Identification and Quantification of Deranged Metabolites in Critically Ill Patients Using NMR-Based Metabolomics

Summary

Nuclear magnetic resonance (NMR) spectroscopy is used to identify dysregulation in the metabolites in patients with various diseases. This technique allows the quantification of the deranged metabolites, unraveling the pathophysiological insights. Here, we describe the step-by-step procedure of the NMR-based approach for the metabolic characterization of the patients.

Abstract

Metabolomics is emerging as a significant approach to reflect the individual's response to pathophysiological conditions. Nuclear magnetic resonance (NMR) spectroscopy has evolved as a tool to identify metabolic dysregulations in critically ill patients afflicted with conditions like acute respiratory distress syndrome (ARDS), severe acute pancreatitis (SAP), acute kidney injury (AKI), and sepsis. The spectral data from the serum sample of the study and control group are recorded using an 800 MHz NMR spectrometer and processed using NMR processing and analysis tools. Furthermore, a rigorous statistical analysis, such as univariate and multivariate tests, is performed to pinpoint significant metabolites, which are then accurately identified and quantified using NMR metabolite quantification software. Additionally, pathway analysis highlights the deranged biochemical cycles that result in the severity of illness. Through this comprehensive approach, researchers aim to gain deeper insights into the metabolic alterations associated with these critical illnesses, potentially paving the way for a better understanding of the disease and improved diagnostics and treatment strategies.

Introduction

Despite continuous efforts to provide efficient disease diagnosis worldwide, targeted therapy has still not achieved its true potential. Various approaches, such as transcriptomics, proteomics, etc., have resulted in the identification of several biomarkers, but these did not hold enough clinical utility because of the lack of sensitivity and specificity^1,2. Targeted therapy is a big challenge in some multifactorial diseases, eventually leading to higher mortality. There is a need for a better understanding of the underlying mechanism and pathophysiology of complex diseases existing with wide heterogeneity. So, in this regard, the evolvement of metabolomics has revolutionized therapeutic development, which eventually may aid in tailoring the treatment regimens for various critical illnesses such as acute respiratory distress syndrome (ARDS), sepsis, and severe acute pancreatitis (SAP).

Metabolomics is a comprehensive approach aimed at identifying and quantifying small molecular weight molecules (metabolites such as amino acids, lipids, peptides, organic acids, and vitamins) across various biofluids, cells, or tissue extracts. These metabolites, which typically weigh less than 1500 Da, play active roles in biochemical processes, reflecting a progressive outline of the organism's biological state. They include substrates for key enzymatic processes, intermediates in biological pathways, and by-products of cellular metabolism. Consequently, metabolomics captures a detailed fingerprint of dietary influences, drug interactions, and disease states. Metabolite changes are highly sensitive indicators of metabolism and biological pathways, allowing for correlations with phenotypic expressions and resultant pathophysiological abnormalities^3,4. Initial variations in metabolites can serve as early indicators of disease severity, while temporal changes may help in monitoring treatment efficacy, disease progression, and clinical outcomes^5,6,7. Metabolomics thus enhances clinical assays and various other omics approaches by redefining diseases through clinical, physiological, and biochemical endpoints^{8,9,10,11,12,13}. The analytical capabilities of metabolomics are employed to monitor and determine disease susceptibility through altered metabolite concentrations^14,15.

