Non-Targeted Metabolomics for Food Authentication: A Comprehensive Guide for Researchers and Scientists

Abstract

This article provides a comprehensive overview of the application of non-targeted metabolomics in food authentication, a critical field for ensuring food safety, quality, and traceability. Aimed at researchers, scientists, and professionals in drug development and food science, it covers the foundational principles of using metabolic fingerprints to combat food fraud, including origin misrepresentation, species substitution, and adulteration. The scope extends from core concepts and analytical methodologies—highlighting advances in mass spectrometry (MS) and nuclear magnetic resonance (NMR)—to the application of machine learning for data analysis. It further addresses the critical challenges of method validation, harmonization, and quality assurance, while comparing metabolomics with other omics technologies. The goal is to equip professionals with the knowledge to develop, optimize, and implement robust, non-targeted methods for authenticating food in complex global supply chains.

The Principles and Scope of Food Metabolomics in Authenticity

Defining Food Authentication and Its Global Challenges

Food authentication represents a critical frontier in analytical chemistry and food science, employing advanced analytical techniques to verify food product integrity, composition, and origin within increasingly complex global supply chains. This application note examines the foundational principles, methodological approaches, and significant challenges in food authentication, with particular emphasis on the emerging role of non-targeted metabolomics. We provide detailed experimental protocols for metabolite profiling, comprehensive data analysis workflows, and specialized reagent solutions to support implementation in research settings. The escalating economic and public health impacts of food fraud—estimated at $10-15 billion annually—underscore the urgent need for robust, high-throughput authentication technologies that can detect sophisticated adulteration practices and mislabeling across diverse food matrices [1].

Food authentication encompasses the analytical procedures and regulatory frameworks designed to verify that food products conform to their label descriptions regarding composition, origin, processing methods, and quality attributes. It addresses deliberate misrepresentation for economic gain, commonly termed "food fraud," which includes practices such as adulteration (adding unauthorized substances), substitution (replacing valuable ingredients with inferior alternatives), and mislabeling (providing false geographic, species, or quality information) [2] [1]. The fundamental objective of authentication is to protect consumers from health risks, ensure fair trade practices, and maintain trust in food supply chains.

The global significance of food authentication has intensified due to several converging factors: the expansion and complexity of international food supply networks, increasing consumer awareness and demand for premium products with specific attributes (e.g., organic, geographic origin, traditional production), and the escalating economic incentives for fraudulent activities [2] [3]. High-value products such as extra-virgin olive oil, manuka honey, wine, and seafood consistently rank among the most frequently adulterated commodities, with fraudulent practices becoming increasingly sophisticated and difficult to detect through conventional analytical methods [2] [1]. For instance, global sales of manuka honey reportedly reach 10,000 tonnes annually despite only 1,700 tonnes being produced in New Zealand, indicating widespread misrepresentation in the marketplace [1].

Global Challenges in Food Authentication

Economic and Public Health Impact

Food fraud inflicts substantial economic damage and poses serious public health risks worldwide. The economic burden is staggering, with estimates indicating annual global losses between $10-15 billion across the food industry [1]. Beyond financial impacts, adulteration incidents have led to severe health crises, most notably the 2008 melamine contamination of infant formula in China that resulted in 54,000 hospitalizations and 6 infant deaths [1]. These incidents highlight the critical intersection between economic fraud and food safety emergencies, necessitating robust detection and prevention systems.

Table 1: Documented Food Fraud Incidents and Impacts

Product Category	Type of Fraud	Economic/Health Impact
Infant Formula	Melamine adulteration	54,000 hospitalizations, 6 deaths [1]
Manuka Honey	Mislabeling as premium product	10,000 tonnes sold globally vs. 1,700 tonnes produced [1]
Olive Oil	Adulteration with cheaper oils	9 of 20 Italian brands failed quality verification [1]
Meat Products	Horsemeat in beef products	€300 million market value drop for Tesco [1]
Dried Oregano	Adulteration with other leaves	19 of 78 samples contained 30-70% foreign matter [1]

Technical and Regulatory Complexities

The technical landscape of food authentication presents multifaceted challenges. Food matrices exhibit tremendous chemical complexity, with composition variations arising from natural biological diversity, environmental conditions, and processing methods. This complexity is compounded by the globalization of supply chains, where ingredients may traverse multiple countries and processing stages before reaching consumers, significantly complicating origin verification and traceability efforts [2] [4]. Regulatory frameworks struggle to maintain pace with evolving fraudulent practices, often lagging behind emerging threats due to lengthy policy development and implementation cycles [2]. The absence of uniform international standards and enforcement mechanisms further creates vulnerabilities, particularly for products involving numerous jurisdictions with varying regulatory rigor [2] [4].

Non-Targeted Metabolomics in Food Authentication

Theoretical Foundations

Non-targeted metabolomics has emerged as a powerful approach for food authentication by comprehensively analyzing the small molecule metabolites (typically <1500 Da) present in biological samples. Unlike targeted methods that quantify predefined analytes, non-targeted strategies aim to capture global biochemical profiles, enabling detection of unexpected alterations resulting from adulteration, substitution, or misrepresentation [5] [6]. This methodology is particularly well-suited to authentication because the metabolome provides a sensitive record of a food's biological history, reflecting factors such as geographic origin, botanical variety, agricultural practices, and processing methods [5] [3].

The conceptual framework for applying non-targeted metabolomics to food authentication centers on identifying distinctive chemical patterns or "fingerprints" that are characteristic of authentic products. These patterns may derive from environmentally influenced metabolic pathways (the "terroir" effect), species-specific biochemical processes, or production method signatures [5]. By establishing reference metabolomic profiles for authentic materials, researchers can develop classification models capable of detecting deviations indicative of fraud. This approach has demonstrated particular utility for verifying geographic origin—the most prevalent focus in food authentication research—with successful applications across diverse commodities including wine, rice, olive oil, spices, and honey [5].

Analytical Platforms and Workflows

Liquid chromatography coupled with high-resolution mass spectrometry (LC-HRMS) represents the predominant analytical platform for non-targeted metabolomics in food authentication due to its sensitivity, broad dynamic range, and capability to detect diverse chemical classes without derivatization [6] [7]. Common instrumental configurations include Q-TOF (quadrupole time-of-flight) and Orbitrap mass analyzers, which provide the mass accuracy and resolution necessary for confident compound annotation [6] [7]. Effective non-targeted workflows typically incorporate complementary separation techniques, most frequently combining reversed-phase chromatography (for lipophilic compounds) with hydrophilic interaction liquid chromatography (HILIC) for polar metabolites, thereby expanding metabolome coverage [6].

Figure 1: Non-Targeted Metabolomics Workflow for Food Authentication

Experimental Protocols for Non-Targeted Metabolomics

Sample Preparation and Metabolite Extraction

Principle: Effective metabolite extraction is critical for comprehensive metabolome coverage, requiring optimization to capture chemically diverse compounds while minimizing bias and degradation.

Reagents and Materials:

• LC/MS-grade water, acetonitrile, and methanol
• LC/MS-grade formic acid (99.0+%)
• Ammonium formate
• Stable isotope-labeled internal standards (e.g., l-Phenylalanine-d8, l-Valine-d8)
• Analytical balance (precision 0.1 mg)
• Vortex mixer and ultrasonic bath
• Refrigerated centrifuge
• 1.5 mL microcentrifuge tubes

Procedure:

• Weigh 50 mg (±5 mg) of homogenized food sample into a 1.5 mL microcentrifuge tube.
• Add 800 μL of ice-cold extraction solvent (acetonitrile:methanol:formic acid, 74.9:24.9:0.2, v/v/v) containing internal standards (0.1 μg/mL l-Phenylalanine-d8 and 0.2 μg/mL l-Valine-d8).
• Vortex vigorously for 60 seconds until completely mixed.
• Sonicate for 15 minutes in an ice-cold water bath.
• Centrifuge at 14,000 × g for 10 minutes at 4°C.
• Transfer 600 μL of supernatant to a new LC-MS vial.
• Evaporate to dryness under a gentle nitrogen stream at 30°C.
• Reconstitute in 100 μL of starting mobile phase (for HILIC: 90% acetonitrile with 0.1% formic acid; for RP-LC: 95% water with 0.1% formic acid).
• Centrifuge again at 14,000 × g for 5 minutes before LC-MS analysis [6].

LC-HRMS Analysis for Food Authentication

Principle: This protocol describes HILIC-MS analysis optimized for polar metabolites relevant to food authentication, particularly useful for geographic origin discrimination.

Chromatographic Conditions:

• Column: Waters Atlantis HILIC Silica (150 × 2.1 mm, 3 μm)
• Mobile Phase A: 10 mM ammonium formate with 0.1% formic acid in water
• Mobile Phase B: 0.1% formic acid in acetonitrile
• Gradient Program: 0-2 min: 90% B; 2-15 min: 90%→30% B; 15-18 min: 30% B; 18-18.1 min: 30%→90% B; 18.1-23 min: 90% B (re-equilibration)
• Flow Rate: 0.3 mL/min
• Injection Volume: 5 μL
• Column Temperature: 30°C

Mass Spectrometry Conditions (Orbitrap):

• Ionization Mode: Electrospray ionization (ESI) positive and negative modes
• Spray Voltage: +3.5 kV (positive), -2.8 kV (negative)
• Capillary Temperature: 320°C
• Sheath Gas: 40 arbitrary units
• Auxiliary Gas: 15 arbitrary units
• Scan Range: m/z 70-1050
• Resolution: 70,000 (at m/z 200)
• Data Acquisition: Full MS with fragmentation (data-dependent MS/MS) [6] [7]

Data Processing and Chemometric Analysis

Principle: Transforming raw LC-HRMS data into meaningful authentication models requires specialized computational workflows for feature detection, multivariate statistics, and classification.

Software Tools:

• Feature Detection: Compound Discoverer, XCMS, MS-DIAL
• Statistical Analysis: SIMCA-P, MetaboAnalyst, R packages
• Machine Learning: Python scikit-learn, KNIME

Procedure:

• Convert raw files to open formats (mzML, mzXML) using vendor converters or ProteoWizard.
• Perform peak picking and alignment across all samples with retention time correction.
• Annotate metabolites using accurate mass (±5 ppm), isotopic patterns, and MS/MS fragmentation against databases (HMDB, FoodDB, KEGG).
• Normalize data using internal standards and quality control samples.
• Apply Pareto scaling or unit variance scaling to reduce dominance of high-abundance metabolites.
• Conduct unsupervised pattern recognition using Principal Component Analysis (PCA) to identify natural clustering and outliers.
• Apply supervised methods such as Partial Least Squares-Discriminant Analysis (PLS-DA) or machine learning algorithms (random forests, support vector machines) to build classification models.
• Validate model performance through cross-validation (7-fold) and external validation with independent sample sets [6] [7].

The Scientist's Toolkit: Research Reagent Solutions

Table 2: Essential Research Reagents and Materials for Food Metabolomics

Reagent/Material	Function in Protocol	Application Example
HILIC Silica Column	Separation of polar metabolites	Geographic origin discrimination of black tea [7]
Stable Isotope-Labeled Internal Standards	Quality control & quantification	l-Phenylalanine-d8 for extraction efficiency monitoring [6]
Ammonium Formate	Mobile phase additive for improved ionization	HILIC-MS analysis of honey carbohydrates [6]
Formic Acid	Mobile phase modifier for protonation	LC-MS analysis of olive oil phenolics [6]
Acetonitrile: Methanol Extraction Solvent	Comprehensive metabolite extraction	Polar metabolite profiling from diverse food matrices [6]
C18 Reversed-Phase Column	Separation of non-polar metabolites	Lipid profiling for oil authentication [3]
Trimethoprim-13C3	Trimethoprim-13C3, CAS:1189970-95-3, MF:C14H18N4O3, MW:293.30 g/mol	Chemical Reagent
Desethyl Chloroquine-d4	Desethyl Chloroquine-d4, CAS:1189971-72-9, MF:C16H22ClN3, MW:295.84 g/mol	Chemical Reagent

Application Case Study: Black Tea Geographic Origin Authentication

A recent investigation demonstrated the power of non-targeted metabolomics for authenticating black tea geographical origins. Researchers analyzed 302 black tea samples from 9 distinct geographical indication regions using LC-QToF mass spectrometry. The workflow identified 229-145 metabolite biomarkers that enabled perfect discrimination (100% accuracy) between origins through internal 7-fold cross-validation and external validation [7]. This case exemplifies how non-targeted fingerprinting, coupled with machine learning, can address one of the most challenging authentication problems—multi-class geographical origin discrimination in complex plant-based products.

Figure 2: Black Tea Geographical Origin Authentication Workflow

Non-targeted metabolomics represents a transformative approach to food authentication, offering unprecedented capabilities for detecting sophisticated fraudulent practices across global supply chains. The methodologies and protocols detailed in this application note provide researchers with robust frameworks for implementing these powerful analytical strategies. As food fraud continues to evolve in complexity and scale, further advancements in high-resolution mass spectrometry, computational metabolomics, and multi-omics integration will be essential for developing increasingly sensitive, rapid, and accessible authentication platforms. These technological innovations, coupled with enhanced international collaboration and data sharing, will play a pivotal role in safeguarding food integrity and protecting consumers worldwide.

Metabolomics, defined as the comprehensive analysis of small molecule metabolites in a biological system, has emerged as a powerful tool in food science. In the context of food authentication, it provides a snapshot of the chemical fingerprint of a food product, which is influenced by its geographical origin, production method, and processing techniques [8] [3]. This approach is technically implemented to ensure consumer protection through the strict inspection and enforcement of food labeling, detecting adulterants or ingredients that are added deliberately to compromise the authenticity or quality of food products [8]. The two primary methodological paradigms in this field are untargeted and targeted metabolomics, each with distinct purposes, workflows, and applications.

