Artificial neural networks for quantitative online NMR spectroscopy

Explainable AI to Facilitate Understanding of Neural Network-Based Metabolite Profiling Using NMR Spectroscopy

Neural networks (NNs) are emerging as a rapid and scalable method for quantifying metabolites directly from nuclear magnetic resonance (NMR) spectra, but the nonlinear nature of NNs precludes understanding of how a model makes predictions. This study implements an explainable artificial intelligence algorithm called integrated gradients (IG) to elucidate which regions of input spectra are the most important for the quantification of specific analytes. The approach is first validated in simulated mixture spectra of eight aqueous metabolites and then investigated in experimentally acquired lipid spectra of a reference standard mixture and a murine hepatic extract. The IG method revealed that, like a human spectroscopist, NNs recognize and quantify analytes based on an analyte's respective resonance line-shapes, amplitudes, and frequencies. NNs can compensate for peak overlap and prioritize specific resonances most important for concentration determination. Further, we show how modifying a NN training dataset can affect how a model makes decisions, and we provide examples of how this approach can be used to de-bug issues with model performance. Overall, results show that the IG technique facilitates a visual and quantitative understanding of how model inputs relate to model outputs, potentially making NNs a more attractive option for targeted and automated NMR-based metabolomics.

Polarization transfer methods for quantitative analysis of flowing mixtures with benchtop 13C NMR spectroscopy

Benchtop NMR spectroscopy is attractive for process monitoring; however, there are still drawbacks that often hamper its use, namely, the comparatively low spectral resolution in 1H NMR, as well as the low signal intensities and problems with the premagnetization of flowing samples in 13C NMR. We show here that all these problems can be overcome by using 1H-13C polarization transfer methods. Two ternary test mixtures (one with overlapping peaks in the 1H NMR spectrum and one with well-separated peaks, which was used as a reference) were studied with a 1 T benchtop NMR spectrometer using the polarization transfer sequence PENDANT (polarization enhancement that is nurtured during attached nucleus testing). The mixtures were analyzed quantitatively in stationary as well as in flow experiments by PENDANT enhanced 13C NMR experiments, and the results were compared with those from the gravimetric sample preparation and from standard 1H and 13C NMR spectroscopy. Furthermore, as a proxy for a process monitoring application, continuous dilution experiments were carried out, and the composition of the mixture was monitored in a flow setup by 13C NMR benchtop spectroscopy with PENDANT. The results demonstrate the high potential of polarization transfer methods for applications in quantitative process analysis with benchtop NMR instruments, in particular with flowing samples.

Deconvolution and Analysis of the 1H NMR Spectra of Crude Reaction Mixtures

Nuclear magnetic resonance (NMR) spectroscopy is an important analytical technique in synthetic organic chemistry, but its integration into high-throughput experimentation workflows has been limited by the necessity of manually analyzing the NMR spectra of new chemical entities. Current efforts to automate the analysis of NMR spectra rely on comparisons to databases of reported spectra for known compounds and, therefore, are incompatible with the exploration of new chemical space. By reframing the NMR spectrum of a reaction mixture as a joint probability distribution, we have used Hamiltonian Monte Carlo Markov Chain and density functional theory to fit the predicted NMR spectra to those of crude reaction mixtures. This approach enables the deconvolution and analysis of the spectra of mixtures of compounds without relying on reported spectra. The utility of our approach to analyze crude reaction mixtures is demonstrated with the experimental spectra of reactions that generate a mixture of isomers, such as Wittig olefination and C-H functionalization reactions. The correct identification of compounds in a reaction mixture and their relative concentrations is achieved with a mean absolute error as low as 1%.

Mixtures Recomposition by Neural Nets: A Multidisciplinary Overview

Artificial Neural Networks (ANNs) are transforming how we understand chemical mixtures, providing an expressive view of the chemical space and multiscale processes. Their hybridization with physical knowledge can bridge the gap between predictivity and understanding of the underlying processes. This overview explores recent progress in ANNs, particularly their potential in the 'recomposition' of chemical mixtures. Graph-based representations reveal patterns among mixture components, and deep learning models excel in capturing complexity and symmetries when compared to traditional Quantitative Structure-Property Relationship models. Key components, such as Hamiltonian networks and convolution operations, play a central role in representing multiscale mixtures. The integration of ANNs with Chemical Reaction Networks and Physics-Informed Neural Networks for inverse chemical kinetic problems is also examined. The combination of sensors with ANNs shows promise in optical and biomimetic applications. A common ground is identified in the context of statistical physics, where ANN-based methods iteratively adapt their models by blending their initial states with training data. The concept of mixture recomposition unveils a reciprocal inspiration between ANNs and reactive mixtures, highlighting learning behaviors influenced by the training environment.