In this context, both mass spectrometry (MS) and nuclear magnetic resonance (NMR) spectroscopy have emerged as the primary analytical platforms for metabolite profiling in biological samples. These methods are used for both targeted and untargeted identification and quantification of metabolites^16,17,18. Each platform has its advantages and limitations, but the non-destructive nature of NMR makes it preferable in various in vivo studies and for the characterization of the structure of unknown compounds, particularly in the initial stages of metabolomics research. The sample fractionation, derivatization, and ionization required before MS can introduce biases and often result in sample loss, affecting the dynamic features that NMR spectroscopy can capture with minimal or no sample preparation. The primary limitation of NMR is its lower sensitivity compared to MS, which offers a lower limit of detection, making it challenging to detect less abundant metabolites¹⁹. However, advancements such as high-resolution superconducting magnets, cryogenically cooled NMR probes, and techniques that enhance sensitivity have mitigated this limitation^20,21,22. As a complementary approach to genomics and proteomics, metabolic profiling using NMR spectroscopy is gaining traction as a preferred technique^23,24,25. NMR's minimal sample preparation, reproducibility, and repeatability make it a valuable tool for capturing the inherent dynamic features of metabolites despite its sensitivity challenges²⁶.

Several research groups have performed metabolomics successfully, pinpointing the dysregulated metabolic profile of patients for various diseases²⁷ such as ARDS^{28,29,30,31,32,33}, pneumonia³⁴, sepsis⁷, gallstones³⁵, and pancreatitis³⁶. NMR-based metabolomics studies of critically ill patients have been instrumental in tracking the progression from systemic inflammatory response syndrome (SIRS) to multiple organ dysfunction syndrome (MODS), which is a leading cause of intensive care unit (ICU) mortality and morbidity³⁷. In a study by Stringer et al., plasma samples were used to examine metabolic alterations in patients with sepsis-induced acute lung injury (ALI) compared to control patients³⁸. Key metabolites found elevated in this pilot study reflected the metabolic pathways involved and their association with clinical scores. This research was extended to serum metabolomics to differentiate sepsis from the early stages of lung injury mechanisms in ARDS³². Additionally, another study in ARDS identified potent serum biomarkers that distinctly differentiate between acute lung injury/ARDS and healthy controls, offering insights into systemic metabolic changes corresponding to the acute onset of lung injury^39,40.

NMR spectroscopy is a high-throughput and automated analytical technique that delivers robust and unbiased information about the metabolic fingerprint revealing the underlying pathophysiology³⁸. The clinical application and biological interpretation of NMR data hinge on obtaining high-quality spectra that contain rich information. Therefore, ensuring accurate, uniform, and well-formulated data collection, processing, and analysis is crucial. Therefore, the objective of this study is to leverage the essential steps of NMR-based metabolomics for the identification and quantification of metabolites. This study highlights the key steps of the protocol required for clinical metabolomics study (Figure 1), such as the selection of appropriate samples, collection and storage, sample processing and preparation, data acquisition and analysis, identifying and quantifying the metabolite of interest, and eventually, interpretation of the results in a clinical context to derive relevant insights. Each of these steps is essential for leveraging NMR spectroscopy in metabolomics to uncover significant biological and clinical insights.

Protocol

Ethical approval (IEC code: 2022-71-PhD-126) was obtained from the IEC of Sanjay Gandhi Postgraduate Institute of Medical Sciences (SGPGIMS), Lucknow. Written and informed consents were taken from the patients or their relatives to perform the study and publish the data for research purpose. Moreover, the research was conducted following the institutional guidelines.

1. Study design and ethical clearance

1. Determine the sample size for each group. Carefully review and establish the selection criteria for participants. Additionally, if the study design requires the inclusion of human or animal samples, then obtain ethical clearance from the appropriate institutional ethical committee (IEC) before beginning sample collection.
2. Select ARDS patients categorized as mild, moderate, and severe, based on the diagnostic criteria of the Berlin 2012 definition based on the partial pressure of oxygen in arterial blood (PaO2) to the fraction of inspired oxygen (FiO2) (P/F) ratio. The ARDS patients with a P/F range of 300-200 are regarded as mild, 200-100 as moderate, and below 100 as severe³⁸. Here, we have focused on two groups: mild and moderate ARDS patients (see Table 1).
   NOTE: The study involves the use of serum obtained from blood samples to perform NMR-based metabolomics. In this study, the participants included for reference had a mean age of 42 years (± 10.9 years), with a gender distribution of 2 males and 6 females.