Foodomics integrates metabolomics with other omics technologies like proteomics and genomics, using advanced biostatistics and bioinformatics to address complex challenges in food authenticity and safety from field to table [3]. The stability and complexity of the metabolome make it an ideal target for distinguishing authentic high-value foods from fraudulent substitutes.

Core Principles and Comparative Analysis

Untargeted Metabolomics

The untargeted approach is a hypothesis-generating methodology that aims to comprehensively analyze all detectable analytes in a sample without prior knowledge of which metabolites will be found [8] [9]. It is considered a "soft" authentication technique because it can detect both known and unknown forms of food fraud without targeting a specific adulterant, making it particularly valuable for detecting emerging and unpredictable fraudulent practices [9]. A key challenge is the immense amount of raw data generated, which requires sophisticated chemometric analysis for interpretation [8].

Targeted Metabolomics

In contrast, targeted metabolomics is a hypothesis-driven approach where the chemical attributes of the metabolites to be analyzed are known before data acquisition begins [8]. Analytical methods are specifically designed and validated to provide high precision, selectivity, and reliability for these predefined compounds [8]. This method leverages established knowledge of metabolic enzymes, their kinetics, and biochemical pathways, allowing for a focused investigation of specific metabolites or pathways of interest [8].

Table 1: Fundamental Characteristics of Untargeted and Targeted Metabolomics

Feature	Untargeted Metabolomics	Targeted Metabolomics
Objective	Hypothesis generation, comprehensive profiling, discovery of novel markers [8]	Hypothesis testing, precise quantification of predefined metabolites [8]
Scope	Global analysis of all detectable metabolites [8] [9]	Analysis of a predefined set of metabolites [8]
Nature	Non-targeted, "soft" authentication method [9]	Targeted, focused analysis
Data Complexity	High, requires advanced chemometrics [8]	Lower, focused data analysis
Identification Level	Unknowns and knowns, with challenges in annotation [10] [8]	Known metabolites, based on authentic standards

Table 2: Analytical and Practical Considerations for Metabolomics Approaches

Consideration	Untargeted Metabolomics	Targeted Metabolomics
Throughput	High-throughput for screening [11]	Lower throughput, focused analysis
Data Processing	Time-consuming; requires advanced tools (e.g., MS-DIAL, Compound Discoverer) [10] [8]	Streamlined, immediate biological interpretation [8]
Standardization	Challenging due to comprehensive nature [8]	Easier to standardize and validate
Ideal Application	Geographical discrimination, detection of unknown adulterants [10] [3]	Verification of specific adulteration, compliance testing [8]

Workflow and Experimental Design

The general workflow for metabolomics in food authentication involves several key stages, from sample preparation to data interpretation. The specific requirements, however, diverge significantly between untargeted and targeted strategies.

Figure 1: Comparative Workflows for Untargeted and Targeted Metabolomics

Sample Preparation Protocols

A. Generic Protocol for Untargeted Analysis of Plant Materials (e.g., Herbs and Spices) This protocol is adapted from studies on the geographical discrimination of thyme and other herbs [10].

• Homogenization: Weigh 200.00 ± 0.01 mg of the sample. For solid materials like dried thyme, grind to a fine powder (e.g., 0.2 mm particle size) using an ultra-centrifugal mill to ensure homogeneity [10].
• Extraction: Add 4 mL of chilled, GC-MS grade ethyl acetate to the sample in a 15 mL polypropylene tube.
• Sonication: Place the sample in an ultrasonic bath for 30 minutes at 37 kHz and room temperature to facilitate metabolite extraction [10].
• Centrifugation: Centrifuge the extract at 4400 × g (5500 rpm) for 10 minutes to pellet insoluble debris.
• Filtration: Filter the supernatant through a 0.45 µm nylon filter to remove any remaining particulates.
• Storage: Store the final extract at -21 °C until analysis to preserve metabolite stability.

B. Protocol for Animal Tissues (e.g., Liver or Muscle) This protocol is derived from toxicology and meat quality studies [12] [13].

• Quenching and Homogenization: Snap-freeze approximately 100 mg of tissue in liquid nitrogen and homogenize it to a fine powder using a mortar and pestle, kept cold with liquid nitrogen.
• Protein Precipitation: Resuspend the homogenized powder in 1 mL of pre-chilled 80% methanol. Vortex the mixture thoroughly.
• Incubation: Incubate the homogenate on ice for 5 minutes.
• Centrifugation: Centrifuge at 15,000 × g for 20 minutes at 4 °C to pellet proteins and other macromolecules.
• Dilution: Dilute a portion of the supernatant with LC-MS grade water to a final methanol concentration of 53% [12].
• Second Centrifugation: Centrifuge again at 15,000 × g for 20 minutes at 4 °C.
• Collection: Collect the final supernatant for LC-MS/MS analysis.

Instrumental Analysis and Data Acquisition

The choice of analytical platform is critical and depends on the chosen metabolomics approach.

Untargeted Analysis typically employs High-Resolution Mass Spectrometry (HRMS) coupled with chromatography (LC or GC) to achieve broad metabolite coverage.

• GC-Orbitrap-HRMS Protocol for Herbs [10]:

  • Instrument: Trace 1310 GC coupled to Q-Exactive Orbitrap mass analyzer.
  • Column: Standard capillary GC column.
  • Sample Injection: Follows established chromatographic methods for volatile compounds.
  • Data Acquisition: Full-scan mode with high mass accuracy (< 5 ppm) to record all detectable ions.

• LC-MS/MS Protocol for Animal Tissues [12] [14]:

  • Instrument: Vanquish UHPLC system coupled with an Orbitrap Q Exactive HF or HF-X mass spectrometer.
  • Column: Hypersil Gold column (100 × 2.1 mm, 1.9 μm).
  • Flow Rate: 0.2 mL/min with a 12-minute linear gradient.
  • Mobile Phases:
    • Positive ion mode: (A) 0.1% formic acid in water, (B) methanol.
    • Negative ion mode: (A) 5 mM ammonium acetate (pH 9.0), (B) methanol.
  • Ionization: Electrospray Ionization (ESI) in both positive and negative modes.

Targeted Analysis often uses triple quadrupole (QQQ) mass spectrometers operating in Selected Reaction Monitoring (SRM) or Multiple Reaction Monitoring (MRM) mode for high sensitivity and specific quantification of pre-defined metabolite panels.

Data Processing and Analysis

Untargeted Data Processing

The processing of untargeted HRMS data is a key and time-consuming challenge [10]. The workflow involves:

• Feature Extraction: Detecting all ion signals (features) from the raw data, comprising a mass-to-charge ratio (m/z), retention time, and intensity. Software tools like the open-source MS-DIAL and commercial Compound Discoverer are widely used for this purpose [10]. The performance of these tools can vary, leading to different subsets of detected features from the same dataset [10].
• Metabolite Annotation: Assigning a putative identity to features using accurate mass and fragmentation spectra (MS/MS) by querying metabolic databases such as the Human Metabolome Database (HMDB), FoodDB (www.foodb.ca), and MassBank [8] [15]. Confidence levels for identification should be reported (e.g., Level 1: confirmed with standard, Level 2: putative annotation) [10].
• Multivariate Statistical Analysis: Using techniques like Principal Component Analysis (PCA) and Orthogonal Projections to Latent Structures-Discriminant Analysis (OPLS-DA) to identify patterns and metabolites that differentiate sample groups (e.g., different geographical origins) [10] [13].
• Pathway Analysis: Enriching the biological interpretation by mapping differentially abundant metabolites to biochemical pathways using databases like KEGG (Kyoto Encyclopedia of Genes and Genomes) [12] [13].

Targeted Data Processing

Targeted data processing is more straightforward, focusing on:

• Peak Integration: Quantifying the area under the chromatographic peak for each targeted metabolite.
• Quantification: Calculating concentrations by comparing peak areas to a calibration curve created from authentic standards.
• Statistical Validation: Using univariate statistics (e.g., t-tests, ANOVA) to validate hypotheses about specific metabolite changes.

Figure 2: Pathway and Biomarker Analysis Workflow

Application in Food Authentication: A Case Study

Case Study: Geographical Discrimination of Thyme using GC-Orbitrap-HRMS [10]

• Objective: To differentiate thyme samples from Spain (Castilla-La Mancha) and Poland (Lublin) and identify marker metabolites.
• Approach: Untargeted metabolomics.
• Sample Preparation: Ultrasound-assisted extraction with ethyl acetate, as detailed in Section 3.1.A.
• Analysis: GC-Orbitrap-HRMS.
• Data Processing: Both MS-DIAL and Compound Discoverer software were compared for feature extraction and annotation.
• Results:
  • The data processing approach significantly influenced the results. Compound Discoverer putatively annotated 52 compounds, while MS-DIAL annotated 115 compounds (both at Level 2 confidence) [10].
  • Multivariate data analysis of the data from both software tools successfully identified differential compounds that served as markers for geographical discrimination [10].
  • This study highlights that the putative identification of markers in untargeted analysis heavily depends on the data processing parameters and the databases used [10].

The Scientist's Toolkit: Essential Reagents and Materials

Table 3: Key Research Reagent Solutions for Metabolomics in Food Authentication

Reagent / Material	Function	Example Use Case
GC-MS Grade Solvents (e.g., Ethyl Acetate)	High-purity extraction solvent for volatile and semi-volatile metabolites, minimizing background interference [10].	Ultrasound-assisted extraction of herbs and spices (e.g., thyme) [10].
LC-MS Grade Solvents & Additives (Methanol, Water, Formic Acid)	High-purity mobile phase components for LC-MS, essential for stable retention times and high sensitivity [12] [14].	Metabolite extraction and UHPLC separation of liver or muscle tissue [12] [13].
Authentic Chemical Standards	Unambiguous identification and absolute quantification of metabolites in targeted analyses [8].	Validation and quantification of potential biomarker molecules.
Alkane Standard Mixture (C7-C40)	Calculation of Kovats Retention Indices (KI) in GC-MS, aiding in the identification of metabolites [10].	Reliable annotation of volatile compounds in herb profiling [10].
Stable Isotope-Labeled Internal Standards (e.g., 13C, 15N)	Correction for matrix effects and losses during sample preparation, improving quantification accuracy [8].	Used in both targeted and untargeted workflows for data normalization.
Protein Precipitation Solvents (e.g., 80% Methanol)	Efficiently deproteinate complex biological samples like tissue or serum, releasing metabolites for analysis [12].	Preparation of liver tissue extracts for metabolomic profiling [12].
Nicosulfuron-d6	Nicosulfuron-d6, CAS:1189419-41-7, MF:C15H18N6O6S, MW:416.4 g/mol	Chemical Reagent
Carbosulfan-d18	Carbosulfan-d18, CAS:1189903-75-0, MF:C20H32N2O3S, MW:398.7 g/mol	Chemical Reagent

Food authentication is a critical scientific frontier in protecting public health, ensuring economic fairness, and combating food fraud, which costs the global economy an estimated $40 billion annually [16]. In an era of increasingly complex and globalized supply chains, verifying key product attributes—geographic origin, production system, and absence of adulteration—is paramount. Non-targeted metabolomics has emerged as a powerful tool for this purpose, capable of detecting unexpected deviations by comprehensively profiling the small-molecule composition of a food sample. This document provides detailed application notes and protocols for using non-targeted metabolomics to address these three primary authentication targets within a research framework.

Non-Targeted Metabolomics Workflow for Food Authentication

The standard non-targeted metabolomics workflow involves a sequence of steps from experimental design through data interpretation. The following diagram illustrates this integrated workflow, highlighting the key stages and decision points.

Figure 1: A generalized workflow for non-targeted metabolomics in food authentication, from sample preparation to model deployment.

Targeted Application Notes and Protocols

Authentication of Geographic Origin

3.1.1 Application Note Geographic origin is one of the most challenging authenticity targets due to the complex "terroir" effect—the interaction of genotype, environment, and agricultural practice that creates a unique biochemical fingerprint in a food product [5]. Non-targeted metabolomics can capture this fingerprint by analyzing a wide range of metabolites. A seminal study on 302 black tea samples from 9 geographical regions successfully used LC-QToF-based non-targeted fingerprinting combined with machine learning to discriminate origins with 100% accuracy in both internal and external validation [7]. This demonstrates the power of the approach to manage complex, multi-class discrimination problems.

3.1.2 Detailed Experimental Protocol

• Sample Preparation:

  • Materials: Lyophilizer, cryomill, analytical balance, methanol, water, methyl tert-butyl ether (MTBE).
  • Procedure:
    • Freeze-dry the samples (e.g., tea leaves) to a constant weight.
    • Homogenize into a fine powder using a cryomill.
    • Precisely weigh 50 mg of powdered sample into a 2 mL microcentrifuge tube.
    • Add 1 mL of a pre-cooled methanol/MTBE/water mixture (1.5:5:1.94, v/v/v).
    • Vortex vigorously for 1 minute, then sonicate in an ice-water bath for 30 minutes.
    • Centrifuge at 14,000 × g for 15 minutes at 4°C.
    • Collect the supernatant and filter through a 0.22 µm PVDF syringe filter into an LC-MS vial for analysis.