Key steps for improving bacterial SERS signals in complex samples: Separation, recognition, detection, and analysis

Rapid and reliable detection of pathogenic bacteria is absolutely essential for research in environmental science, food quality, and medical diagnostics. Surface-enhanced Raman spectroscopy (SERS), as an emerging spectroscopic technique, has the advantages of high sensitivity, good selectivity, rapid detection speed, and portable operation, which has been broadly used in the detection of pathogenic bacteria in different kinds of complex samples. However, the SERS detection method is also challenging in dealing with the detection difficulties of bacterial samples in complex matrices, such as interference from complex matrices, confusion of similar bacteria, and complexity of data processing. Therefore, researchers have developed some technologies to assist in SERS detection of bacteria, including both the front-end process of obtaining bacterial sample data and the back-end data processing process. The review summarizes the key steps for improving bacterial SERS signals in complex samples: separation, recognition, detection, and analysis, highlighting the principles of each step and the key roles for SERS pathogenic bacteria analysis, and the interconnectivity between each step. In addition, the current challenges in the practical application of SERS technology and the development trends are discussed. The purpose of this review is to deepen researchers' understanding of the various stages of using SERS technology to detect bacteria in complex sample matrices, and help them find new breakthroughs in different stages to facilitate the detection and control of bacteria in complex samples.

In Situ Hyperpolarization Enables 15N and 13C Benchtop NMR at Natural Isotopic Abundance

Without employing isotopic labeling, we demonstrate the generation of 15N and 13C NMR signals for molecules containing -NH2 motifs using benchtop NMR spectrometers (1-1.4 T). Specifically, high-SNR (>50) detection of ammonia, 4-aminopyridine, benzylamine, and phenethylamine dissolved in methanol or dichloromethane is demonstrated after only 10 s of parahydrogen bubbling using signal amplification by reversible exchange and applying a pulse sequence based on spin-lock-induced crossing. Optimization of the sequence parameters allows us to achieve up to 12% 15N and 0.4% 13C polarization in situ without the need for the sample transfer typically employed in other hyperpolarization methods. Moreover, hyperpolarization is generated continuously without having to stop the parahydrogen bubbling to reset magnetization, paving the way toward fast 2D spectroscopic methods and relaxometry. The provided methodology may find application for the identification of diluted chemicals relevant to industry and research with the aid of affordable benchtop NMR spectrometers.

Successful combination of benchtop nuclear magnetic resonance spectroscopy and chemometric tools: A review

Low-field nuclear magnetic resonance (NMR) has three general modalities: spectroscopy, imaging, and relaxometry. In the last twelve years, the modality of spectroscopy, also known as benchtop NMR, compact NMR, or just low-field NMR, has undergone instrumental development due to new permanent magnetic materials and design. As a result, benchtop NMR has emerged as a powerful analytical tool for use in process analytical control (PAC). Nevertheless, the successful application of NMR devices as an analytical tool in several areas is intrinsically linked to its coupling with different chemometric methods. This review focuses on the evolution of benchtop NMR and chemometrics in chemical analysis, including applications in fuels, foods, pharmaceuticals, biochemicals, drugs, metabolomics, and polymers. The review also presents different low-resolution NMR methods for spectrum acquisition and chemometric techniques for calibration, classification, discrimination, data fusion, calibration transfer, multi-block and multi-way.

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.

Rapid Quantitative Analysis of IR Absorption Spectra for Trace Gas Detection by Artificial Neural Networks Trained with Synthetic Data

Infrared absorption spectroscopy is a widely used tool to quantify and monitor compositions of gases. The concentration information is often retrieved by fitting absorption profiles to the acquired spectra, utilizing spectroscopic databases. In complex gas matrices an expanded parameter space leads to long computation times of the fitting routines due to the increased number of spectral features that need to be computed for each iteration during the fit. This hinders the capability of real-time analysis of the gas matrix. Here, an artificial neural network (ANN) is employed for rapid prediction of gas concentrations in complex infrared absorption spectra composed of mixtures of CO and N₂O. Experimental data is acquired with a mid-infrared dual frequency comb spectrometer. To circumvent the experimental collection of huge amounts of training data, the network is trained on synthetically generated spectra. The spectra are based on simulated absorption profiles making use of the HITRAN database. In addition, the spectrometer’s influence on the measured spectra is characterized and included in the synthetic training data generation. The ANN was tested on measured spectra and compared to a non-linear least squares fitting algorithm. An average evaluation time of 303 µs for a single measured spectrum was achieved. Coefficients of determination were 0.99997 for the predictions of N₂O concentrations and 0.99987 for the predictions of CO concentrations, with uncertainties on the predicted concentrations between 0.04 and 0.18 ppm for 0 to 100 ppm N₂O and between 0.05 and 0.18 ppm for 0 to 60 ppm CO.

Standardization and Control of Grignard Reactions in a Universal Chemical Synthesis Machine using online NMR

A big problem with the chemistry literature is that it is not standardized with respect to precise operational parameters, and real time corrections are hard to make without expert knowledge. This lack of context means difficult reproducibility because many steps are ambiguous, and hence depend on tacit knowledge. Here we present the integration of online NMR into an automated chemical synthesis machine (CSM aka. "Chemputer" which is capable of small-molecule synthesis using a universal programming language) to allow automated analysis and adjustment of reactions on the fly. The system was validated and benchmarked by using Grignard reactions which were chosen due to their importance in synthesis. The system was monitored in real time using online-NMR, and spectra were measured continuously during the reactions. This shows that the synthesis being done in the Chemputer can be dynamically controlled in response to feedback optimizing the reaction conditions according to the user requirements.

