data, availability of experimentally acquired data, and model explainability.

SpecDB: A relational database for archiving biomolecular NMR spectral data

NMR is a valuable experimental tool in the structural biologist's toolkit to elucidate the structures, functions, and motions of biomolecules. The progress of machine learning, particularly in structural biology, reveals the critical importance of large, diverse, and reliable datasets in developing new methods and understanding in structural biology and science more broadly. Biomolecular NMR research groups produce large amounts of data, and there is renewed interest in organizing these data to train new, sophisticated machine learning architectures and to improve biomolecular NMR analysis pipelines. The foundational data type in NMR is the free-induction decay (FID). There are opportunities to build sophisticated machine learning methods to tackle long-standing problems in NMR data processing, resonance assignment, dynamics analysis, and structure determination using NMR FIDs. Our goal in this study is to provide a lightweight, broadly available tool for archiving FID data as it is generated at the spectrometer, and grow a new resource of FID data and associated metadata. This study presents a relational schema for storing and organizing the metadata items that describe an NMR sample and FID data, which we call Spectral Database (SpecDB). SpecDB is implemented in SQLite and includes a Python software library providing a command-line application to create, organize, query, backup, share, and maintain the database. This set of software tools and database schema allow users to store, organize, share, and learn from NMR time domain data. SpecDB is freely available under an open source license at https://github.rpi.edu/RPIBioinformatics/SpecDB.

Deep Learning-Based Method for Compound Identification in NMR Spectra of Mixtures

Nuclear magnetic resonance (NMR) spectroscopy is highly unbiased and reproducible, which provides us a powerful tool to analyze mixtures consisting of small molecules. However, the compound identification in NMR spectra of mixtures is highly challenging because of chemical shift variations of the same compound in different mixtures and peak overlapping among molecules. Here, we present a pseudo-Siamese convolutional neural network method (pSCNN) to identify compounds in mixtures for NMR spectroscopy. A data augmentation method was implemented for the superposition of several NMR spectra sampled from a spectral database with random noises. The augmented dataset was split and used to train, validate and test the pSCNN model. Two experimental NMR datasets (flavor mixtures and additional flavor mixture) were acquired to benchmark its performance in real applications. The results show that the proposed method can achieve good performances in the augmented test set (ACC = 99.80%, TPR = 99.70% and FPR = 0.10%), the flavor mixtures dataset (ACC = 97.62%, TPR = 96.44% and FPR = 2.29%) and the additional flavor mixture dataset (ACC = 91.67%, TPR = 100.00% and FPR = 10.53%). We have demonstrated that the translational invariance of convolutional neural networks can solve the chemical shift variation problem in NMR spectra. In summary, pSCNN is an off-the-shelf method to identify compounds in mixtures for NMR spectroscopy because of its accuracy in compound identification and robustness to chemical shift variation.

Towards autonomous analysis of chemical exchange saturation transfer experiments using deep neural networks

Macromolecules often exchange between functional states on timescales that can be accessed with NMR spectroscopy and many NMR tools have been developed to characterise the kinetics and thermodynamics of the exchange processes, as well as the structure of the conformers that are involved. However, analysis of the NMR data that report on exchanging macromolecules often hinges on complex least-squares fitting procedures as well as human experience and intuition, which, in some cases, limits the widespread use of the methods. The applications of deep neural networks (DNNs) and artificial intelligence have increased significantly in the sciences, and recently, specifically, within the field of biomolecular NMR, where DNNs are now available for tasks such as the reconstruction of sparsely sampled spectra, peak picking, and virtual decoupling. Here we present a DNN for the analysis of chemical exchange saturation transfer (CEST) data reporting on two- or three-site chemical exchange involving sparse state lifetimes of between approximately 3-60 ms, the range most frequently observed via experiment. The work presented here focuses on the ¹H CEST class of methods that are further complicated, in relation to applications to other nuclei, by anti-phase features. The developed DNNs accurately predict the chemical shifts of nuclei in the exchanging species directly from anti-phase ¹H^N CEST profiles, along with an uncertainty associated with the predictions. The performance of the DNN was quantitatively assessed using both synthetic and experimental anti-phase CEST profiles. The assessments show that the DNN accurately determines chemical shifts and their associated uncertainties. The DNNs developed here do not contain any parameters for the end-user to adjust and the method therefore allows for autonomous analysis of complex NMR data that report on conformational exchange.