• LC-QToF Analysis:

  • Instrumentation: Agilent 1290 Infinity II LC system coupled to an Agilent 6546 QToF mass spectrometer.
  • Chromatography:
    • Column: ZORBAX Eclipse Plus C18 (2.1 × 100 mm, 1.8 µm).
    • Mobile Phase A: Water with 0.1% formic acid.
    • Mobile Phase B: Acetonitrile with 0.1% formic acid.
    • Gradient: 0-2 min, 5% B; 2-15 min, 5-95% B; 15-18 min, 95% B; 18-18.1 min, 95-5% B; 18.1-20 min, 5% B for re-equilibration.
    • Flow Rate: 0.3 mL/min.
    • Column Temperature: 40°C.
    • Injection Volume: 2 µL.
  • Mass Spectrometry:
    • Ionization: Dual AJS ESI, positive and negative ion modes.
    • Data Acquisition: Full scan mode (m/z 50-1700) and data-dependent MS/MS (Top 10) for biomarker identification.
    • Source Parameters: Drying gas temperature 325°C, flow 8 L/min, nebulizer 35 psi, sheath gas temperature 350°C, flow 11 L/min, VCap 3500 V.

• Data Processing and Analysis:

  • Convert raw data to open formats (e.g., mzML) using tools like ThermoRawFileParser [17].
  • Process using software like Metabox 2.0 or XCMS Online for peak picking, alignment, and integration [18].
  • For studies requiring high quantitative fidelity, apply the CCMN normalization method followed by square root transformation to best approximate absolute quantitative data [18].
  • Export the final peak intensity table for statistical analysis.

Verification of Production System

3.2.1 Application Note The production system (e.g., organic vs. conventional, free-range vs. caged) directly influences a food's metabolite profile due to differences in fertilizer use, animal feed, and overall management practices. Non-targeted metabolomics can detect markers associated with these inputs and stresses. For instance, it can identify the unauthorized use of synthetic fertilizers in products labeled as "organic" or distinguish between different farming practices [19].

3.2.2 Detailed Experimental Protocol

• Experimental Design:

  • Crucial: Collect paired samples from well-documented organic and conventional production systems, controlling for other variables like geographic location, cultivar, and harvest time.
  • Include a sufficient number of biological replicates (recommended n > 10 per group) to ensure statistical power.

• Metabolite Profiling:

  • The sample preparation and LC-QToF analysis can follow the protocol outlined in Section 3.1.2.
  • Focus on specific metabolite classes: The analytical method can be tuned to target specific classes known to be affected by production systems, such as polyphenols, alkaloids, or specific lipids.

• Statistical Analysis:

  • Perform unsupervised analysis (Principal Component Analysis - PCA) to observe natural clustering and identify potential outliers.
  • Use supervised methods like Orthogonal Projections to Latent Structures-Discriminant Analysis (OPLS-DA) to maximize the separation between organic and conventional groups and identify discriminant features.
  • Apply false discovery rate (FDR) correction to p-values to account for multiple testing.
  • Select features with a Variable Importance in Projection (VIP) score > 1.5 and a p-value (FDR-corrected) < 0.05 as potential biomarkers.

Detection of Adulteration and Substitution

3.2.1 Application Note Adulteration involves the addition of undeclared, inferior, or cheaper substances to a product. Common examples include adding cassava starch to sweet potato vermicelli [20], diluting olive oil with cheaper vegetable oils [16], or misrepresenting the species in meat and seafood products. Non-targeted metabolomics is highly effective here because it does not require a priori knowledge of the adulterant; it can detect unexpected compositional changes.

3.2.2 Detailed Experimental Protocol & Reverse Metabolomics

A powerful emerging strategy for adulteration detection is "reverse metabolomics" [17]. This approach leverages public data repositories to discover adulteration-relevant biomarkers, flipping the traditional workflow.

Figure 2: A comparison of the traditional and reverse metabolomics workflows. Reverse metabolomics begins with known spectra to discover biological or, in this context, adulteration-related associations from public data [17].

• Protocol for Reverse Metabolomics in Adulteration Detection:
  • Obtain MS/MS Spectra of Interest: Start with the MS/MS spectrum of a known marker for an authentic product or a suspected adulterant. This can come from in-house libraries or public databases like MassBank or GNPS [17]. Assign a Universal Spectrum Identifier (USI) if possible.
  • Mass Spectrometry Search Tool (MASST) Search: Use the MASST tool (or its domain-specific versions like foodMASST) to query public metabolomics data repositories (e.g., MetaboLights, GNPS) for datasets containing the same MS/MS spectrum [17].
  • Link Files with Metadata: Use frameworks like the Reanalysis Data User Interface (ReDU) to link the MASST search results (positive data files) with their associated sample metadata (e.g., sample type, disease state, geographic origin, or processing method) [17].
  • Validation: Statistically validate the association between the metabolite and a specific sample type (e.g., authentic vs. adulterated) found in the public data by conducting controlled, targeted experiments in the lab.

The Scientist's Toolkit

Table 1: Essential Reagents, Materials, and Software for Non-Targeted Metabolomics

Category	Item	Function / Application
Chemicals & Solvents	LC-MS Grade Methanol, Acetonitrile, Water	Mobile phase preparation, ensuring minimal background noise and ion suppression.
	Methyl tert-butyl ether (MTBE)	For lipid-rich sample extraction in biphasic systems.
	Formic Acid / Ammonium Acetate	Mobile phase additives to promote protonation/deprotonation in positive/negative ESI mode.
	Internal Standards (e.g., Heptanoic methyl ester)	Used for data normalization (e.g., CCMN method) to correct for technical variation [18].
Consumables	Syringe Filters (PVDF, 0.22 µm)	Filtering sample extracts prior to LC-MS injection to remove particulates.
	LC Vials and Caps	Safe holding of samples in the autosampler.
	Solid Phase Extraction (SPE) Cartridges (C18)	Clean-up of complex samples to reduce matrix effects.
Software & Databases	Metabox 2.0 / XCMS	Data processing pipeline: peak picking, alignment, normalization, and statistical analysis [18].
	GNPS (Global Natural Products Social Molecular Networking)	Platform for MS/MS spectral library matching, molecular networking, and performing MASST searches [17].
	FoodMASST	A specialized MASST tool for searching metabolomics data against foods and beverages [17].
	PubChem / HMDB	Chemical databases for metabolite annotation and structural information.
Molindone-d8	Molindone-d8, CAS:1189805-13-7, MF:C16H24N2O2, MW:284.42 g/mol	Chemical Reagent
Canniprene	Canniprene, CAS:70677-47-3, MF:C21H26O4, MW:342.4 g/mol	Chemical Reagent

Data Analysis and Biomarker Workflow

The journey from raw data to robust authentication biomarkers involves a critical feature selection and validation process, as visualized below.

Figure 3: The biomarker selection and validation workflow, which refines thousands of metabolic features into a concise, validated panel for model building [7].

Table 2: Summary of Quantitative Performance from a Non-Targeted Metabolomics Study on Black Tea [7]

Parameter	Result / Value
Sample Size	302 black tea samples
Number of Geographical Origins	9 regions
Metabolites Detected (Features)	229 - 145 selected as biomarkers
Model Validation	7-fold cross-validation & external validation
Reported Discrimination Accuracy	100%

Non-targeted metabolomics, supported by robust protocols for geographic origin, production system, and adulteration analysis, provides a comprehensive solution for modern food authentication challenges. The integration of advanced instrumentation, rigorous data processing techniques like CCMN normalization [18], and innovative discovery frameworks like reverse metabolomics [17] creates a powerful toolkit for researchers. As public metabolomics data repositories continue to grow, the potential for developing highly accurate, standardized, and globally applicable authentication models will only increase, ultimately leading to greater transparency and security in the global food supply chain.

The concept of terroir, traditionally associated with wine, refers to the unique combination of environmental factors that give an agricultural product its distinctive character. Scientifically, terroir encompasses the interactive ecosystem of a given place, including climate, soil, topography, and the associated biological communities such as the plant microbiome [21]. Modern metabolomics technologies now allow researchers to move beyond subjective tasting notes and objectively characterize the biochemical signatures imparted by terroir. This is particularly relevant for food authentication research, where non-targeted metabolomics serves as a powerful tool to verify geographical origin and combat fraud by detecting the unique metabolic fingerprints that arise from specific growing conditions [22] [21].

Metabolomics, the large-scale systematic study of small molecules or metabolites, is ideally suited to this task as it provides a snapshot of the physiological state of an organism, bridging the gap between genotype and phenotype [23] [24]. The metabolome is highly dynamic and can be perturbed by biology, phenotype, chemicals, or the environment, making it a sensitive marker for terroir-induced variation [24]. By employing non-targeted approaches, which comprehensively analyze a sample's metabolite profile without prior hypothesis, researchers can uncover the complex ways in which environment shapes food chemistry, thus providing a scientific basis for the terroir concept [25].

Key Metabolomic Findings on Terroir

Metabolic Pathways Influenced by Terroir

Research has consistently shown that specific metabolic pathways are particularly plastic and responsive to environmental conditions. The table below summarizes key pathways and metabolites affected by terroir in various agricultural products, as identified through non-targeted metabolomics studies.

Table 1: Key Metabolic Pathways and Metabolites Influenced by Terroir

Agricultural Product	Key Metabolic Pathways Affected	Specific Metabolites of Interest
Grape (Wine)	Phenylpropanoid pathway, Resveratrol biosynthesis, Tricarboxylic Acid (TCA) cycle, Fatty acid metabolism	Anthocyanins, Flavonoids, Tannins, Stilbenes, Organic acids (tartaric, malic) [22]
Coffee	Not specified in search results; requires non-targeted profiling	Aromas (Jasmine, Tangerine, Bergamot) linked to volatile organic compounds (VOCs) [21]
General Plant Products	Amino acid metabolism, Lipid metabolism, Carbohydrate metabolism	Amino acids (proline, arginine), Sugars (glucose, fructose), Fatty acids, Phenolic acids [22]

Studies on a single clone of the Corvina grape variety cultivated across different vineyards revealed that the phenylpropanoid pathway, especially resveratrol biosynthesis, was one of the most environmentally-dependent metabolic components [22]. This demonstrates that even without genetic variation, the environment can profoundly shape the phytochemical profile of a crop. Furthermore, environmental stress, such as limited nitrogen or high altitude, can trigger the accumulation of specific compounds like sugars, phenolics, anthocyanins, and tannins, which directly impact product quality and sensory characteristics [21].

The Role of the Plant Microbiome

A critical and often overlooked component of terroir is the plant microbiome. The collective communities of bacteria, fungi, and other microorganisms associated with plant organs (the rhizosphere, endosphere, and phyllosphere) form a holobiont with the host plant [21]. This microbiome contributes to terroir by:

• Altering Host Metabolism: Microbes can increase the nutrients absorbed by roots, which are then deposited in leaves, seeds, and fruits [21].
• Modifying the Metabolome: They can consume plant molecules, thereby removing them, or contribute their own metabolites, which directly add to the smells and flavors of the final product [21].

Advanced metagenomics and metabolomics have made it possible to correlate the diversity of a plant's microbiome with the chemical variation in its derived products, solidifying the microbiome's role as a key contributor to agricultural terroir [21].

Experimental Protocols for Non-Targeted Metabolomics in Terroir Research

This section details a standardized protocol for non-targeted metabolomics, adapted for characterizing the terroir of food products.

Sample Preparation and Metabolite Extraction

Principle: To reproducibly isolate a wide range of small molecules from solid food matrices (e.g., berries, beans, leaves) while minimizing degradation.

Protocol (Based on Grape Berry Metabolomics) [22]:

• Sampling: Collect plant material (e.g., 30 clusters from different positions along vine rows) at the desired physiological stage (e.g., véraison, mid-ripening, full maturity). Avoid damaged or infected tissues.
• Freezing and Grinding: Immediately freeze the selected samples in liquid nitrogen. Prior to extraction, crush and finely grind the frozen material (seeds removed) to a homogeneous powder using a pre-chilled mortar and pestle or a laboratory mill.
• Metabolite Extraction: Extract metabolites at room temperature using a methanol-based solvent.
  • Add three volumes (w/v) of methanol acidified with 0.1% (v/v) formic acid to the powdered tissue.
  • Sonicate in an ultrasonic bath at 40 kHz for 15 minutes.
  • Centrifuge the extract twice for 10 minutes at 16,000 × g at 4°C.
  • Dilute the supernatant 1:2 (v/v) with milliQ water.
  • Filter the diluted extract through a 0.2-μm syringe filter before instrumental analysis.

Standardized Approach (PTFI Platform) [25]: For greater cross-study comparability, the Partnership for Food Metabolomics Innovation (PTFI) platform uses a standardized protocol involving solid phase extraction (SPE) to isolate small molecules. A key feature is the incorporation of a unique internal retention standard reagent containing 33 compounds not found endogenously in food, which allows for data harmonization across different laboratories.

Instrumental Analysis: LC-HRMS

Principle: To separate, detect, and accurately mass-measure the vast array of metabolites in a complex extract.

Protocol [22] [25]:

• Chromatography: Use Reverse-Phase Liquid Chromatography (RP-LC).
  • Column: An analytical C18 column (e.g., 150 × 2.1 mm, 3 μm particle size).
  • Mobile Phase: Solvent A (5% acetonitrile, 0.5% formic acid in water) and Solvent B (100% acetonitrile).
  • Gradient: Employ a linear gradient, for example: 0-10% B in 5 min, 10-20% B in 20 min, 20-25% B in 5 min, and 25-70% B in 15 min, at a constant flow rate of 0.2 mL/min.
• Mass Spectrometry: Use High-Resolution Mass Spectrometry (HRMS) such as an Orbitrap or FT-MS instrument.
  • Ionization: Electrospray Ionization (ESI), alternating between positive and negative ion modes.
  • Scanning: Full scan mode in the range of 50-1500 m/z.
  • Data-Dependent Acquisition (optional): For metabolite identification, trigger MS/MS or MSn scans for the most intense ions with a defined fragmentation amplitude.