Automated metabolic assignment: Semi-supervised learning in metabolic analysis employing two dimensional Nuclear Magnetic Resonance (NMR)

•

Automatic assignment of metabolites of 2D-TOCSY NMR spectra.

•

Semi-supervised learning for metabolic profiling.

•

Deconvolution and metabolic profiling of 2D NMR spectra using Machine Learning.

•

Accurate Automatic multicomponent assignment of 2D NMR spectrum.

Metabolomics is an expanding field of medical diagnostics since many diseases cause metabolic reprogramming alteration. Additionally, the metabolic point of view offers an insight into the molecular mechanisms of diseases. Due to the complexity of metabolic assignment dependent on the 1D NMR spectral analysis, 2D NMR techniques are preferred because of spectral resolution issues. Thus, in this work, we introduce an automated metabolite identification and assignment from ¹H-¹H TOCSY (total correlation spectroscopy) using real breast cancer tissue. The new approach is based on customized and extended semi-supervised classifiers: KNFST, SVM, third (PC3) and fourth (PC4) degree polynomial. In our approach, metabolic assignment is based only on the vertical and horizontal frequencies of the metabolites in the ¹H–¹H TOCSY. KNFST and SVM show high performance (high accuracy and low mislabeling rate) in relatively low size of initially labeled training data. PC3 and PC4 classifiers showed lower accuracy and high mislabeling rates, and both classifiers fail to provide an acceptable accuracy at extremely low size (≤9% of the entire dataset) of initial training data. Additionally, semi-supervised classifiers were implemented to obtain a fully automatic procedure for signal assignment and deconvolution of TOCSY, which is a big step forward in NMR metabolic profiling. A set of 27 metabolites were deduced from the TOCSY, and their assignments agreed with the metabolites deduced from a 1D NMR spectrum of the same sample analyzed by conventional human-based methodology.

Sampling and Analysis in Flow: The Keys to Smarter, More Controllable, and Sustainable Fine-Chemical Manufacturing

Process analytical technology (PAT) is a system designed to help chemists better understand and control manufacturing processes. PAT systems operate through the combination of analytical devices, reactor control elements, and mathematical models to ensure the quality of the final product through a quality by design (QbD) approach. The expansion of continuous manufacturing in the pharmaceutical and fine-chemical industry requires the development of PAT tools suitable for continuous operation in the environment of flow reactors. This requires innovative approaches to sampling and analysis from flowing media to maintain the integrity of the reactor content and the analyte of interest. The following Review discusses examples of PAT tools implemented in flow chemistry for the preparation of small organic molecules, and applications of self-optimization tools.

Advanced Real-Time Process Analytics for Multistep Synthesis in Continuous Flow*

In multistep continuous flow chemistry, studying complex reaction mixtures in real time is a significant challenge, but provides an opportunity to enhance reaction understanding and control. We report the integration of four complementary process analytical technology tools (NMR, UV/Vis, IR and UHPLC) in the multistep synthesis of an active pharmaceutical ingredient, mesalazine. This synthetic route exploits flow processing for nitration, high temperature hydrolysis and hydrogenation reactions, as well as three inline separations. Advanced data analysis models were developed (indirect hard modeling, deep learning and partial least squares regression), to quantify the desired products, intermediates and impurities in real time, at multiple points along the synthetic pathway. The capabilities of the system have been demonstrated by operating both steady state and dynamic experiments and represents a significant step forward in data-driven continuous flow synthesis.

Fourier and Laplace-like low-field NMR spectroscopy: The perspectives of multivariate and artificial neural networks analyses

Low field Nuclear Magnetic Resonance (LF-NMR) is a rich source of information for a wide range of samples types. These can be hard or soft solids, such as plastics or elastomers; bulk liquids or liquids absorbed in porous materials, and can come from biomaterials, biological tissues, archaeological artifacts, cultural heritage objects. LF-NMR instruments present a significant advance especially for in situ, ex situ and in vivo measurement of relaxation and diffusion. Moreover, high resolution 1D and 2D spectroscopy, as well as magnetic resonance (MR) imaging are available in these fields. In this work we discuss the advanced analysis of the data measured in LF-NMR from the perspectives of tertiary level that implies the analysis on principal components (PCA), and on the quaternary analysis that uses an artificial neural network (ANN). The principles of PCA and ANN are largely discussed. For the PCA analysis, a series of 52 spectra were analyzed, having been recorded in vivo by LF-NMR. Of these spectra, 38 were generated from normal uterus, 7 by uterus tissue with endometrial cancer, and another 7 were obtained from tissues of women with uterine cervical cancer. The PC1 vs PC2 plot was further analyzed using an artificial neural network, and the results are presented as 2D maps of probability. Furthermore, the perspectives of applying an ANN to solve the problem of Laplace-like inversion are discussed. An example of such ANN was presented and the performance was discussed. Finally, a model of complex ANN, capable to sequentially solve this kind of problems specific to LF-NMR is proposed and discussed.