Evaluation of Multivariate Classification Models for Analyzing NMR Metabolomics Data

Shifting-corrected regularized regression for 1H NMR metabolomics identification and quantification

The process of identifying and quantifying metabolites in complex mixtures plays a critical role in metabolomics studies to obtain an informative interpretation of underlying biological processes. Manual approaches are time-consuming and heavily reliant on the knowledge and assessment of nuclear magnetic resonance (NMR) experts. We propose a shifting-corrected regularized regression method, which identifies and quantifies metabolites in a mixture automatically. A detailed algorithm is also proposed to implement the proposed method. Using a novel weight function, the proposed method is able to detect and correct peak shifting errors caused by fluctuations in experimental procedures. Simulation studies show that the proposed method performs better with regard to the identification and quantification of metabolites in a complex mixture. We also demonstrate real data applications of our method using experimental and biological NMR mixtures.

Applications of machine learning in metabolomics: Disease modeling and classification

Metabolomics research has recently gained popularity because it enables the study of biological traits at the biochemical level and, as a result, can directly reveal what occurs in a cell or a tissue based on health or disease status, complementing other omics such as genomics and transcriptomics. Like other high-throughput biological experiments, metabolomics produces vast volumes of complex data. The application of machine learning (ML) to analyze data, recognize patterns, and build models is expanding across multiple fields. In the same way, ML methods are utilized for the classification, regression, or clustering of highly complex metabolomic data. This review discusses how disease modeling and diagnosis can be enhanced via deep and comprehensive metabolomic profiling using ML. We discuss the general layout of a metabolic workflow and the fundamental ML techniques used to analyze metabolomic data, including support vector machines (SVM), decision trees, random forests (RF), neural networks (NN), and deep learning (DL). Finally, we present the advantages and disadvantages of various ML methods and provide suggestions for different metabolic data analysis scenarios.

Effects of enteral nutrition with different energy supplies on metabolic changes and organ damage in burned rats

BACKGROUND

Enteral nutrition (EN) is an important treatment for burn patients. However, severe gastrointestinal damage caused by major burns often leads to EN intolerance. Trophic EN solves this problem basically, but how to transition from trophic EN to standard EN smoothly is still a challenge in burn clinical nutrition. The aim of this study is to investigate the effects of EN with different energy supplies on metabolic changes, organ damage and prognosis in burned rats.

METHODS

Different feeding regimens were designed based on the continuous monitoring of resting energy expenditure in rats. Thirty-two Sprague-Dawley rats were randomly divided into a normal control group, burn +50% REE group, burn +75% REE group and burn +100% REE group. At the end of a nutritional treatment cycle (14th day), nuclear magnetic resonance spectroscopy, blood biochemistry analysis and quantification of subscab bacteria were performed to explore the differences in metabolic changes, degrees of organ damage and prognoses between the groups.

RESULTS

Sixteen metabolites involving seven metabolic pathways were identified from the different energy supply groups. After burn injury, resting energy consumption and body weight loss increased obviously. Meanwhile, weight loss was inversely related to energy supply. The greatest changes in the degree of organ damage, the level of plasma proteins, lipids and endotoxins, as well as the quantification of subscab bacteria were observed in the 50% REE group, followed by the 75 and 100% groups.

CONCLUSIONS

Achieving an early balance between energy supply and expenditure is conducive to mitigating metabolic disorders and improving prognosis after burn injury.

Metabolite quantification: A fluorescence-based method for urine sample normalization prior to 1H-NMR analysis

INTRODUCTION

Metabolomics is a multi-discipline approach to systems biology that provides a snapshot of the metabolic status of a cell, tissue, or organism. Metabolomics uses mass spectroscopy (MS) and nuclear magnetic resonance (NMR) to analyze biological samples for low molecular weight metabolites.

OBJECTIVE

Normalize urine sample pre-acquisition to perform a targeted quantitative analysis of selected metabolites in rat urine.