The following workflow diagram summarizes the key steps from sample to data, highlighting the parallel paths for MS1 and MS/MS data, which are crucial for identification in non-targeted studies.

Data Processing and Functional Analysis

Principle: To convert raw spectral data into meaningful biological insights about pathway activity.

Protocol [23] [26]:

• Data Preprocessing: Use software like XCMS, MZmine, or MetaboAnalyst to perform:
  • Noise filtering and peak detection.
  • Retention time alignment and correction.
  • Peak integration and deconvolution.
  • Creation of a data matrix (samples × metabolic features with intensities).
• Quality Control: Use Quality Control (QC) samples to monitor and correct for technical variance. Features with high variance in QCs are typically removed.
• Compound Identification & Functional Analysis:
  • For MS1 peak lists: Upload a table containing m/z, p-values, and/or t-scores/fold-changes into a tool like MetaboAnalyst. Use algorithms like mummichog to predict pathway activity directly from the m/z features, bypassing the need for complete metabolite identification [26].
  • For MS/MS data: Use spectral matching against reference libraries (e.g., MassBank, GNPS) to achieve a higher level of identification confidence (e.g., Level 2 or higher per the Metabolomics Standards Initiative) [23].

Computational Analysis & Data Integration

Non-targeted metabolomics generates complex, high-dimensional data. Effective analysis requires specialized statistical and bioinformatics tools.

• Multivariate Statistics: Techniques like Principal Component Analysis (PCA) and Orthogonal Projections to Latent Structures-Discriminant Analysis (OPLS-DA) are essential for visualizing clustering patterns and identifying the metabolic features that most contribute to the differentiation between terroirs [24] [27].
• Pathway Analysis: Tools like mummichog (in MetaboAnalyst) use a priori knowledge of metabolic pathways to infer biological activity from significant m/z features, providing a functional interpretation of the terroir effect [26].
• Data Integration: Integrating metabolomics data with other omics data (transcriptomics, proteomics) and metadata (soil composition, climate data) is recommended to obtain an exhaustive description of the biological processes underlying terroir [23]. This requires specialized integration algorithms and software.

The diagram below illustrates the logical flow of computational analysis, from raw data to biological interpretation, showcasing how different data types are integrated.

The Scientist's Toolkit: Research Reagent Solutions

Successful non-targeted metabolomics relies on a suite of reliable reagents and materials. The table below lists essential items for a terroir study based on LC-HRMS.

Table 2: Essential Research Reagents and Materials for Non-Targeted Metabolomics

Item	Function / Purpose	Example / Specification
LC-MS Grade Solvents	To minimize background noise and ion suppression during MS analysis; essential for high-sensitivity detection.	Acetonitrile, Methanol, Water, Formic Acid (all LC-MS grade) [22]
Solid Phase Extraction (SPE) Cartridges	To clean up samples and pre-concentrate metabolites, reducing matrix effects and improving data quality.	Reverse-phase C18 cartridges [25]
Internal Standard Mixture	To correct for retention time shifts and enable data harmonization across multiple batches and labs.	PTFI's mixture of 33 nonendogenous compounds [25]
Authenticated Chemical Standards	For confident metabolite identification (Level 1 according to MSI) by matching retention time and MS/MS spectrum.	Commercial standards for key metabolites (e.g., resveratrol, malic acid) [23]
Quality Control (QC) Pool Sample	To monitor instrument stability, balance analytical bias, and correct for technical noise throughout a run.	A pool created by combining small aliquots of all experimental samples [23]
Chromatography Column	To separate a complex metabolite mixture based on chemical polarity, reducing MS complexity and increasing ID confidence.	Reverse-phase C18 column (e.g., 150 x 2.1 mm, 3 μm) [22]
Methocarbamol-d5	Methocarbamol-d5, CAS:1189699-70-4, MF:C11H15NO5, MW:246.27 g/mol	Chemical Reagent
Oxyphenbutazone-d9	Oxyphenbutazone-d9, CAS:1189693-23-9, MF:C19H20N2O3, MW:333.4 g/mol	Chemical Reagent

Non-targeted metabolomics has emerged as a powerful analytical strategy for food authentication, offering a comprehensive snapshot of the complex metabolite profiles in food commodities. This approach is particularly vital for combating economically motivated adulteration in high-value products such as wine, olive oil, honey, spices, and cereals. Unlike targeted methods that focus on predefined compounds, non-targeted metabolomics enables the detection of unexpected adulterants and emerging fraud patterns by analyzing the entire metabolome [28] [29]. This application note provides detailed protocols and data analysis workflows for authenticating these top five commodities of concern, supporting researchers in implementing robust food integrity programs.

Experimental Protocols

Sample Preparation Standards

Universal Metabolite Extraction Protocol:

• Homogenization: Cryogenically grind solid samples (grains, spices) using liquid nitrogen, mortar, and pestle to preserve labile metabolites [30] [31].
• Weighing: Accurately weigh 100±5 mg of homogeneous sample into extraction tubes.
• Extraction: Add 1 mL of cold extraction solvent (80:20 methanol:water with 0.1% formic acid) per 100 mg sample [32] [33].
• Spiking: Introduce internal standards (e.g., creatine-D3, leucine-D3, L-tryptophan-D3) at 0.5 ng/μL final concentration [32] [33].
• Mixing: Vortex vigorously for 60 seconds, then shake for 15 minutes at room temperature.
• Centrifugation: Spin at 18,000× g for 10 minutes at 4°C [32].
• Collection: Transfer 500 μL of supernatant to LC-MS vial for analysis.

Note: For oily matrices (olive oil), prior liquid-liquid extraction with hexane may be required to remove lipids that interfere with analysis [28].

LC-HRMS Non-Targeted Analysis

Chromatographic Conditions:

• Column: HILIC or C18 (e.g., DB-5, 30 m × 0.25 mm × 0.25 μm) [30] [33]
• Mobile Phase A: LC-MS grade water with 0.1% formic acid
• Mobile Phase B: Acetonitrile with 0.1% formic acid
• Gradient: 5-95% B over 25 minutes, hold at 95% B for 5 minutes
• Flow Rate: 0.3 mL/min
• Injection Volume: 5 μL [32] [34]

Mass Spectrometry Parameters:

• Instrument: UPLC-QTOF or UPLC-Orbitrap
• Ionization: ESI positive/negative mode switching
• Mass Range: m/z 50-1000
• Resolution: >30,000
• Collision Energy: 10-40 eV ramp for MS/MS
• Source Temperature: 300°C
• Drying Gas: 8 L/min [32] [34] [33]

Data Processing Workflow

• Raw Data Conversion: Convert vendor files to .mzXML or .netCDF format
• Peak Detection: Use AMDIS or MS-DIAL for peak picking and deconvolution
• Alignment: Correct retention time drift across samples
• Normalization: Apply internal standard and quality control-based correction
• Metabolite Annotation: Query databases (HMDB, MassBank, mzCloud) with mass accuracy <5 ppm [30] [33]

Commodity-Specific Authentication Data

Table 1: Quality Indices and Adulteration Markers in Olive Oil

Parameter	EVOO Standard	Adulterated/Low Quality	Analytical Method
Free Fatty Acids	≤0.8% oleic acid	>0.8% oleic acid	Titration (AOCS Ca 5a-40) [28]
Peroxide Value	≤20 meq O₂/kg	>20 meq O₂/kg	Titration (AOCS Cd 8-53) [28]
Pyropheophytins	≤17%	>17%	HPLC-DAD (ISO 29841:2009) [28]
Phenolic Compounds	Specific profile	Altered profile	LC-QTOF-MS [35]
Fatty Acid Profile	Specific composition	Deviations	GC-FID [28]

Table 2: Metabolomic Profiling of Cereal Phenolics (μg/g)

Phenolic Compound	Barley	Corn	Oats	Rice	Rye	Wheat
Catechin	1.31-2.38	7.36	0.56±0.05	0-1.39	+	0.83-1.79 [31]
Quercetin	0.0004-18.41	0.09-1.58	10.18±0.06	0-1.87	+	1.96-10.48 [31]
Cyanidin	0.86-23.93	0.6-260.1	npr	0-302.22	0.29	0-7.1 [31]
Apigenin	+	+	npr	1.44-2.85	0-1.52	20.0-36.5 [31]
Ferulic Acid	High	Medium	Medium	High	Medium	High [31]

+ = present but not quantified; npr = no published results

The Scientist's Toolkit

Table 3: Essential Research Reagents and Materials

Reagent/Material	Function	Example Applications
HILIC Chromatography Column	Polar metabolite separation	Cereal sugars, wine acids [33]
C18 Reverse Phase Column	Non-polar metabolite separation	Olive oil phenolics, spice oils [30] [35]
Stable Isotope Standards	Quantification & normalization	Absolute metabolite quantification [33]
Divinylbenzene/CAR/PDMS Fiber	SPME for volatile capture	Spice aroma profiling [30]
Methanol:Water (80:20)	Metabolite extraction	Universal metabolite extraction [32] [33]
Quality Control Pool	System performance monitoring	All non-targeted experiments [33]
Climbazole-d4	Climbazole-d4, CAS:1185117-79-6, MF:C15H17ClN2O2, MW:296.78 g/mol	Chemical Reagent
Prazobind-d8	Prazobind-d8, MF:C23H27N5O3, MW:429.5 g/mol	Chemical Reagent

Data Analysis and Chemometric Modeling

Statistical Workflow for Authentication

Step 1: Data Preprocessing

• Apply quality control-based robust LOESS signal correction
• Use Pareto or Unit Variance scaling for normalization
• Implement missing value imputation (KNN or minimum value)

Step 2: Exploratory Analysis

• Perform Principal Component Analysis (PCA) to identify outliers and natural clustering
• Generate hierarchical clustering to visualize sample relationships

Step 3: Supervised Modeling

• Develop Partial Least Squares-Discriminant Analysis (PLS-DA) models to maximize class separation
• Apply Orthogonal PLS-DA (OPLS-DA) to separate predictive and non-predictive variation
• Validate models using cross-validation and permutation testing (n>100) [28] [34] [35]

Step 4: Marker Selection

• Calculate Variable Importance in Projection (VIP) scores
• Perform ANOVA with false discovery rate correction
• Assess fold-change thresholds (typically >1.5 or <0.67) [35]

Case Study: Wine Clone Discrimination

Non-targeted UPLC-FT-ICR-MS successfully distinguished wines from three different Vitis vinifera cv. Pinot noir clones grown under identical conditions. The sensory analysis panel detected significant differences in astringency, bitterness, and acidity that correlated with specific non-volatile metabolite profiles. The OPLS-DA model demonstrated excellent separation (R2Y=0.95, Q2=0.87) with 25 molecular features contributing most to discrimination [34].

Case Study: Honey Adulteration Detection

LC-HRMS non-targeted metabolomics identified a specific marker for sugar syrup adulteration in honey that was present in beet, corn, and wheat syrups but absent in authentic honeys. The method demonstrated a limit of quantification of approximately 5% fortification, showing a linear trend in intentionally adulterated samples [32].

Non-targeted metabolomics provides an powerful framework for authenticating high-risk food commodities. The protocols and data analysis workflows presented here enable researchers to implement comprehensive food authentication programs that can detect both known and emerging adulteration practices. Through integration of advanced analytical techniques with robust chemometric modeling, these methods offer the sensitivity, specificity, and breadth needed to address evolving challenges in food fraud prevention.

Analytical Platforms and Data Analysis Strategies

Non-targeted metabolomics has emerged as a powerful analytical strategy for food authentication, enabling the comprehensive detection and identification of metabolites without prior hypothesis [36] [37]. This approach is particularly valuable for addressing food fraud challenges, including mislabeling of geographical origin, species substitution, and detection of undeclared adulterants [3] [37]. The analytical workflow encompasses multiple critical stages, from initial sample collection through to data acquisition and processing, each requiring meticulous execution to ensure data quality and biological relevance. This protocol outlines a standardized workflow specifically tailored for food authentication studies, incorporating recent methodological advances to enhance cross-laboratory reproducibility and data reliability [38].

Experimental Protocols

Sample Preparation and Extraction

Principle: The objective of sample preparation is to extract a comprehensive range of metabolites while maintaining sample integrity and minimizing analytical bias. Proper sample handling is crucial for obtaining metabolomic profiles that accurately represent the food sample's biochemical composition [37].

Protocol Steps:

• Homogenization: For solid food matrices (meat, grains, cheese), rapidly freeze samples in liquid nitrogen and pulverize using a laboratory mill until a fine, homogeneous powder is achieved. For liquid matrices (oil, milk, juice), vortex thoroughly for 30-60 seconds to ensure uniformity [37].
• Metabolite Extraction: Weigh 100 ± 5 mg of homogenized solid sample or aliquot 1 mL of liquid sample into a 2 mL microcentrifuge tube.
  • Add 1 mL of pre-chilled extraction solvent (typically methanol:water:chloroform in a 2:1:1 ratio) to simultaneously extract both polar and non-polar metabolites [37].
  • Vortex vigorously for 60 seconds, then sonicate in an ice-water bath for 15 minutes.
• Precipitation and Recovery: Centrifuge at 14,000 × g for 15 minutes at 4°C to pellet proteins and insoluble debris. Carefully transfer the supernatant (containing the metabolites) to a new vial.
• Concentration and Reconstitution: Evaporate the solvent to dryness under a gentle stream of nitrogen gas. Reconstitute the dried metabolite extract in 100 µL of solvent compatible with the subsequent analytical method (e.g., water:acetonitrile, 95:5 for LC-MS). Vortex for 30 seconds to ensure complete dissolution.
• Quality Control (QC): Pool equal aliquots from all samples to create a quality control sample. This QC sample is analyzed repeatedly throughout the analytical sequence to monitor instrument performance and stability [38].