Category	P/F ratio
Mild	300-200
Moderate	200-100
Severe	100-0

Table 1: ARDS patients' categorization.

2. Sample selection, collection, and processing

1. Perform serum isolation by following the steps described below.
   1. For experimenting, collect 2 mL of blood samples from the arteries using a sterile needle in a plain (with no additives) and sterile vial. Leave the sample to clot for 30 min at room temperature. For this study, 4 patients diagnosed with mild ARDS, and 4 patients having moderate ARDS were selected.
   2. Centrifuge the sample at 3100 x g for 15 min to separate the serum²⁵. Transfer the serum into sterile microcentrifuge tubes, making aliquots of smaller volume (i.e., 400-500 µL), appropriately label them (Patient name, CR no., day point, etc.), and store them at -80 °C for future analysis.
      NOTE: At this point, the experiment can be paused and resumed at a convenient time.
2. Process the serum sample before NMR experiments using the following steps.
   1. Thaw the sample before performing the NMR experiments. Mix 250 µL of aliquot of the serum with 250 µL of saline phosphate buffer solution (containing 100% D₂O, 0.9% NaCl, 50 mM sodium phosphate buffer, pH 7.4) to minimize pH variation.
   2. Vortex the mixture for a few seconds (~15 s) to ensure it is homogenously mixed. Finally, transfer the prepared sample to a clean NMR tube with a coaxial insert containing trimethylsilyl propionate (TSP) at a concentration of 0.05 mM, which serves as an external standard for calibration and quantification⁴¹.
      NOTE: For reference, an inert and non-reactive compound that does not interact with the analytes is used. Other commonly used reference compounds include Tetramethylsilane (TMS) and sodium trimethylsilylpropanesulfonate (DSS).

3. NMR experimentation

NOTE: The focus is on identifying small molecules in the serum samples, therefore, we used the Carr-Purcell-Meiboom-Gill (CPMG) pulse sequence, which suppresses macromolecule signals. All the serum samples in this study were recorded using an 800 MHz NMR spectrometer equipped with a cryogenically cooled triple-resonance TCI 5 mm broadband inverse probe-head and shielded z-gradient.

1. Place the sample into the magnet. Write the command wrpa (mention experiment number) and press enter to set up a proton experiment with the cpmg pulse program. Mention the patient's name and other details necessary by clicking on the option Title.
2. Lock the magnetic field by Writing the command Lock, press enter, and select the option 90% H₂O and 10% D₂O further. Tune and match the probe manually by ATMM to optimize the efficiency of radio frequency pulses and maximize sensitivity. Carry out 1D gradient shimming by clicking on topshim.
3. Set the following acquisition parameters, including a relaxation delay of 5 s, a spectral sweep width of 12 parts per million (ppm), and an echo time of 300 µs. All these parameters can be easily modified by selecting the option Acquisition Parameters in the upper panel.
4. Apply line broadening of 0.3 Hz using an exponential window function, collecting data into 64,000 points over 128 scans. These parameters can be similarly modified by selecting the option Acquisition Parameters in the upper panel. By clicking on Acquisition Parameters, a list of specifications will appear, and the mentioned attributes, along with other parameters, can be modified during optimization.
5. Acquire the NMR spectra using NMR processing and analysis tools. Once the final spectra are obtained, write command apk and absn and press enter to perform phase correction and baseline correction, respectively. Click on the Calibrate Axis option present on the upper panel and then either calibrate the TSP peak to 0 ppm automatically or manually.
6. Transfer the acquired data from the system to the workstation, where further processing and analysis are done.
   NOTE: The experimental phase is now complete and can be paused. Data preprocessing, analysis, and interpretation can be performed conveniently.