METHODS

Urine samples were provided from rats on a control diet (n = 10) and moderate sucrose diet (n = 8) collected in a metabolic cage during an eight hour fast. Urine from each sample was prepared by two different methods. One sample was a non-normalized sample of 1200 µL and the second sample was a variable volume-normalized to the concentration of urobilin in a standard sample of urine. The urobilin concentration in all samples was determined by fluorescence. Ten metabolites for each non-normalized and normalized urine sample were quantified by integration to an internal standard of DSS.

RESULTS

Both groups showed an improvement in pH range going from non-normalized to normalized samples. In the group on the control diet, eight metabolites had significant improvement in range, while the remaining two metabolites had insignificant improvement in range comparing the non-normalized sample to the normalized sample. In the group on the moderate sucrose diet all ten metabolites showed significant improvement in range going from non-normalized to normalized samples.

CONCLUSIONS

These findings describe a pre-acquisition method of urine normalization to adjust for differences in hydration state of each organism. This results in a narrower concentration range in a targeted analysis.

Effect of fixed-dose combination of insulin degludec and liraglutide on apoB-containing lipoprotein subclasses and HDL lipidome in type 2 diabetes

AIMS

Administration of insulin degludec and liraglutide (IDegLira) correlates to fasting lipid profile changes of diabetic patients, while data concerning apoB-containing lipoprotein subclasses and HDL lipidome are scarce. We evaluated its effect on fasting lipid parameters, apolipoproteins, apoB-containing lipoprotein subclasses and HDL lipidome in patients with type 2 diabetes.

METHODS

Sixty three patients with HbA1c > 7 % on oral glucose-lowering drugs received either IDegLira or insulin degludec for 3 months. Lipoprotein subfraction profile was determined through Lipoprint method, whereas HDL lipid composition via ¹H NMR.

RESULTS

Compared to insulin degludec, IDegLira administration resulted in significantly greater reduction of total and LDL-cholesterol. On the other hand, the effect of the two drugs on apolipoprotein-B-containing lipoprotein subfractions concentration was minimal and did not differ between the 2 interventions. IDegLira, but not insulin degludec, induced an atheroprotective shift in HDL's fatty acid composition and particle core depletion in triglycerides.

CONCLUSIONS

IDegLira administration is accompanied by total and LDL-cholesterol reduction, while sdLDL concentration only reduced in patients experiencing triglyceride reduction. IDegLira induced compositional changes of HDL particles. These changes may contribute to the cardioprotective properties of liraglutide.

Effects of Different Ratios of Carbohydrate-Fat in Enteral Nutrition on Metabolic Pattern and Organ Damage in Burned Rats

(1) Background: Nutritional support is one of the most important cornerstones in the management of patients with severe burns, but the carbohydrate-to-fat ratios in burn nutrition therapy remain highly controversial. In this study, we aimed to discuss the effects of different ratios of carbohydrate–fat through enteral nutrition on the metabolic changes and organ damage in burned rats. (2) Methods: Twenty-four burned rats were randomly divided into 5%, 10%, 20% and 30% fat nutritional groups. REE and body weight were measured individually for each rat daily. Then, 75% of REE was given in the first week after burns, and the full dose was given in the second week. Glucose tolerance of the rats was measured on days 1, 3, 7, 10 and 14. Blood biochemistry analysis and organ damage analysis were performed after 7 and 14 days of nutritional therapy, and nuclear magnetic resonance (NMR) and insulin content analysis were performed after 14 days. (3) Results: NMR spectra showed significant differences of glucose, lipid and amino acid metabolic pathways. The energy expenditure increased, and body weight decreased significantly after burn injury, with larger change in the 20%, 5% and 30% fat groups, and minimal change in the 10% fat group. The obvious changes in the level of plasma protein, glucose, lipids and insulin, as well as the organ damage, were in the 30%, 20% and 5% fat groups. In relative terms, the 10% fat group showed the least variation and was closest to normal group. (4) Conclusion: Lower fat intake is beneficial to maintaining metabolic stability and lessening organ damage after burns, but percentage of fat supply should not be less than 10% in burned rats.