Data Acquisition via Liquid Chromatography-Mass Spectrometry (LC-MS)

Principle: LC-MS combines chromatographic separation with high-sensitivity mass spectrometric detection, making it the cornerstone platform for non-targeted metabolomics due to its broad coverage of metabolites [39] [36]. The choice between Data-Dependent Acquisition (DDA) and Data-Independent Acquisition (DIA) is a key consideration, as performance is dependent on sample complexity [39].

Protocol Steps:

• Chromatographic Separation:
  • Column: Utilize a reversed-phase C18 column (e.g., 2.1 × 100 mm, 1.7 µm) maintained at 40°C.
  • Mobile Phase: A) Water with 0.1% formic acid and B) Acetonitrile with 0.1% formic acid.
  • Gradient: Employ a linear gradient from 2% to 98% B over 20 minutes, followed by a 5-minute wash at 98% B and a 7-minute re-equilibration at 2% B.
  • Flow Rate: 0.3 mL/min.
  • Injection Volume: 5 µL.
• Mass Spectrometric Detection:
  • Ion Source: Electrospray Ionization (ESI), operated in both positive and negative ionization modes to maximize metabolite coverage.
  • Source Parameters: Capillary voltage: 3.0 kV (ESI+), 2.5 kV (ESI-); Source temperature: 150°C; Desolvation temperature: 350°C.
  • Mass Analyzer: Time-of-Flight (TOF) mass analyzer for high-resolution and accurate mass measurement.
  • Acquisition Mode: Operate in full-scan mode over a mass range of m/z 50-1200 for MS¹ profiling. The decision to use DDA or DIA should be guided by sample complexity. DIA (e.g., MSE, SWATH) fragments all ions within sequential isolation windows, providing comprehensive fragmentation data and is superior when few compounds elute simultaneously. DDA selects the most abundant ions from the MS¹ scan for fragmentation, which can be more effective as ion overlap increases in complex samples [39].

Data Processing and Statistical Analysis

Principle: Raw LC-MS data must be processed to extract meaningful metabolic features, which are then subjected to statistical analysis to identify metabolites that differentiate authentic from adulterated food samples [40] [37].

Protocol Steps:

• Spectral Processing and Feature Detection: Process raw data files using software platforms (e.g., MZmine, XCMS, or the Global Natural Products Social Molecular Networking (GNPS) pipeline). Key steps include:
  • Peak picking and deconvolution.
  • Retention time alignment and correction using an Internal Retention Time Standard (IRTS) mixture of compounds non-endogenous to food to enable robust cross-laboratory data alignment [38].
  • Isotope and adduct annotation.
  • Generation of a feature table containing m/z, retention time, and intensity for all detected peaks.
• Multivariate Statistical Analysis: Import the normalized feature table into statistical software (R, Python, or SIMCA-P).
  • Unsupervised Analysis: Perform Principal Component Analysis (PCA) to visualize natural clustering and identify outliers.
  • Supervised Analysis: Apply Orthogonal Projections to Latent Structures-Discriminant Analysis (OPLS-DA) to maximize separation between pre-defined sample classes (e.g., authentic vs. adulterated) and identify candidate biomarker ions.
• Metabolite Identification and Validation: Query the accurate mass and fragmentation spectra (MS/MS) of significant features against metabolomic databases such as HMDB, METLIN, and FoodDB [36] [37]. Confirm putative identifications by comparing retention times and fragmentation patterns with authentic chemical standards, when available.

Workflow Visualization

The following diagram summarizes the complete analytical workflow for non-targeted metabolomics in food authentication.

The Scientist's Toolkit: Essential Research Reagents and Materials

The following table details key reagents, solvents, and materials essential for executing the non-targeted metabolomics workflow for food authentication.

Table 1: Essential Research Reagents and Materials for Non-Targeted Metabolomics

Item	Function/Application in Workflow
Internal Retention Time Standard (IRTS)	A proprietary mixture of compounds non-endogenous to food; enables robust chromatographic alignment and cross-laboratory comparison of data [38].
Methanol, Acetonitrile, Chloroform (HPLC/MS Grade)	High-purity solvents used for metabolite extraction and as mobile phases in LC-MS to minimize background noise and ion suppression.
Formic Acid (Optima LC/MS Grade)	Mobile phase additive (0.1%) used to promote protonation of analytes in positive ESI mode, improving ionization efficiency and chromatographic peak shape.
Water (HPLC/MS Grade)	Used for sample reconstitution and as a mobile phase component; high purity is critical to reduce chemical noise.
Reversed-Phase C18 LC Column	The core separation media (e.g., 2.1 x 100 mm, 1.7 µm) for resolving a wide range of metabolites based on hydrophobicity prior to mass spectrometry.
Quality Control (QC) Sample	A pooled sample from all test samples used to condition the instrument and injected at regular intervals throughout the run to monitor system stability and performance.
Mass Calibration Standard	A reference standard (e.g., sodium formate) used to calibrate the mass axis of the mass spectrometer, ensuring high mass accuracy for metabolite identification.
(R)-Bromoenol lactone	(R)-Bromoenol lactone|Selective iPLA2γ Inhibitor
Trofosfamide-d4	Trofosfamide-d4, CAS:1189884-36-3, MF:C9H18Cl3N2O2P, MW:327.6 g/mol

Data Presentation: Key Analytical Parameters

The table below summarizes the core quantitative parameters and specifications for the major stages of the LC-MS-based non-targeted metabolomics workflow.

Table 2: Key Parameters for LC-MS-Based Non-Targeted Metabolomics Workflow

Workflow Stage	Parameter	Specification / Value	Purpose / Rationale
Sample Preparation	Sample Amount	100 mg (solid); 1 mL (liquid)	Provides sufficient material for comprehensive metabolite extraction.
	Extraction Solvent	Methanol:Water:Chloroform (2:1:1)	Simultaneous extraction of polar and non-polar metabolites.
LC Separation	Column Type	Reversed-Phase C18 (e.g., 1.7 µm)	High-resolution separation of complex metabolite mixtures.
	Run Time	~32 minutes (incl. equilibration)	Balances throughput with sufficient chromatographic resolution.
	Mobile Phase	Water/Acetonitrile + 0.1% Formic Acid	Facilitates efficient separation and ionization in ESI-MS.
MS Detection	Mass Analyzer	Time-of-Flight (TOF)	Provides high-resolution and accurate mass data for metabolite identification.
	Mass Range	m/z 50 - 1200	Covers a broad range of small molecule metabolites.
	Ionization Mode	ESI+ and ESI-	Maximizes coverage of ionizable metabolites.
Data Processing	IRTS	Included in all samples	Enables cross-laboratory chromatographic alignment [38].
	Software	MZmine, XCMS, GNPS	For feature detection, alignment, and statistical analysis [40].

This application note provides a detailed protocol for an analytical workflow in non-targeted metabolomics, specifically contextualized for food authentication research. The standardized method—from rigorous sample preparation through to advanced data acquisition and processing strategies—ensures the generation of high-quality, reproducible data. The integration of internal standards for cross-laboratory alignment and clear guidelines for handling complex food matrices makes this workflow a robust tool for combating food fraud. By adhering to this structured approach, researchers can reliably identify metabolite markers that are essential for verifying food authenticity, ensuring safety, and protecting consumer interests.

Mass spectrometry (MS) platforms are indispensable tools in modern food authentication research, enabling the precise detection and identification of metabolites that serve as chemical fingerprints for food origin, quality, and authenticity. Non-targeted metabolomics has emerged as a powerful hypothesis-generating approach that comprehensively analyzes small molecule metabolites to distinguish food products based on geographical origin, variety, and production methods [41]. The versatility of MS platforms allows researchers to address complex food fraud challenges through detailed chemical profiling.

The fundamental strength of mass spectrometry lies in its ability to provide both qualitative and quantitative information on a wide range of compounds in complex food matrices. When hyphenated with separation techniques like gas chromatography (GC) and liquid chromatography (LC), MS becomes exceptionally powerful for resolving complex mixtures encountered in food analysis [42] [41]. The continuous advancements in high-resolution mass spectrometry (HRMS) have significantly enhanced these capabilities, providing greater confidence in metabolite identification through accurate mass measurement [43].

This article explores the principal MS platforms used in food authentication research, with a specific focus on their application in non-targeted metabolomics for addressing food integrity challenges. We will examine the complementary strengths of GC-MS and LC-MS systems, the transformative role of high-resolution techniques, and provide detailed application notes and protocols for implementing these methodologies in food authentication research.

The selection of an appropriate MS platform represents a critical decision point in designing food authentication studies. Each platform offers distinct advantages and limitations that must be aligned with research objectives, sample characteristics, and target metabolome coverage.

GC-MS systems excel in separating and analyzing volatile and semi-volatile compounds, making them ideal for aroma profiling and primary metabolite analysis [42]. The technique provides high chromatographic resolution and excellent reproducibility, with electron ionization (EI) generating consistent, library-searchable fragmentation patterns. A key limitation, however, is the requirement for volatile analytes, often necessitating chemical derivatization for non-volatile compounds like sugars, organic acids, and amino acids [44] [45]. This additional sample preparation step can introduce variability but enables coverage of central carbon metabolism intermediates.

LC-MS platforms, particularly those coupled to high-resolution mass spectrometers, offer complementary capabilities for analyzing non-volatile, thermally labile, and high molecular weight compounds without derivatization [46]. This includes important biomarker classes like polyphenols, lipids, and carotenoids that are intractable to GC-MS analysis. The soft ionization techniques employed in LC-MS (electrospray ionization (ESI) and atmospheric pressure chemical ionization (APCI)) primarily generate molecular ion information, with structural characterization achieved through tandem MS experiments [41].

High-resolution mass spectrometry represents a significant advancement, with Orbitrap and time-of-flight (TOF) analyzers providing accurate mass measurements (<5 ppm mass accuracy) that enable confident elemental composition assignment and facilitate the identification of unknown metabolites [43] [41]. The high resolving power (>25,000) allows separation of isobaric compounds that would co-elute on lower resolution instruments, while full-scan data acquisition enables retrospective data mining without re-injection [42].

Table 1: Comparison of Mass Spectrometry Platforms for Food Authentication

Platform	Mass Analyzer	Resolving Power	Mass Accuracy	Key Applications in Food Authentication	Limitations
GC-MS	Quadrupole, TOF	Unit resolution (Quadrupole), >5,000 (TOF)	>100 ppm	Geographical origin discrimination, variety differentiation, volatile profiling	Requires volatility/derivatization, limited to lower molecular weight compounds
LC-MS/MS	QqQ, Ion Trap	Unit resolution	>100 ppm	Targeted analysis of specific biomarker classes, adulterant detection	Limited compound identification in untargeted mode
GC-Orbitrap	Orbitrap	25,000-120,000	<3 ppm	Untargeted analysis for geographical discrimination, marker identification [47]	Higher instrument cost, requires derivatization
LC-Orbitrap	Orbitrap	25,000-240,000	<3 ppm	Comprehensive lipidomics, polyphenol profiling, food intake biomarker discovery [48]	Higher instrument cost, matrix effects in ESI
LC-TOF	TOF	20,000-60,000	<5 ppm	Food contaminant screening, metabolite fingerprinting [41]	Requires frequent mass calibration

The choice between these platforms should be guided by the specific research question. For volatile profiling or central metabolite analysis, GC-MS provides robust, reproducible data. For comprehensive analysis of secondary metabolites and complex lipids, LC-HRMS is indispensable. Many advanced laboratories now employ complementary platforms to maximize metabolome coverage, as the combined data provides a more complete chemical signature for authentication purposes [41].

Detailed Experimental Protocols

GC-Orbitrap-HRMS for Geographical Discrimination of Herbs

This protocol details the application of GC-Orbitrap-HRMS for geographical discrimination of thyme, based on a published case study that demonstrated successful differentiation of Spanish and Polish origins [47].

Sample Preparation:

• Weigh 200.0 ± 0.1 mg of homogenized thyme sample (particle size 0.2 mm) into a 15 mL polypropylene tube.
• Add 4 mL of GC-MS grade ethyl acetate (≥99.5% purity).
• Perform ultrasound-assisted extraction for 30 minutes at 37 kHz and room temperature.
• Centrifuge at 5,500 rpm (4,400 × g) for 10 minutes.
• Filter supernatant through 0.45 μm nylon filters.
• Store extracts at -21°C until analysis.
• Prepare procedure blanks to identify background signals.

Instrumental Analysis:

• System: Thermo Scientific Trace 1310 GC coupled to Q-Exactive Orbitrap mass analyzer.
• Column: BP5MS capillary column (30 m × 0.25 mm i.d., 0.25 μm film thickness).
• Injection: 1 μL with split/splitless injector.
• Carrier Gas: Helium, constant flow.
• Oven Program: Initial temperature 60°C (hold 1 min), ramp to 300°C at 10°C/min, final hold 5 min.
• Transfer Line Temperature: 280°C.
• Ionization: Electron ionization (EI) at 70 eV.
• Mass Resolution: 60,000 at m/z 200.
• Mass Range: m/z 50-500.
• Quality Control: Inject pooled quality control (QC) samples throughout the sequence to monitor system performance.