The exposome paradigm to predict environmental health in terms of systemic homeostasis and resource balance based on NMR data science

The environment, from microbial ecosystems to recycled resources, fluctuates dynamically due to many physical, chemical and biological factors, the profile of which reflects changes in overall state, such as environmental illness caused by a collapse of homeostasis. To evaluate and predict environmental health in terms of systemic homeostasis and resource balance, a comprehensive understanding of these factors requires an approach based on the "exposome paradigm", namely the totality of exposure to all substances. Furthermore, in considering sustainable development to meet global population growth, it is important to gain an understanding of both the circulation of biological resources and waste recycling in human society. From this perspective, natural environment, agriculture, aquaculture, wastewater treatment in industry, biomass degradation and biodegradable materials design are at the forefront of current research. In this respect, nuclear magnetic resonance (NMR) offers tremendous advantages in the analysis of samples of molecular complexity, such as crude bio-extracts, intact cells and tissues, fibres, foods, feeds, fertilizers and environmental samples. Here we outline examples to promote an understanding of recent applications of solution-state, solid-state, time-domain NMR and magnetic resonance imaging (MRI) to the complex evaluation of organisms, materials and the environment. We also describe useful databases and informatics tools, as well as machine learning techniques for NMR analysis, demonstrating that NMR data science can be used to evaluate the exposome in both the natural environment and human society towards a sustainable future.

FID-Net: A versatile deep neural network architecture for NMR spectral reconstruction and virtual decoupling

In recent years, the transformative potential of deep neural networks (DNNs) for analysing and interpreting NMR data has clearly been recognised. However, most applications of DNNs in NMR to date either struggle to outperform existing methodologies or are limited in scope to a narrow range of data that closely resemble the data that the network was trained on. These limitations have prevented a widescale uptake of DNNs in NMR. Addressing this, we introduce FID-Net, a deep neural network architecture inspired by WaveNet, for performing analyses on time domain NMR data. We first demonstrate the effectiveness of this architecture in reconstructing non-uniformly sampled (NUS) biomolecular NMR spectra. It is shown that a single network is able to reconstruct a diverse range of 2D NUS spectra that have been obtained with arbitrary sampling schedules, with a range of sweep widths, and a variety of other acquisition parameters. The performance of the trained FID-Net in this case exceeds or matches existing methods currently used for the reconstruction of NUS NMR spectra. Secondly, we present a network based on the FID-Net architecture that can efficiently virtually decouple ¹³C_α-¹³C_β couplings in HNCA protein NMR spectra in a single shot analysis, while at the same time leaving glycine residues unmodulated. The ability for these DNNs to work effectively in a wide range of scenarios, without retraining, paves the way for their widespread usage in analysing NMR data.

NMR as a "Gold Standard" Method in Drug Design and Discovery

Studying disease models at the molecular level is vital for drug development in order to improve treatment and prevent a wide range of human pathologies. Microbial infections are still a major challenge because pathogens rapidly and continually evolve developing drug resistance. Cancer cells also change genetically, and current therapeutic techniques may be (or may become) ineffective in many cases. The pathology of many neurological diseases remains an enigma, and the exact etiology and underlying mechanisms are still largely unknown. Viral infections spread and develop much more quickly than does the corresponding research needed to prevent and combat these infections; the present and most relevant outbreak of SARS-CoV-2, which originated in Wuhan, China, illustrates the critical and immediate need to improve drug design and development techniques. Modern day drug discovery is a time-consuming, expensive process. Each new drug takes in excess of 10 years to develop and costs on average more than a billion US dollars. This demonstrates the need of a complete redesign or novel strategies. Nuclear Magnetic Resonance (NMR) has played a critical role in drug discovery ever since its introduction several decades ago. In just three decades, NMR has become a "gold standard" platform technology in medical and pharmacology studies. In this review, we present the major applications of NMR spectroscopy in medical drug discovery and development. The basic concepts, theories, and applications of the most commonly used NMR techniques are presented. We also summarize the advantages and limitations of the primary NMR methods in drug development.