4. Data preprocessing

1. Conduct the data preprocessing using NMR processing and analysis tools and NMR metabolite quantification software.
   1. Sometimes, only automatic correction is not enough, and in that case, manual correction is also necessary. To perform phase correction manually, click on the option Process on the top menu bar of NMR processing and analysis tools. Then click on Adjust Phase, drag the mouse, and observe until the spectra are phase corrected. Once done, click on the Save and Return option (present in the form of an icon) available in the same menu bar.
   2. To perform manual baseline (Whittaker method) correction, open the spectral file in the processor module of the NMR metabolite quantification software. Select the option Baseline Correction. Then select the option Whittaker Method. Now put the dots at the base of the peaks in the entire spectra (0.0 to 10 ppm usually in the case of serum or as per the availability of the metabolite in the spectra) to adjust the base. After all the spectral files have been corrected baseline, save them in a single folder. Further processing, either using the processor or the profiler module of the NMR metabolite quantification software, can be performed using the same file saved in the specific format.
   3. Perform binning using the profiler module of the same software. Select the option Tools, then Batch Process, and then Spectral Binning.
2. To generate the final binning sheet to be used for statistical analysis, select the folder containing all the baseline-corrected spectral files after clicking on Spectral binning.
3. Divide the spectra into a set number of equal-sized bins with a defined spectral width, ranging from 0.01 to 0.04 ppm (can be optimized as per the study). This process simplifies and organizes the spectral data, facilitating more consistent analysis. Specify the bucket size along with the start and end ppm values and designate the folder where the output binning sheets will be saved.
4. Exclude the chemical shift regions corresponding to water and the solvent TSP (4.8-5.2 and 0.0-0.7 ppm) to prevent spectral interference^42,43,44.
5. This spreadsheet contains the sample names in one row and the different bin values along with ppm in the other rows. Before using this sheet for statistical analysis, modify it by organizing the samples into groups for analysis and saving it in comma-delimited (CSV) format only.
   NOTE: For instance, when comparing the mild group with the moderate group of ARDS patients, we arranged the data with mild and moderate labels, added a serial number, and saved the file in CSV format.

5. Statistical analysis

NOTE: The binning sheet obtained from the previous step serves as the input file for statistical analysis. In this study, the statistical analysis (one factor) module of metabolomics statistical analysis software was used.

1. Before statistical analysis, perform the normalization step using sum, log, and Pareto but it can vary across different studies⁴⁵.
2. Perform univariate and multivariate analysis to distinguish the different groups and determine the accuracy of the identified metabolites.
   1. Perform principal component analysis (PCA), partial least-square discriminant analysis (PLS-DA), and orthogonal partial least-square discriminant analysis (OPLS-DA) to observe discrimination between the groups yielding variable importance of projection (VIP) scores for every metabolite responsible for the difference in the metabolic profile.
   2. Perform univariate methods, such as the student t-test and ANOVA, to identify the significant metabolites based on the criteria of having a p-value less than 0.05. Both tests yield box whisker plots showing the difference in the relative concentration of the metabolites.
   3. Perform biomarker analysis, which yields an area under the receiver operating characteristic curve (AUROC) graph. The AUC (area under the curve) value > 0.8 is considered significant⁴⁶.
   4. Select the final list of significant metabolites using the criteria VIP score>1, Bonferroni-corrected p-value<0.05⁴⁴, and AUC > 0.8⁴⁶.
3. Identify metabolites as described below.
   1. Binning provides significant metabolites in the form of ppm values rather than names. Identify the metabolites corresponding to specific ppm values using databases like the Biological Magnetic Resonance Bank (BMRB)⁴⁷and the Human Metabolome Database (HMDB).
   2. Other than these, use previously published literature^42,44 for identification and validation.
   3. Use the profiler module of NMR metabolite quantification software to confirm the identified metabolites with their ppm values.

6. Quantification of metabolites using NMR metabolite quantification software

NOTE: The profiler module of the software is used widely to quantify the identified metabolites.