Data Processing:

• Process raw data using either commercial (Compound Discoverer) or open-source (MS-DIAL) software.
• Perform peak picking, deconvolution, and alignment.
• Annotate metabolites using authentic standards (Level 1 identification) or spectral libraries (Level 2 identification) [47].
• Apply multivariate statistical analysis (PCA, PLS-DA) to identify discriminatory features.

Figure 1: GC-Orbitrap-HRMS Workflow for Food Authentication

LC-HRMS Metabolomic Profiling for Dietary Biomarker Discovery

This protocol describes an untargeted LC-MS/MS approach for discovering biomarkers of dietary patterns, specifically applied to identify plasma biomarkers of Mediterranean diet adherence [48] [49].

Sample Preparation:

• Thaw plasma samples slowly on ice for 30 minutes.
• Aliquot 100 μL of plasma into a clean microcentrifuge tube.
• Add 300 μL of ice-cold methanol for protein precipitation.
• Mix for 10 minutes at 700 rpm.
• Centrifuge at 13,000 × g for 15 minutes at 4°C.
• Filter supernatant through 0.22 μm centrifugal filters at 8,000 × g for 5 minutes.
• Transfer to maximum recovery vials for LC-MS analysis.

LC-MS Analysis:

• System: Dionex Ultimate 3000 UHPLC coupled to LTQ Orbitrap Elite mass spectrometer.
• Column: C18 column (e.g., 150 × 2.1 mm, 1.9 μm).
• Mobile Phase: A) water with 0.1% formic acid; B) acetonitrile with 0.1% formic acid.
• Gradient: 5-95% B over 25 minutes.
• Flow Rate: 0.3 mL/min.
• Column Temperature: 40°C.
• Injection Volume: 5-10 μL.
• Ionization: Heated electrospray ionization (H-ESI) in positive and negative modes.
• Mass Resolution: 60,000-120,000.
• Mass Range: m/z 100-1500.
• Fragmentation: Data-dependent MS/MS for top 10 ions.

Data Processing and Biomarker Panel Development:

• Process raw data using XCMS Online or similar software for peak detection, alignment, and normalization.
• Perform statistical analysis using multivariate methods (PCA, OPLS-DA) to identify differentially abundant features.
• Validate model quality with cross-validation and permutation tests (R2 > Q2).
• Annotate significant features using accurate mass, isotopic patterns, and MS/MS fragmentation.
• Develop biomarker panel through stepwise elimination, selecting 3-5 highly discriminatory metabolites.
• Build logistic regression model to classify dietary adherence and evaluate using ROC curve analysis (AUC >0.8 indicates good performance) [48].

GC-MS Metabolomics for Rice Variety Differentiation

This protocol describes a non-targeted GC-MS approach for discriminating rice varieties based on their seed metabolic profiles [44] [45].

Sample Preparation and Derivatization:

• Grind brown rice seeds to fine powder using a mixer mill.
• Weigh 50 mg into a 2 mL microcentrifuge tube.
• Add 0.5 mL of methanol:chloroform (3:1, v/v) and 10 μL of ribitol (2 mg/mL in water) as internal standard.
• Vortex for 30 seconds, then grind at 45 Hz for 4 minutes.
• Incubate in ice bath for 5 minutes.
• Repeat steps 4-5 three times for complete extraction.
• Centrifuge at 12,000 rpm for 15 minutes.
• Collect 300 μL of polar (upper) phase.
• Dry completely in a centrifugal concentrator for 3 hours.
• Add 60 μL of methoxyamine hydrochloride (20 mg/mL in pyridine) and incubate at 80°C for 30 minutes for methoximation.
• Add 70 μL of BSTFA (with 1% TMCS) and incubate at 70°C for 1.5 hours for trimethylsilylation.
• Cool to room temperature and add 5 μL of FAMEs (for retention index calibration).

GC-TOF-MS Analysis:

• System: Agilent 7890 GC coupled to Pegasus HT TOF mass spectrometer.
• Column: DB-5MS capillary column (30 m × 0.25 mm, 0.25 μm).
• Injection: 1 μL in split mode (split ratio 10:1).
• Carrier Gas: Helium, constant flow 3.0 mL/min.
• Oven Program: 50°C (hold 1 min), to 310°C at 10°C/min, hold 8 min.
• Transfer Line: 280°C.
• Ion Source: 250°C.
• Ionization: EI at 70 eV.
• Mass Range: m/z 50-500.
• Acquisition Rate: 10 spectra/second.

Data Processing:

• Process raw data using Chroma TOF software.
• Perform peak detection, deconvolution, and alignment.
• Identify metabolites using LECO-Fiehn Rtx5 database by matching mass spectra and retention indices.
• Filter peaks: remove those detected in <50% of QC samples or with RSD >30% in QCs.
• Export peak table for statistical analysis.

Statistical Analysis:

• Perform ANOVA to identify significant differences between varieties.
• Conduct unsupervised PCA to observe natural clustering.
• Apply supervised PLS-DA to maximize separation between groups.
• Validate models with cross-validation and permutation tests.
• Select discriminating metabolites with VIP >1.0 and p <0.05.
• Perform pathway analysis using MetaboAnalyst 4.0 to identify impacted metabolic pathways.

Applications in Food Authentication

Geographical Origin Discrimination

The discrimination of geographical origin represents one of the most prominent applications of MS-based metabolomics in food authentication. The comprehensive metabolite profiling capabilities of HRMS platforms enable the detection of subtle chemical differences imparted by environmental factors including soil composition, climate, and agricultural practices.

In a landmark study applying GC-Orbitrap-HRMS to thyme geographical differentiation, researchers successfully distinguished Spanish and Polish thyme samples through comprehensive metabolic profiling [47]. The data processing strategies employed significantly influenced the results, with Compound Discoverer and MS-DIAL software putatively annotating 52 and 115 compounds at Level 2 confidence, respectively. The study highlighted that feature detection is considerably affected by unknown metabolites, background signals, and duplicate features that require careful evaluation before multivariate analysis. Both data processing approaches proved viable for untargeted analysis of GC-Orbitrap-HRMS data, with the selection depending on researcher availability and specific project requirements.

Similarly, comprehensive profiling of roasted hazelnuts from nine geographical regions using HS-SPME and GC×GC-qMS demonstrated the power of advanced GC techniques for geographical discrimination [42]. The two-dimensional GC separation provided enhanced peak capacity, allowing researchers to establish measurable parameters linking volatile profiles to sensory properties and geographical origin. Such approaches are particularly valuable for protecting protected designations of origin (PDO) and ensuring product authenticity in premium food markets.

Food Variety and Quality Differentiation

MS-based metabolomics enables precise differentiation of food varieties and quality grades, providing scientific basis for quality control and preventing economic fraud through variety misrepresentation.

Research on six Indica rice varieties using GC-TOF-MS revealed distinct metabolic profiles that enabled clear variety discrimination [44] [45]. The study identified 221 metabolites classified into amino acids, sugars, organic acids, fatty acids, alcohols, esters, and other compounds. Organic acids (27-33%), amino acids (7-11%), and sugars (10-25%) accounted for the majority of metabolites in all rice varieties. Significant differences in metabolite profiles were observed, with specific varieties showing up-regulation of particular metabolites: phenylalanine and 1,5-anhydroglucitol in NX rice; glycine in YX rice; and lactulose in HM, HY, and MX rice. These metabolic differences not only enabled variety discrimination but also provided insights into potential nutritional implications, demonstrating how metabolomics can inform both authentication and health-related research.

Table 2: Key Metabolite Classes in Food Authentication Studies

Metabolite Class	Analytical Platform	Food Matrix	Authentication Purpose	Key Findings
Amino Acids	GC-MS, LC-MS	Rice, herbs, spices	Variety differentiation, geographical origin	Phenylalanine and glycine content varied significantly between rice varieties [44]
Sugars and Sugar Alcohols	GC-MS	Rice, fruits, cereals	Quality assessment, variety differentiation	1,5-anhydroglucitol and lactulose up-regulated in specific rice varieties [45]
Lipids and Phospholipids	LC-HRMS	Plasma, olive oil, dairy	Dietary pattern assessment, adulteration detection	Lysophospholipids, phosphatidylcholines, and monoacylglycerides identified as Mediterranean diet biomarkers [48]
Polyphenols	LC-HRMS	Fruits, vegetables, herbs	Geographical origin, authenticity	Specific polyphenol profiles enabled food characterization and classification [46]
Volatile Organic Compounds	GC-MS, GC×GC-TOF-MS	Nuts, spices, beverages	Flavor quality, geographical origin	Comprehensive volatile profiling enabled prediction of sensory properties and origin [42]

Dietary Intake Biomarker Discovery

LC-HRMS-based metabolomics has emerged as a powerful approach for discovering biomarkers of food intake, enabling objective assessment of dietary patterns and adherence to specific diets like the Mediterranean diet.

A controlled study investigating MD adherence performed untargeted metabolomics on 135 plasma samples from 58 patients using LC-MS/MS [48] [49]. The strongest association with Mediterranean Diet Score (MDS) was pectenotoxin 2 seco acid, a non-toxic marine xenobiotic metabolite. Several lipids served as useful biomarkers, including eicosapentaenoic acid, a structurally related lysophospholipid, a phosphatidylcholine, and xi-8-hydroxyhexadecanedioic acid. Two metabolites were negatively correlated with MDS. Through stepwise elimination, the researchers selected a panel of three highly discriminatory metabolites and developed a linear regression model that identified high MDS individuals with high sensitivity and specificity [AUC (95% CI) 0.83 (0.76–0.97)].

This study highlights several important aspects of dietary biomarker discovery: the prominence of lipids as discriminatory metabolites, the utility of xenobiotic metabolites as specific intake markers, and the importance of developing optimized biomarker panels rather than relying on single metabolites. Such approaches provide valuable tools for nutritional epidemiology and clinical studies investigating diet-disease relationships.

The Scientist's Toolkit

Essential Research Reagent Solutions

Table 3: Essential Research Reagents for MS-Based Metabolomics

Reagent/Chemical	Function	Application Examples	Technical Considerations
Methoxyamine hydrochloride	Methoximation of carbonyl groups	GC-MS analysis of sugars, organic acids	Prevents ring formation of reducing sugars during derivatization [44]
BSTFA with 1% TMCS	Trimethylsilylation of polar functional groups	GC-MS sample derivatization	Adds volatile TMS groups to -OH, -COOH, -NH groups; TMCS acts as catalyst [45]
Ribitol (internal standard)	Quality control for extraction efficiency	GC-MS metabolomics	Added before extraction to monitor technical variability [44]
Deuterated internal standards	Quantification and quality control	LC-HRMS metabolomics	Corrects for matrix effects and instrument variability
FAMEs mixture	Retention index calibration	GC-MS method setup	Enables calculation of Kovats retention indices for compound identification [45]
LC-MS grade solvents	Mobile phase preparation	LC-HRMS analysis	Minimizes background contamination and ion suppression
Physcion-d3	Physcion-d3, CAS:1215751-27-1, MF:C16H12O5, MW:287.28 g/mol	Chemical Reagent	Bench Chemicals
3-Methyl Hippuric Acid-d7	3-Methyl Hippuric Acid-d7, MF:C10H11NO3, MW:200.24 g/mol	Chemical Reagent	Bench Chemicals

Critical Software and Databases

Effective data processing and metabolite identification rely on specialized software tools and comprehensive databases. The choice between commercial and open-source solutions depends on research resources, objectives, and required level of technical support.

Commercial software such as Compound Discoverer (designed for Orbitrap data processing) offers integrated workflows with instrument systems and dedicated technical support [47]. These platforms typically provide user-friendly interfaces and validated processing parameters, making them accessible to researchers with varying levels of bioinformatics expertise.

Open-source platforms like MS-DIAL have gained popularity due to their flexibility, transparency, and cost-effectiveness [47]. These tools are particularly valuable for method development and customization, though they may require greater bioinformatics expertise for optimal implementation.

For metabolite identification, several databases are essential resources. The LECO-Fiehn Rtx5 database provides comprehensive GC-MS spectra and retention indices for metabolite identification [44]. For LC-HRMS data, platforms like HMDB, MassBank, and mzCloud offer extensive MS/MS spectra for structural annotation [43] [41]. The use of multiple databases and confirmation with authentic standards when possible enhances confidence in metabolite identifications.

The field of MS-based food authentication continues to evolve rapidly, with several emerging trends shaping future research directions. The integration of multiple analytical platforms provides complementary data that enhances metabolome coverage and authentication confidence [41]. Advanced data integration strategies, including data fusion and multiblock analysis, will enable more effective utilization of these complementary datasets.

The implementation of ion mobility spectrometry (IMS) adds a new separation dimension based on molecular size and shape, providing collision cross-section (CCS) values that serve as additional molecular descriptors for confident identification [41]. The combination of IMS with HRMS in platforms like LC-IMS-HRMS increases peak capacity and provides isomeric separation that challenges conventional chromatographic approaches.

Data-independent acquisition (DIA) methods are gaining traction in untargeted metabolomics, providing comprehensive MS/MS data without precursor ion selection [43]. These approaches eliminate the stochastic sampling limitations of data-dependent acquisition and ensure fragmentation data is collected for all detected ions, though at the cost of more complex data interpretation.

Non-targeted metabolomics using advanced MS platforms has established itself as an indispensable approach for food authentication research [47] [41]. The complementary strengths of GC-MS and LC-MS systems, enhanced by the high resolution and mass accuracy of modern Orbitrap and TOF instruments, provide powerful tools for addressing food fraud challenges. As instrumentation continues to advance and data processing strategies become more sophisticated, MS-based metabolomics will play an increasingly vital role in ensuring food integrity, protecting consumers, and supporting regulatory compliance.