Maternal early-pregnancy body mass index-associated metabolomic component and mental and behavioral disorders in children

Maternal pre-pregnancy obesity and/or higher body mass index (BMI) have been associated with neurodevelopmental and mental health adversities in children. While maternal metabolomic perturbations during pregnancy may underpin these associations, the existing evidence is limited to studying individual metabolites, not capturing metabolic variation specific to maternal BMI, and not accounting for the correlated nature of the metabolomic measures. By using multivariate supervised analytical methods, we first identified maternal early-pregnancy BMI-associated metabolomic component during pregnancy. We then examined whether this component was associated with mental and behavioral disorders in children, improved the prediction of the child outcomes over maternal BMI, and what proportion of the effect of maternal BMI on the child outcomes this component mediated. Early-pregnancy BMI of 425 mothers participating in the PREDO study was extracted from the national Medical Birth Register. During pregnancy, mothers donated up to three blood samples, from which a targeted panel of 68 metabolites were measured. Mental and behavioral disorders in children followed-up from birth until 8.4-12.8 years came from the Care Register for Health Care. Of the 68 metabolites averaged across the three sampling points, 43 associated significantly with maternal early-pregnancy BMI yielding a maternal early-pregnancy BMI-associated metabolomic component (total variance explained, 55.4%; predictive ability, 52.0%). This metabolomic component was significantly associated with higher hazard of any mental and behavioral disorder [HR 1.45, 95%CI(1.15, 1.84)] and relative risk of having a higher number of co-morbid disorders [RR 1.43, 95%CI(1.12, 1.69)] in children. It improved the goodness-of-model-fit over maternal BMI by 37.7-65.6%, and hence the predictive significance of the model, and mediated 60.8-75.8% of the effect of maternal BMI on the child outcomes. Maternal BMI-related metabolomic perturbations during pregnancy are associated with a higher risk of mental and behavioral disorders in children. These findings may allow identifying metabolomic targets for personalized interventions.

Towards a biochemical approach to occupational stress management

Given the immense and growing cost of occupational stress to society through lost productivity and the burden to healthcare systems, current best practices for detecting, managing and reducing stress in the workplace are clearly sub-optimal and substantially better methods are required. Subjective, self-reported psychology and psychiatry-based instruments are prone to biases whereas current objective, biology-based measures produce conflicting results and are far from reliable. A multivariate approach to occupational stress research is required that reflects the broad, coordinated, physiological response to demands placed on the body by exposure to diverse occupational stressors. A literature review was conducted to determine the extent of application of the emerging multivariate technology of metabolomics to occupational stress research. Of 170 articles meeting the search criteria, three were identified that specifically studied occupational stressors using metabolomics. A further ten studies were not specifically occupational or were of indirect or peripheral relevance. The occupational studies, although limited in number highlight the technological challenges associated with the application of metabolomics to investigate occupational stress. They also demonstrate the utility to evaluate stress more comprehensively than univariate biomarker studies. The potential of this multivariate approach to enhance our understanding of occupational stress has yet to be established. This will require more studies with broader analytical coverage of the metabolome, longitudinal sampling, combination with experience sampling methods and comparison with psychometric models of occupational stress. Progress will likely involve combining multi-omic data into a holistic, systems biology approach to detecting, managing and reducing occupational stress and optimizing workplace performance.

Metabolomic Approaches to Study Chemical Exposure-Related Metabolism Alterations in Mammalian Cell Cultures