1. Before beginning the quantification, calibrate the reference compound selected for the study (tsp used in this study) using the processor module. Click on the option Calibrate CSI (chemical shape indicator). Mention the concentration of the reference compound (0.05 mM for this study). Then, drag the software peak onto the spectral reference peak and adjust its height and width properly. Once the spectral reference peaks fit the software's peak, click on Accept.
2. Open the desired spectra in the profiler module and select the appropriate compound library for study.
3. At the bottom of the screen, a list of metabolites is displayed. Select the metabolite of interest from this list. Once selected, the possible ppm of the chosen metabolite will appear in the top corner of the spectra.
4. Select the Ppm Value at the top corner and zoom in on the spectra to the specific ppm scale of the metabolite.
5. At this point, two peaks will appear on the screen: one representing the recorded data and the other, displayed as a dotted line, representing the software peak. Align both peaks. Drag the software peak and adjust its height and position to match the peak present in the sample.
6. Once properly aligned, the concentration of the metabolite can be obtained from the value displayed under the heading concentration in mM for that metabolite (Figure 1).
7. Follow the same procedure for all metabolites and samples and export the quantified values in a spreadsheet. To generate the output file, click on the Batch Operations option under the Tools menu. Specify the desired folder for saving the output file.

7. Pathway analysis

NOTE: The significant metabolites identified after analysis are utilized to determine the major pathways directly influencing the outcome in disease groups. The pathway analysis module of metabolomics statistical analysis software and the KEGG database are generally employed for this purpose.

1. Input the obtained list of significant metabolites into the module. A list containing the names of pathways is generated.
2. Select the final pathways based on a criterion of having an impact factor greater than 0.1. Significant impact values indicate disease-related pathways³⁰.
3. Analyze the resulting pathways thoroughly to elucidate the dysregulation associated with the severity of the disease. One of the demerits of this method is its inefficiency in identifying the dysregulated pathways when the significant metabolites are less than 3.
4. To cope with the same, the identified metabolite and its correlation with the disease severity can also be determined by using the previously published literature. Perform a thorough literature survey and observe if the identified metabolite is also following the same trend as per your interest.

Figure 1: Fundamental steps in NMR-based metabolomics. The figure presents key steps in NMR-based metabolomics: collecting and preparing samples, performing NMR spectroscopy, preprocessing data, conducting statistical analysis, identifying and quantifying metabolites, and interpreting biological significance. Please click here to view a larger version of this figure.

Results

To conduct a metabolomics study, it is essential to determine the sample size and the specific groups that will be analyzed. Selecting an adequate sample size is essential for obtaining significant results that accurately correlate with disease severity⁴⁸. However, in this particular work, we used a small sample size to demonstrate the steps involved in the identification and quantification of metabolites using NMR-based metabolomics, which was intended primarily for reference. In this study, we examined the metabolic differences between patients in the mild and moderate stages of ARDS on day 1 of their ICU admission, including a total of 8 subjects. Following the protocol, NMR spectra of all the samples were acquired and processed, and eventually, a spreadsheet was created with the binning results.

We then performed a rigorous statistical analysis, where the binning sheets were used as input files, normalizing the data using sum, log, and Pareto. We conducted PCA (Figure 2A) and OPLS-DA (Figure 2B), where the former showed clustering, but the latter revealed clear discrimination between the groups and highlighted metabolic differences. Permutation test statistical analysis (20x) yielded values of R2Y=0.99 and Q2=0.47. Subsequently, we performed a student's t-test (Figure 2C) and biomarker analysis (Figure 2D) to identify dysregulated metabolites. Following the criteria outlined in the protocol section, we filtered the results to the significant metabolites (Table 2).

Figure 2: Multivariate and univariate analysis. (A) Principal component analysis (PCA) and (B) Orthogonal partial least-square discriminant analysis (OPLS-DA) of mild ARDS patients (represented by red color) and moderate ARDS patients (represented by green color). (C) The box-cum-whisker plots highlightthe differences in the relative concentration of metabolites between mild and moderate ARDS patients detected by proton nuclear magnetic resonance spectroscopy in serum samples. The x-axis represents the study groups, and the y-axis shows the relative concentration when mild ARDS patients (represented by red color) and moderate ARDS patients (represented by green color) were compared on day 1 of ICU admission. (D) The area under the curve (AUC) plot of significant metabolites predicting severity obtained after performing biomarker analysis comparing mild and moderate ARDS patients on day 1 of ICU admission. Please click here to view a larger version of this figure.