Figure 2: MS Platform Selection Logic for Food Authentication

Nuclear Magnetic Resonance (NMR) Spectroscopy for Non-Destructive Analysis

Nuclear Magnetic Resonance (NMR) spectroscopy has emerged as a powerful analytical technique in food authentication research, particularly within the framework of non-targeted metabolomics. As food fraud becomes increasingly sophisticated, the demand for robust, reproducible analytical methods that can comprehensively characterize food matrices has grown substantially. NMR spectroscopy meets this need by providing a non-destructive platform for simultaneously identifying and quantifying a wide range of metabolites without prior knowledge of the food composition. This capability makes it exceptionally valuable for verifying geographical origin, production methods, and authenticity while detecting adulteration in complex food systems.

The fundamental principle of NMR spectroscopy revolves around the magnetic properties of certain atomic nuclei (e.g., ¹H, ¹³C, ³¹P). When placed in a strong magnetic field, these nuclei absorb and re-emit electromagnetic radiation at frequencies characteristic of their molecular environment. This produces spectra that serve as comprehensive metabolic fingerprints of the analyzed sample. The quantitative nature, high reproducibility, and minimal sample preparation requirements of NMR have positioned it as an indispensable tool in modern food analytical laboratories, especially for non-targeted approaches that require holistic analysis of food composition [50] [51].

Quantitative Validation of NMR for Food Authentication

The application of NMR in food authentication requires rigorous validation to ensure analytical robustness. Recent interlaboratory studies have demonstrated the exceptional reproducibility of NMR spectra across different instruments and operators, supporting its use in official control methods.

Table 1: Validation Parameters for NMR in Food Authentication

Validation Parameter	Experimental Findings	Significance for Food Authentication
Inter-laboratory Reproducibility	97.62% correct classification of tomato geographical origin between two independent laboratories [52]	Enables reliable implementation across control laboratories
Spectral Reproducibility	Statistical equivalence of spectra from differently configured spectrometers with optimized acquisition parameters [52]	Ensures consistent data quality regardless of instrumentation
Sample Preparation Robustness	Low relative standard deviation (%RSD = 1.32) for homogenized tomato samples across multiple preparations [52]	Minimizes analytical variance introduced during sample preparation
Multivariate Model Performance	Successful discrimination of Grana Padano cheese from non-PDO competitors using NMR with multivariate analysis [53]	Provides statistical foundation for authentication decisions

The reliability of NMR spectroscopy has been demonstrated through a systematic study involving 63 tomato samples prepared and analyzed independently by two laboratories. The resulting classification model achieved 97.62% correct classification for geographical origin, underscoring the technique's robustness even when different operators perform sample preparation and measurements using their own equipment [52]. This level of reproducibility is crucial for regulatory acceptance and implementation in food control routines.

Experimental Protocols for Food Authentication

Sample Preparation Protocol for Solid Food Matrices

Protocol Title: NMR Sample Preparation from Homogenized Solid Foods (Adapted from Tomato Sample Preparation [52])

Principle: This protocol aims to extract water-soluble metabolites from solid food matrices while maintaining reproducibility and sample integrity. The homogenization approach has demonstrated superior extraction capability and measurement repeatability compared to simple mechanical squeezing or lyophilization.

Materials:

• Liquid nitrogen for flash freezing
• Laboratory homogenizer or grinder
• Centrifuge capable of 14,000 × g
• Lyophilizer (optional)
• NMR tube (5 mm standard)
• Deuterated solvent (Dâ‚‚O with 0.1% TSP-dâ‚„ as internal standard)
• pH meter and buffers

Procedure:

• Homogenization: Flash-freeze the food sample with liquid nitrogen and homogenize to a fine powder using a laboratory grinder.
• Extraction: Weigh 500 mg of homogenized material into a centrifuge tube. Add 1.5 mL of extraction solvent (deuterated phosphate buffer in Dâ‚‚O, pH 4.2).
• Vortexing and Centrifugation: Vortex the mixture vigorously for 1 minute, then centrifuge at 14,000 × g for 15 minutes at 4°C.
• Supernatant Collection: Transfer 600 μL of the supernatant to a standard 5 mm NMR tube.
• Quality Control: Check sample pH and adjust if necessary to ensure spectral consistency.

Critical Notes:

• The pH value should be carefully controlled as it affects the chemical shift position of certain metabolites, particularly glutamine amide protons [52].
• For high-fat content foods, a dual extraction (aqueous and organic phases) may be necessary to capture the full metabolomic profile [53].
• Sample stability should be verified over time, with recommended analysis within 48 hours of preparation.

NMR Acquisition Parameters for Food Metabolomics

Protocol Title: 1D ¹H NMR Spectroscopy for Food Metabolomic Profiling [54] [52]

Instrumentation: High-field NMR spectrometer (≥400 MHz) with temperature control and automated sample changer.

Standard Acquisition Parameters:

• Pulse Sequence: 1D NOESY (noesygppr1d) with presaturation for water suppression
• Spectral Width: 20 ppm
• Number of Scans: 64-128 (depending on sample concentration)
• Relaxation Delay: 4 seconds
• Acquisition Time: 2-4 seconds
• Temperature: 298 K (25°C)
• Data Points: 64k

Processing Parameters:

• Fourier Transformation: Apply with 0.3-1.0 Hz line broadening
• Phase Correction: Manual adjustment for optimal baseline
• Referencing: TSP-dâ‚„ at 0.0 ppm or residual solvent peak
• Baseline Correction: Polynomial function application

Quality Assessment:

• Verify signal-to-noise ratio > 100:1 for quantitative analysis
• Check line width at half height (< 1 Hz for reference peak)
• Confirm proper water suppression without affecting nearby metabolites

Protocol for Non-Targeted Analysis and Data Processing

Protocol Title: NMR Data Processing for Non-Targeted Food Authentication [52] [50]

Workflow:

• Data Reduction: Segment spectra into bins (0.01-0.04 ppm) to account for slight shifts
• Normalization: Apply probabilistic quotient normalization to correct for dilution effects
• Spectral Alignment: Use Icoshift algorithm to correct for misalignments [54]
• Multivariate Analysis: Implement Principal Component Analysis (PCA) and Orthogonal Projections to Latent Structures Discriminant Analysis (OPLS-DA)
• Model Validation: Apply cross-validation and permutation tests to prevent overfitting

Validation Metrics:

• Q² (predictive ability) > 0.5 for robust models
• R²X (goodness of fit) indicating explained variance
• Discrimination accuracy > 90% for authentication purposes

Workflow Visualization

Diagram 1: NMR-based Food Authentication Workflow - This diagram illustrates the comprehensive workflow from sample preparation to authentication decision, highlighting the key stages in NMR-based food analysis.

Application Case Studies in Food Authentication

Geographical Origin Discrimination

The capability of NMR to verify geographical origin has been demonstrated across various food commodities. In a comprehensive study of Italian tomatoes, NMR metabolomics successfully differentiated samples from Lazio and Sicily with 97.62% classification accuracy [52]. The statistical model built from spectroscopic fingerprint data identified specific metabolic patterns characteristic of each growing region, highlighting the technique's precision for Protected Designation of Origin (PDO) verification.

Similarly, NMR spectroscopy has been applied to authenticate Grana Padano cheese, a frequently adulterated Italian product. The analysis distinguished authentic PDO Grana Padano from competitors and non-PDO cheeses through detectable differences in their metabolic profiles, particularly in the aqueous fraction containing amino acids, organic acids, and carbohydrates [53]. This approach proved especially valuable for verifying shredded cheese, where traditional authentication markers like crust imprints are absent.

Adulteration Detection in High-Value Products

Table 2: NMR Applications in Detecting Food Adulteration

Food Product	Adulterant	NMR Detection Method	Key Metabolite Markers
Olive Oil	Hazelnut Oil	¹H NMR Profiling [50]	Absence of linolenic acid and squalene in hazelnut oils
Milk	Whey, Urea, Synthetic Milk	Tâ‚‚ Relaxometry [50]	Increased spin-spin relaxation time (Tâ‚‚) with adulteration
Honey	Sugar Syrups	¹H NMR Profiling [55] [56]	Abnormal carbohydrate profiles, absence of specific botanical markers
Coffee	Robusta in Arabica	¹H NMR Spectroscopy [50]	Detection of 16-O-methylcafestol specific to Robusta
Saffron	Turmeric, Safflower	Benchtop NMR [50]	Characteristic pigment and metabolite profiles

NMR spectroscopy has proven particularly effective for detecting adulteration in high-value food products. For olive oil, ¹H NMR analysis capitalizes on the distinct fatty acid profiles between authentic olive oil and potential adulterants like hazelnut oil. The near absence of linolenic acid and squalene in hazelnut oils provides a clear spectroscopic signature for detection [50]. This application is commercially significant given the premium price of high-quality olive oil and its frequent adulteration with cheaper alternatives.

In dairy products, NMR relaxometry has emerged as a powerful tool for detecting milk adulteration. Studies have demonstrated that the spin-spin relaxation time (Tâ‚‚) significantly increases with the addition of whey, urea, synthetic urine, or synthetic milk [50]. This physical parameter serves as a sensitive indicator of compositional changes resulting from adulteration, enabling rapid screening without extensive sample preparation.

Quality Assessment and Processing Verification

Beyond authentication, NMR spectroscopy provides insights into food quality parameters and processing history. In cheese production, NMR-based metabolomics has been employed to monitor ripening stages in Grana Padano by identifying metabolic "biomarkers" associated with different aging periods [53]. This application addresses potential mislabeling of ripening duration, which directly impacts product value.

Similarly, the effects of alternative processing techniques like bactofugation—a centrifugation step to remove microorganisms and spores—have been investigated using NMR. Comparative analysis revealed that this additional processing step primarily affects the aqueous fraction of cheese, which is responsible for organoleptic properties [53]. Such findings inform regulatory decisions about permitted production methods for PDO products.

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 3: Essential Research Reagents and Materials for NMR-based Food Authentication

Item	Function	Application Notes
Deuterated Solvents (D₂O, CDCl₃)	Provides field frequency lock; suppresses solvent signals	Use with 0.1% TSP-d₄ as internal standard for chemical shift referencing [52]
Internal Standards (TSP-dâ‚„, DSS)	Chemical shift reference; quantitative calibration	TSP-dâ‚„ recommended for aqueous solutions at pH 4.2-7.0 [52]
Buffer Salts (deuterated)	pH control for reproducible chemical shifts	Phosphate buffer preferred for biological pH range; pH critical for amide proton detection [52]
NMR Tubes (5 mm)	Sample containment with precise dimensional tolerance	High-quality tubes essential for reproducible shimming; consider disposable options for screening
Cryoprobes	Signal sensitivity enhancement	Reduces acquisition time; enables detection of low-abundance metabolites [57]
Magic Angle Spinning (MAS) Equipment	Solid/semi-solid sample analysis	Reduces line broadening in heterogeneous samples; essential for intact tissue analysis [50]
Automated Sample Changers	High-throughput analysis	Enables unsupervised operation; critical for large-scale authentication studies [52]
qNMR Reference Standards	Quantitative concentration determination	Certified reference materials with precisely known purity for absolute quantification [50]
Nitrofurazone-13C,15N2	Nitrofurazone-13C,15N2, CAS:1217220-85-3, MF:C6H6N4O4, MW:201.12 g/mol	Chemical Reagent
Tinidazole-d5	Tinidazole-d5, MF:C8H13N3O4S, MW:252.30 g/mol	Chemical Reagent

Advanced Applications and Integration with Artificial Intelligence

The integration of Artificial Intelligence (AI) with NMR spectroscopy represents a transformative advancement in food authentication. Machine learning algorithms, particularly supervised and deep learning approaches, enhance the interpretation of complex NMR spectra by identifying subtle patterns that may elude conventional analysis [57]. This synergy addresses key challenges in food metabolomics, including spectral overlap and the need for rapid classification of multi-dimensional data.

AI-enhanced NMR techniques have demonstrated particular utility in several domains:

• Spectral Interpretation: Machine learning models facilitate precise prediction of molecular structures from NMR data, enabling faster identification of unknown metabolites in complex food matrices [57].

• Adulteration Detection: AI algorithms improve detection limits for sophisticated adulterants by recognizing complex, multi-metabolite patterns indicative of contamination or dilution [57].

• Quality Prediction: Deep learning models, particularly BP-ANN (Backpropagation Artificial Neural Network), have successfully predicted post-thaw quality attributes in frozen foods using LF-NMR relaxation data with superior accuracy compared to traditional multivariate methods [58].

The continued development of AI-NMR hybrid approaches promises to further automate food authentication processes, reducing reliance on expert interpretation while increasing throughput and accuracy. However, challenges remain in standardizing these methodologies and establishing validated protocols for regulatory acceptance [57].

Methodological Considerations and Quality Assurance

NMR Platform Selection

The choice of NMR platform depends on the specific authentication application and required analytical depth:

• High-Field NMR (>400 MHz): Provides superior resolution and sensitivity for comprehensive metabolomic profiling, essential for detecting minor components and structural elucidation [50].

• Low-Field NMR (LF-NMR, 1-80 MHz): Offers rapid, cost-effective screening based on relaxation time measurements, ideal for high-throughput quality assessment and process monitoring [58] [59].

• Benchtop NMR: Emerging as a practical alternative for routine analysis, with applications including saffron authentication and fat content determination [50].

Quality Assurance Protocols

Robust quality assurance is essential for reliable food authentication:

• Instrument Qualification: Regular verification of magnetic field homogeneity, temperature calibration, and signal-to-noise performance [52].