Biological organisms are constantly exposed to an immense repertoire of molecules that cover environmental or food-derived molecules and drugs, triggering a continuous flow of stimuli-dependent adaptations. The diversity of these chemicals as well as their concentrations contribute to the multiplicity of induced effects, including activation, stimulation, or inhibition of physiological processes and toxicity. Metabolism, as the foremost phenotype and manifestation of life, has proven to be immensely sensitive and highly adaptive to chemical stimuli. Therefore, studying the effect of endo- or xenobiotics over cellular metabolism delivers valuable knowledge to apprehend potential cellular activity of individual molecules and evaluate their acute or chronic benefits and toxicity. The development of modern metabolomics technologies such as mass spectrometry or nuclear magnetic resonance spectroscopy now offers unprecedented solutions for the rapid and efficient determination of metabolic profiles of cells and more complex biological systems. Combined with the availability of well-established cell culture techniques, these analytical methods appear perfectly suited to determine the biological activity and estimate the positive and negative effects of chemicals in a variety of cell types and models, even at hardly detectable concentrations. Metabolic phenotypes can be estimated from studying intracellular metabolites at homeostasis in vivo, while in vitro cell cultures provide additional access to metabolites exchanged with growth media. This article discusses analytical solutions available for metabolic phenotyping of cell culture metabolism as well as the general metabolomics workflow suitable for testing the biological activity of molecular compounds. We emphasize how metabolic profiling of cell supernatants and intracellular extracts can deliver valuable and complementary insights for evaluating the effects of xenobiotics on cellular metabolism. We note that the concepts and methods discussed primarily for xenobiotics exposure are widely applicable to drug testing in general, including endobiotics that cover active metabolites, nutrients, peptides and proteins, cytokines, hormones, vitamins, etc.

Proton Nuclear Magnetic Resonance Metabolomics Corroborates Serine Hydroxymethyltransferase as the Primary Target of 2-Aminoacrylate in a ridA Mutant of Salmonella enterica

The accumulation of the reactive enamine intermediate 2-aminoacrylate (2AA) elicits global metabolic stress in many prokaryotes and eukaryotes by simultaneously damaging multiple pyridoxal 5′-phosphate (PLP)-dependent enzymes. This work employed ¹H NMR to expand our understanding of the consequence(s) of 2AA stress on metabolite pools and effectively identify the metabolic changes stemming from one damaged target: GlyA. This study shows that nutrient supplementation during ¹H NMR metabolomics experiments can disentangle complex metabolic outcomes stemming from a general metabolic stress. Metabolomics shows great potential to complement classical reductionist approaches to cost-effectively accelerate the rate of progress in expanding our global understanding of metabolic network structure and physiology. To that end, this work demonstrates the utility in implementing nutrient supplementation and genetic perturbation into metabolomics workflows as a means to connect metabolic outputs to physiological phenomena and establish causal relationships.

The reactive intermediate deaminase RidA (EC 3.5.99.10) is conserved across all domains of life and deaminates reactive enamine species. When Salmonella entericaridA mutants are grown in minimal medium, 2-aminoacrylate (2AA) accumulates, damages several pyridoxal 5′-phosphate (PLP)-dependent enzymes, and elicits an observable growth defect. Genetic studies suggested that damage to serine hydroxymethyltransferase (GlyA), and the resultant depletion of 5,10-methelenetetrahydrofolate (5,10-mTHF), was responsible for the observed growth defect. However, the downstream metabolic consequence from GlyA damage by 2AA remains relatively unexplored. This study sought to use untargeted proton nuclear magnetic resonance (¹H NMR) metabolomics to determine whether the metabolic state of an S. entericaridA mutant was accurately reflected by characterizing growth phenotypes. The data supported the conclusion that metabolic changes in a ridA mutant were due to the IlvA-dependent generation of 2AA, and that the majority of these changes were a consequence of damage to GlyA. While many of the metabolic differences for a ridA mutant could be explained, changes in some metabolites were not easily modeled, suggesting that additional levels of metabolic complexity remain to be unraveled.

IMPORTANCE The accumulation of the reactive enamine intermediate 2-aminoacrylate (2AA) elicits global metabolic stress in many prokaryotes and eukaryotes by simultaneously damaging multiple pyridoxal 5′-phosphate (PLP)-dependent enzymes. This work employed ¹H NMR to expand our understanding of the consequence(s) of 2AA stress on metabolite pools and effectively identify the metabolic changes stemming from one damaged target: GlyA. This study shows that nutrient supplementation during ¹H NMR metabolomics experiments can disentangle complex metabolic outcomes stemming from a general metabolic stress. Metabolomics shows great potential to complement classical reductionist approaches to cost-effectively accelerate the rate of progress in expanding our global understanding of metabolic network structure and physiology. To that end, this work demonstrates the utility in implementing nutrient supplementation and genetic perturbation into metabolomics workflows as a means to connect metabolic outputs to physiological phenomena and establish causal relationships.