S.No.	Name	Spectral location (ppm)	p-value	VIP	AUC value
1	Lactate	1.34	0.04	1.99	1
2	Isoleucine	1.42	0.03	2.01	0.875

Table 2: List of significant metabolites obtained after comparing mild and moderate ARDS patients on day 1 of ICU admission.

Since the results were obtained in ppm values, we assigned metabolites using previous literature and validated them through the profiler. After the assignment, lactate and isoleucine were obtained as significant metabolites, which have also been reported in some previous ARDS studies^30,40. We ensured the selection of well-resolved, non-overlapping peaks. After identification, we filtered the list of the significant metabolites using the criteria mentioned in the protocol. We then quantified the metabolites of interest using the profiler module by opening the spectra, searching for the metabolite of interest, and fitting the software peak to the peak obtained in the spectra. This provided the quantified values of the metabolites in mM. A spreadsheet containing sample names, metabolite names, and concentrations was generated for reference. After quantification with the software, validation through other approaches, such as proteomics, can be performed.

Furthermore, the identified metabolites are analyzed thoroughly to determine their correlation with disease severity and to identify the biochemical pathways in which these metabolites are actively involved. Since only two significant metabolites were identified after using the criteria mentioned in the material and method section, pathway analysis using the software was not possible. Therefore, a thorough literature survey was done to highlight the deranged metabolic cycles. The two metabolites identified in this study successfully correlated with the severity of ARDS. Increased lactate denoted lung inflammation and anaerobic respiration, whereas increased isoleucine indicated protein catabolism associated with lung injury, infection, and energy metabolism⁴⁰. Understanding these metabolic pathways serves a dual purpose. Firstly, it reveals the complexity of the disease, providing insights into its biological mechanisms. Secondly, it aids therapeutic advancements by guiding clinicians in tailoring treatments for patients, ultimately improving clinical outcomes.

Discussion

Metabolomics efficiently identifies and quantifies metabolites, targeting the metabolic cycles that become deranged during disease. The quality of the results depends on the meticulous execution of each step in the metabolomics approach. Every stage, from sample selection and collection to pathway identification, is critical in accurately identifying the primary factors contributing to the disease. Before performing metabolomics, a thorough review of the literature is essential, and careful attention must be paid at every stage.

In designing a metabolomics study, the selection of a sample is crucial for extracting significant information and ensuring ideal sample preparation, whether for targeted or untargeted metabolic fingerprinting of the clinical condition in consideration. Proper initial sample processing is crucial, and protocols should be consistently followed for all samples throughout the study⁴⁹. Extra caution is needed when working with clinical samples. Commonly used samples for NMR-based metabolomics include biofluids, cell extracts, and tissue extracts^50,51. Cerebrospinal fluid (CSF) and urine often require minimal pre-treatment and are frequently studied in contexts such as kidney-related ailments, neurodegenerative diseases, and drug toxicity. Whole blood, plasma, and serum provide comprehensive information for a broad spectrum of critical diseases and are preferred because of ease of collection. A wide range of other biofluids, such as seminal fluid, saliva, bile, dialysis fluid, amniotic fluid, exhaled breath condensate, lung aspirates, synovial fluid, and mini bronchoalveolar lavage fluid (mBALF), have been extensively used for their specificity and accuracy using NMR^52,53,54. Each biofluid presents its advantages and challenges concerning collection, availability, and informational content, depending on the study or experimental design⁵⁵. We have focused on serum samples of ARDS patients for this study as blood samples typically offer information at a whole-body level and are not significantly influenced by extreme daily variations. On the other hand, urine samples are subject to various fluctuations and can be impacted by diet, lifestyle, medications, and environmental factors²⁷. Proper sample storage is also important to avoid enzymatic activity or microbial degradation. Samples are typically cryopreserved at -80 °C with minimal freeze-thaw cycles to prevent metabolic breakdown and maintain the stability of the sample.The sample can be stored at -80 °C for extended periods, allowing NMR experimentation to be conducted conveniently. However, planning and performing the experiment as soon as possible is advisable for optimal results.

