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.

Information-Theoretic Neural Decoding Reproduces Several Laws of Human Behavior.

Human response times conform to several regularities including the Hick-Hyman law, the power law of practice, speed-accuracy trade-offs, and the Stroop effect. Each of these has been thoroughly modeled in isolation, but no account describes these phenomena as predictions of a unified framework. We provide such a framework and show that the phenomena arise as decoding times in a simple neural rate code with an entropy stopping threshold. Whereas traditional information-theoretic encoding systems exploit task statistics to optimize encoding strategies, we move this optimization to the decoder, treating it as a Bayesian ideal observer that can track transmission statistics as prior information during decoding. Our approach allays prominent concerns that applying information-theoretic perspectives to modeling brain and behavior requires complex encoding schemes that are incommensurate with neural encoding.

Rapid and automated lipid profiling by nuclear magnetic resonance spectroscopy using neural networks.

Nuclear magnetic resonance (NMR) spectroscopy is a powerful tool for quantitative metabolomics; however, quantification of metabolites from NMR data is often a slow and tedious process requiring user input and expertise. In this study, we propose a neural network approach for rapid, automated lipid identification and quantification from NMR data. Multilayered perceptron (MLP) networks were developed with NMR spectra as the input and lipid concentrations as output. Three large synthetic datasets were generated, each with 55,000 spectra from an original 30 scans of reference standards, by using linear combinations of standards and simulating experimental-like modifications (line broadening, noise, peak shifts, baseline shifts) and common interference signals (water, tetramethylsilane, extraction solvent), and were used to train MLPs for robust prediction of lipid concentrations. The performances of MLPS were first validated on various synthetic datasets to assess the effect of incorporating different modifications on their accuracy. The MLPs were then evaluated on experimentally acquired data from complex lipid mixtures. The MLP-derived lipid concentrations showed high correlations and slopes close to unity for most of the quantified lipid metabolites in experimental mixtures compared with ground-truth concentrations. The most accurate, robust MLP was used to profile lipids in lipophilic hepatic extracts from a rat metabolomics study. The MLP lipid results analyzed by two-way ANOVA for dietary and sex differences were similar to those obtained with a conventional NMR quantification method. In conclusion, this study demonstrates the potential and feasibility of a neural network approach for improving speed and automation in NMR lipid profiling and this approach can be easily tailored to other quantitative, targeted spectroscopic analyses in academia or industry.

Depth-dependence of the underwater noise emission from melting glacier ice.

Submarine-melting of ice at the glacier-ocean interface accounts for a large portion of the ice-loss at tidewater glaciers and produces sound via bubble-release. The sound production is dominant in the sub-surface region near the glacier-ocean interface. This depth-dependence of the sound is studied by melting ice blocks in a glacial bay at various depths up to 20 m and recording their acoustics over a large frequency range. The acoustic energy decreases with depth in line with expectations from the physics of the phenomenon and is fit to an exponentially decaying curve. The estimated variation will be useful for interpreting the sound in marine-terminating glaciers bays in terms of the submarine-melting activity.

Quantitative MRI of diffuse liver diseases: techniques and tissue-mimicking phantoms.

Quantitative magnetic resonance imaging (MRI) techniques are emerging as non-invasive alternatives to biopsy for assessment of diffuse liver diseases of iron overload, steatosis and fibrosis. For testing and validating the accuracy of these techniques, phantoms are often used as stand-ins to human tissue to mimic diffuse liver pathologies. However, currently, there is no standardization in the preparation of MRI-based liver phantoms for mimicking iron overload, steatosis, fibrosis or a combination of these pathologies as various sizes and types of materials are used to mimic the same liver disease. Liver phantoms that mimic specific MR features of diffuse liver diseases observed in vivo are important for testing and calibrating new MRI techniques and for evaluating signal models to accurately quantify these features. In this study, we review the liver morphology associated with these diffuse diseases, discuss the quantitative MR techniques for assessing these liver pathologies, and comprehensively examine published liver phantom studies and discuss their benefits and limitations.

Multi-Tissue Time-Domain NMR Metabolomics Investigation of Time-Restricted Feeding in Male and Female Nile Grass Rats.