In this study, we employed an 800 MHz NMR spectrometer for data acquisition, which provides high-resolution spectral data. For this experiment, we adhered to the same acquisition parameters mentioned above. However, it is important to note that the same acquisition parameters outlined in this protocol can be applied using NMR spectrometers operating at other magnetic field strengths. The number of scans can be increased when operating at a lower magnetic field to improve resolution. While higher field strengths may offer enhanced resolution, the core aspects of the acquisition process, such as pulse sequences, relaxation delays, and temperature control, remain consistent across different instruments, ensuring the reproducibility and reliability of the results.

Following a standardized procedure is essential to maintain reproducibility and impartiality, as deviations can hinder subsequent data analysis and evaluation. Each step of the NMR experimentation must be performed meticulously, as the quality of the spectra depends on them. Special attention should be given to shimming, as it defines peak shape, and acquisition parameters should be optimized according to the sample and study design. For samples with low metabolic concentration in the sample, the number of scans can be increased. Data acquisition in NMR involves standard optimization of both 1D and 2D experiments tailored to the choice of sample and the specific information required. Various NMR experiments have been optimized to obtain high-quality spectra from biofluids, minimizing the impact of unwanted macromolecules. The CPMG pulse sequence⁵⁶ is preferred for samples like serum, plasma, or cerebrospinal fluid. This technique suppresses the broad signals from macromolecules such as lipids and proteins, which have long transverse relaxation times and can obscure the resonances of low molecular weight metabolites.

Data pre-processing encompasses the intermediate methods used to ensure uniform and homogeneous data analysis and its interpretation. This process includes avoiding undesirable signals or spectral regions prone to inconsistencies and correcting peak shifts. In biofluids, excluding unwanted regions often means removing signals, such as those from water, that dominate the 4.6 ppm to 5 ppm range. Peak shifts, caused by variations in pH, temperature, salt concentration, dilution factors, and ionic composition, can be minimized by using buffers at a homeostatic pH. Additionally, the use of standard reference compounds, peak alignment, and binning can help reduce peak shifts. Peak alignment is crucial to correlate patterns across different spectra acquired under identical conditions.

Spectra processing must be done properly to ensure accurate binning later in the analysis. All spectra should be calibrated and aligned uniformly, and base and phase corrections can be performed manually or automatically. Proper phase and base correction, as well as calibration, are critical. Performing baseline correction is essential to avoid distortions that can affect the identification and quantification of metabolites⁵⁷. Phase correction is necessary to ensure no distorted or inverted peaks can affect the interpretation of results. Carefully decide on the criteria for selecting significant metabolites and filter the results accordingly. Since binning provides results in the form of ppm values, accurately identify the metabolite's name using various sources mentioned in the protocol. If multiple metabolites are present at the same ppm values, identify them through peak shape.

Before performing the statistical analysis, normalization is done to reduce variations in the dataset, whether induced or not induced, due to differences in magnitude and fluctuations in metabolite concentrations. This process involves various methods of centering, scaling, and transformation, chosen based on the data's distribution and variability. The goal is to minimize unwanted systematic bias while preserving the biological factors of interest. Multivariate methods, including PCA, PLS-DA, and OPLS-DA, are used to establish differential metabolites with VIP values greater than 1. Model predictability and cross-validation are determined using R2 (goodness of fit) and Q2 (goodness of prediction) values, while permutation test statistics validate the model. A Pearson correlation-based heat map assesses model robustness by enumerating upregulated and downregulated metabolites in sub-phenotypes and the resultant endotypes.

Univariate methods, such as the student t-test and ANOVA, like Bonferroni correction or minimization of FDR (false discovery rate) and Benjamini-Hochberg, are used to reduce the probability of false positives. These methods provide a rough measure of potentially important features independent of their correlation within or between molecules and other confounding factors. Significant metabolites are identified based on a p-value less than 0.05, which are then used to substantiate metabolic endotypes in outcome subgroups⁴⁵.

Individual box whisker plots and AUROC are employed to evaluate the specificity and sensitivity of the endotypes in outcome subgroups. The clinical score is combined with the model in AUROC to validate the model's accuracy. The AUC value > 0.8 is considered significant⁴⁶. Various other tests are available online to test the accuracy of the discrimination and efficacy of metabolites identified. Some of them are random forest classification, heat map, correlation, etc.⁴⁵.

Various approaches are utilized to quantify significant metabolites, including peak integration, which can be challenging for beginners due to their advanced nature. Identify characteristic peaks for metabolites by referring to individual chemical shifts for precise identification. Determine the area of these characteristic peaks relative to the area of the reference peak, TSP. To calculate the absolute concentration of a metabolite (Cm), use the following equation:

Cm = Im.nr.Cr/Ir.nm

Where Cr is the concentration of the reference compound (TSP) in the solution, Im and Ir are the integrated peak areas (can be obtained using NMR processing and analysis tool) of the metabolite and reference (TSP), respectively, and nm and nr are the number of protons representing the metabolite and reference peaks, respectively²⁴.

Alternatively, automated NMR metabolite quantification software is used to obtain metabolite concentrations for single or multiple spectra based on the concentration of the reference standard, TSP. Quantification using software simplifies the process and reduces the likelihood of errors compared to other methods. While this research paper highlights the use of software for metabolite quantification, there can be issues if the metabolite of interest is not present in the software library. Additionally, the software is paid, so only researchers with an authorized license can use it.

Metabolomics has proven instrumental in unraveling the complexities of various diseases by identifying dysregulated metabolic cycles contributing to disease severity. However, translating these findings into clinical practice has been challenging. The medical community is now emphasizing a personalized medicine approach to enhance disease outcomes. While identifying dysregulated metabolites was previously sufficient, current efforts focus on precise quantification to define disease-specific ranges. This information is invaluable for clinicians in customizing treatment strategies. The advancements in quantification techniques have facilitated research efforts, with numerous research groups worldwide successfully employing these approaches using specialized software.

This progress underscores the potential of metabolomics to significantly impact personalized medicine and improve clinical outcomes. Moreover, the integration of metabolomics data with other omics data, such as genomics and proteomics, promises to provide a more comprehensive understanding of disease mechanisms and further refine therapeutic interventions. As technology continues to evolve, the precision and utility of metabolomics in clinical settings are expected to increase, paving the way for more effective and personalized healthcare solutions.

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Centirfuge	Sigma aldrich	3-18KS
Chenomx NMR suite	NMR Suite, v9, Chenomx Inc., Edmonton, Canada		NMR metabolite quantification software
Co-axial insert	Sigma aldrich	Z278513
Deuterim oxide	Sigma aldrich	151882
Eppendorf tubes	Tarsons	500020
Metaboanalyst	Wishart Research Group		Metabolomics statistical analysis software
NMR tube	Wilmad	Z412007	5mm diameter
Pipette	Eppendorf research plus	3123000039	0-100 μl
Sample collection vials	Tarsons cryo chill vials	523194
Sodium azide	Sigma aldrich	S2002
Sodium chloride crystal	Sigma aldrich	S9625
Sodium phosphate dibasic	Sigma aldrich	567550
Sodium phosphate monobasic	Sigma aldrich	S0751
Topspin 3.6.4	Bruker		NMR processing and analysis tool
Tsp salt	Sigma aldrich	269913