Metabolic disease resulting from overnutrition is prevalent and rapidly increasing in incidence in modern society. Time restricted feeding (TRF) dietary regimens have recently shown promise in attenuating some of the negative metabolic effects associated with chronic nutrient stress. The purpose of this study is to utilize a multi-tissue metabolomics approach using nuclear magnetic resonance (NMR) spectroscopy to investigate TRF and sex-specific effects of high-fat diet in a diurnal Nile grass rat model. Animals followed a six-week dietary protocol on one of four diets: chow ad libitum, high-fat ad libitum (HF-AD), high-fat early TRF (HF-AM), or high-fat late TRF (HF-PM), and their liver, heart, and white adipose tissues were harvested at the end of the study and were analyzed by NMR. Time-domain complete reduction to amplitude-frequency table (CRAFT) was used to semi-automate and systematically quantify metabolites in liver, heart, and adipose tissues while minimizing operator bias. Metabolite profiling and statistical analysis revealed lipid remodeling in all three tissues and ectopic accumulation of cardiac and hepatic lipids for HF-AD feeding compared to a standard chow diet. Animals on TRF high-fat diet had lower lipid levels in the heart and liver compared to the ad libitum group; however, no significant differences were noted for adipose tissue. Regardless of diet, females exhibited greater amounts of hepatic lipids compared to males, while no consistent differences were shown in adipose and heart. In conclusion, this study demonstrates the feasibility of performing systematic and time-efficient multi-tissue NMR metabolomics to elucidate metabolites involved in the crosstalk between different metabolic tissues and provides a more holistic approach to better understand the etiology of metabolic disease and the effects of TRF on metabolic profiles.

Early Time-Restricted Feeding Amends Circadian Clock Function and Improves Metabolic Health in Male and Female Nile Grass Rats.

Lengthening the daily eating period contributes to the onset of obesity and metabolic syndrome. Dietary approaches, including energy restriction and time-restricted feeding, are promising methods to combat metabolic disorders. This study explored the effect of early and late time-restricted feeding (TRF) on weight and adiposity, food consumption, glycemic control, clock gene expression, and liver metabolite composition in diurnal Nile grass rats (NGRs). Adult male and female Nile grass rats were randomly assigned to one of three groups: (1) access to a 60% high-fat (HF) diet ad-libitum (HF-AD), (2) time-restricted access to the HF diet for the first 6 h of the 12 h light/active phase (HF-AM) or (3) the second 6 h of the 12 h light/active phase (HF-PM). Animals remained on their respective protocols for six weeks. TRF reduced total energy consumption and weight gain, and early TRF (HF-AM) reduced fasting blood glucose, restored Per1 expression, and reduced liver lipid levels. Although sex-dependent differences were observed for fat storage and lipid composition, TRF improved metabolic parameters in both male and female NGRs. In conclusion, this study demonstrated that early TRF protocol benefits weight management, improves lipid and glycemic control, and restores clock gene expression in NGRs.

CRAFT for NMR lipidomics: Targeting lipid metabolism in leucine-supplemented tumor-bearing mice.

Lipid profiling by 1 H-NMR has gained increasing utility in many fields because of its intrinsically quantitative, nondestructive nature and the ability to differentiate small molecules based on their spectral location. Most nuclear magnetic resonance (NMR) techniques for metabolite quantification use frequency domain analysis that involves many user-dependent steps such as phase and baseline correction and quantification by either manual integration or peak fitting. Recently, Bayesian analysis of time-domain NMR data has been shown to reduce operator bias and increase automation in NMR spectroscopy. In this study, we demonstrate the use of CRAFT (complete reduction to amplitude-frequency table), a Bayesian-based approach to automate processing in NMR-based lipidomics using lipid standards and tissue samples of healthy and tumor-bearing mice supplemented with leucine. Complex mixtures of lipid standards were prepared and examined using CRAFT to validate it against conventional Fourier transform (FT)-NMR and derive a fingerprint to be used for analyzing lipid profiles of serum and liver samples. CRAFT and FT-NMR were comparable in accuracy, with CRAFT achieving higher correlation in quantifying several lipid species. Analysis of the serum lipidome of tumor-bearing mice revealed hyperlipidemia and no signs of hepatic triglyceride accumulation compared with that of the healthy group demonstrating that the tumor-bearing mice were in a state of precachexia. Leucine-supplementation was associated with minimal changes in the lipid profile in both tissues. In conclusion, our study demonstrates that the CRAFT method can accurately identify and quantify lipids in complex lipid mixtures and murine tissue samples and, hence, will increase automation and reproducibility in NMR-based lipidomics.