• Method Validation: Establish precision, accuracy, and detection limits for quantitative applications; demonstrate reproducibility across sample preparations [52].

• Data Standardization: Implement consistent processing parameters and referencing to enable spectral comparability across laboratories and studies [52].

The remarkable reproducibility of NMR spectroscopy, evidenced by successful interlaboratory studies, positions it as an increasingly accepted methodology for official food control purposes. As standardized protocols continue to develop, NMR-based authentication is anticipated to play an expanding role in ensuring food authenticity and protecting consumer interests globally [52] [50].

Food authentication verifies the true identity of food ingredients and components, a critical process for maintaining consumer trust, ensuring nutritional quality, and validating label claims in the food industry [60]. Within this field, non-targeted metabolomics has emerged as a powerful discovery tool, capable of screening for unique chemical identifiers across a wide spectrum of food components [60]. When this high-dimensional, complex chemical data is coupled with modern machine learning technologies, it creates a potent framework for identifying novel food identity markers. This Application Note focuses on the application of the Random Forest (RF) classifier within non-targeted metabolomics workflows for food authentication. RF is an ensemble learning algorithm that constructs a multitude of decision trees during training and outputs the mode of the classes (classification) or mean prediction (regression) of the individual trees [61]. Its robustness against overfitting and ability to handle complex, high-dimensional data make it particularly suited for discovering subtle metabolic patterns that differentiate food types and origins, even in processed products where marker compounds may be diluted or transformed.

Key Principles of the Random Forest Algorithm

Random Forest operates on the principle of the "wisdom of crowds," where a large number of relatively uncorrelated models (trees) operating as a committee will outperform any individual model [62]. The algorithm introduces randomness through two primary mechanisms to ensure this low correlation:

• Bagging (Bootstrap Aggregation): Each tree in the forest is trained on a random subset of the original data, drawn with replacement. This means each tree sees a slightly different version of the dataset, forcing diversity in the learned models [62].
• Feature Randomness: When splitting a node during the construction of a tree, the algorithm is restricted to a random subset of features (e.g., the intensities of specific metabolic features). This prevents any single dominant feature from dictating the structure of all trees and ensures that even weak but informative markers contribute to the model [62].

For classification tasks, the final prediction is determined by majority voting across all trees, while for regression, the average prediction is used [61]. A key advantage for research applications is the RF's inherent ability to provide a measure of feature importance, ranking variables (metabolites) based on their contribution to the model's predictive accuracy [60] [61]. This feature extraction capability is central to its utility in metabolic marker discovery.

Application in Food Authentication: A Case Study

To illustrate a practical implementation, we detail a study that utilized RF to distinguish high-value "superfood" seeds—chia, linseed, and sesame—both in their raw state and as ingredients in processed wheat cookies [60].

Experimental Workflow and Design

The study followed a systematic, iterative process for marker discovery and validation [60]:

• Reference Material Selection: 28 independent batches of seeds from 15 EU vendors, encompassing different color variants.
• Chemical Fractionation: Seeds and cookies were subjected to a broad metabolome coverage scheme, analyzing:
  • Volatile Organic Compounds (VOC)
  • Polar Soluble Metabolites (POL)
  • Solid Fraction after Hydrolysis (SOL)
• Non-Targeted Metabolomics Profiling: Each fraction was analyzed using Gas Chromatography-Mass Spectrometry (GC-MS).
• Data Pre-processing: Metabolic features were extracted using TagFinder software, creating a numerical matrix of mass feature abundances normalized by sample weight and an internal standard [60].
• Machine Learning & Marker Discovery: RF analysis was applied for classification and feature extraction.
• Chemical Annotation: Statistically selected markers were annotated using metabolomic databases.

Table 1: Experimental Sample Overview for Seed Authentication Study

Seed Type	Color Variants	Number of Independent Batches	Processing State
Chia (Salvia hispanica L.)	Brown, Off-white	12 (8 brown, 4 off-white)	Non-processed seeds, Wheat cookies
Linseed (Linum usitatissimum L.)	Golden, Brown	8 (4 golden, 4 brown)	Non-processed seeds, Wheat cookies
Sesame (Sesamum indicum L.)	White, Black	8 (4 white, 4 black)	Non-processed seeds, Wheat cookies

The following workflow diagram summarizes the key experimental and computational steps.

Detailed Protocols

Protocol 1: Metabolomic Profiling of Seed Material

Objective: To generate comprehensive metabolic profiles from raw and processed seed ingredients. Materials: See "The Scientist's Toolkit" section for reagents and software. Procedure:

• Sample Homogenization: Grind seed or cookie material to a fine, homogeneous powder under liquid nitrogen using a mortar and pestle or a ball mill.
• Chemical Fractionation:
  • VOC Extraction: Incubate sample headspace at 40°C and trap volatiles using a solid-phase microextraction (SPME) fiber [60].
  • POL Extraction: Extract soluble polar metabolites from ~20 mg of powder with a methanol-water mixture (e.g., 3:1 v/v). Separate phases by centrifugation and collect the supernatant.
  • SOL Preparation: Dry the remaining pellet after POL extraction and subject it to acid or base hydrolysis to release bound metabolites from the solid matrix [60].
• GC-MS Analysis:
  • Derivatize POL and SOL extracts (e.g., using methoxyamination and silylation) to enhance volatility and thermal stability.
  • Inject samples into the GC-MS system using a standardized temperature ramp program.
  • For VOC analysis, thermally desorb the SPME fiber directly in the GC injector.
  • Acquire mass spectra in a suitable scan range (e.g., m/z 40-600).
• Quality Control: Include procedural blanks and a quality control (QC) sample created by pooling aliquots from all samples to monitor instrument performance.

Protocol 2: Data Pre-processing and RF Model Training

Objective: To convert raw GC-MS data into a structured feature table and build a predictive RF model. Materials: TagFinder software, Python/R environment with scikit-learn or an equivalent library. Procedure:

• Data Pre-processing:
  • Use TagFinder or analogous software for peak detection, deconvolution, and alignment across all chromatograms [60].
  • Perform retention index (RI) calibration using an alkane series.
  • Create a data matrix where rows represent samples, columns represent defined mass features (variables characterized by specific mass-to-charge ratio and RI), and values represent normalized peak intensities.
  • Handle missing values (e.g., by imputation with a minimum value) and apply data scaling if necessary (e.g., Pareto or Unit Variance scaling).
• RF Model Training:
  • Partition the data into training and test sets (e.g., 70/30 or 80/20 split).
  • Initialize the RF classifier (e.g., using RandomForestClassifier in scikit-learn). Key hyperparameters include:
    • n_estimators: Number of trees in the forest (e.g., 100-500).
    • max_features: Number of features to consider for the best split (e.g., 'sqrt' or 'log2').
    • max_depth: Maximum depth of the trees.
    • random_state: Seed for reproducibility.
  • Train the RF model on the training set.
  • Tune hyperparameters using techniques like cross-validated grid search or randomized search [63].
• Model Evaluation: Use the held-out test set to evaluate model performance. Calculate metrics such as accuracy, sensitivity, specificity, and F1-score. For the seed-in-cookie model, an overall error rate of 6.7% was achieved [60].

Results and Marker Discovery

The RF model demonstrated high efficacy. It unambiguously classified the original, non-processed seeds. More notably, it successfully identified the presence of seed ingredients within a complex processed food matrix, classifying cookies with an overall error of only 6.7% [60]. This highlights the model's robustness even when unique metabolites are diluted or lost during processing.

The RF's feature importance analysis revealed key identity markers. For instance, the model identified 4-hydroxybenzaldehyde as a marker for chia and succinic acid monomethylester for linseed additions in cookies [60]. These compounds may represent original seed metabolites or processing-induced transformation products.

Table 2: Key Metabolite Markers Identified by Random Forest in Food Authentication Studies

Study Context	Identified Marker Metabolites	Remarks on Biological Relevance	Classification Performance
Chia, Linseed & Sesame Seeds [60]	Sesamol, Rosmarinic acid (Chia), 4-hydroxybenzaldehyde (Chia in cookies), Succinic acid monomethylester (Linseed in cookies)	Secondary metabolites characteristic of plant family (e.g., Lamiaceae for Chia); Processing markers.	100% for raw seeds; 93.3% accuracy for cookies
Lymph Node Tuberculosis [64]	Leu-Ala, Evodiamine, Fenazaquin, Acetol	Dysregulation in amino acid and tRNA biosynthesis pathways.	Leu-Ala AUC = 0.83
Polycystic Ovary Syndrome (PCOS) [63]	L-Histidine, L-Glutamine, L-Tyrosine	Alterations in amino acid metabolism.	86% Accuracy, 91% Specificity

The Scientist's Toolkit

Table 3: Essential Research Reagent Solutions and Software for RF-Metabolomics

Item	Function/Application	Example/Note
GC-MS System	High-resolution separation and detection of volatile and derivatized metabolites.	Equipped with a non-polar capillary GC column (e.g., DB-5).
SPME Fibers	Extraction and concentration of Volatile Organic Compounds (VOCs) from sample headspace.	Various coatings (e.g., DVB/CAR/PDMS) for different compound classes [60].
Derivatization Reagents	Chemical modification of polar metabolites to increase volatility for GC-MS analysis.	Methoxyamine hydrochloride and N-Methyl-N-(trimethylsilyl)trifluoroacetamide (MSTFA) [60].
Retention Index Markers	Calibration of retention times to a standardized scale for cross-sample comparison.	A homologous series of n-alkanes (e.g., C8-C40) [60].
TagFinder Software	Pre-processing of GC-MS data: peak detection, deconvolution, and alignment to create a data matrix [60].
Python/R with ML Libraries	Programming environment for implementing Random Forest and other statistical analyses.	Scikit-learn (Python) or randomForest (R) packages [61] [63].
Metabolomic Databases	Annotation of statistically significant mass features with putative chemical identities.	NIST, Golm Metabolome Database, HMDB.
(R)-Norfluoxetine-d5	(R)-Norfluoxetine-d5, CAS:1217648-64-0, MF:C16H16F3NO, MW:300.33 g/mol	Chemical Reagent
AGN 193109-d7	AGN 193109-d7, MF:C28H24O2, MW:399.5 g/mol	Chemical Reagent

Critical Analysis and Best Practices

Advantages and Limitations

• Advantages: RF is highly accurate, handles high-dimensional data well, is robust to overfitting, and provides native feature importance rankings, making it excellent for biomarker discovery [60] [61].
• Limitations: The model can be computationally expensive with large numbers of trees and is often viewed as a "black box," though techniques like SHAP (SHapley Additive exPlanations) can improve interpretability [61] [63]. The cited study also noted that RF can be "overdesigned" for simple classification tasks where unique secondary metabolites (e.g., sesamol in chia) already provide unambiguous markers [60].

Recommendations for Implementation

• Data Quality is Paramount: The performance of the RF model is contingent on high-quality metabolomic data. Rigorous quality control during sample preparation and instrumental analysis is non-negotiable.
• Human Supervision is Crucial: RF-based feature extraction should not be used without expert supervision. The selected markers must be evaluated in the context of biological and food chemistry knowledge to avoid spurious correlations [60].
• Combine Analytical Approaches: For comprehensive food authentication, RF-driven metabolomics should be combined with other data analysis technologies and, potentially, genetic or protein-based markers to create a robust, multi-layered verification system [60].
• Model Interpretation: Employ model interpretation frameworks like SHAP to understand the direction and magnitude of each metabolite's contribution, transforming a predictive model into an explanatory tool [63]. The following diagram illustrates the RF decision process and its interpretation.

The integration of non-targeted metabolomics with the Random Forest machine learning algorithm provides a powerful, robust, and discovery-oriented framework for food authentication. Its ability to sift through complex chemical profiles to identify subtle, yet significant, identity markers—even in processed foods—makes it an invaluable tool for researchers and regulatory bodies. Future work should focus on expanding these methodologies to a wider range of food ingredients and processing techniques, further validating the robustness of discovered markers, and enhancing model interpretability to fully unlock the potential of machine learning in ensuring food authenticity and safety.

Application Note

This application note details the implementation of non-targeted metabolomics to address two critical challenges in food authentication: verifying the presence of specific seed ingredients in processed foods and determining the geographical origin of agricultural commodities. These case studies, framed within doctoral research on non-targeted metabolomics, demonstrate the methodology's power in ensuring food integrity and traceability.

Case Study 1: Authentication of Seed Ingredients in Processed Bakery Products

1.1 Research Context and Challenge Food fraud involving high-value seeds such as chia, linseed, and sesame is a growing concern, especially when these ingredients are incorporated into processed foods where visual identification becomes impossible [60]. This study established a standardized procedure to discover metabolic markers that authenticate the presence of these seed ingredients in raw materials and finished products (wheat cookies) [60].

1.2 Key Findings Non-targeted metabolomics successfully differentiated non-processed chia, linseed, and sesame seeds by analyzing their volatile organic compounds (VOCs), polar metabolites (POL), and solid fraction metabolites (SOL) [60]. The random forest (RF) machine learning algorithm classified the seeds with high accuracy. Furthermore, despite the dilution or loss of distinctive markers during cookie production, the model could still identify the presence of seed ingredients in the processed cookies with a 93.3% overall accuracy (6.7% error) [60]. Key processing markers were identified, including 4-hydroxybenzaldehyde for chia and succinic acid monomethylester for linseed additions [60].

Table 1: Authentication Performance for Seed Ingredients in Processed Cookies

Seed Ingredient	Classification Accuracy in Cookies	Identified Processing Marker(s)
Chia	High	4-hydroxybenzaldehyde
Linseed	High	Succinic acid monomethylester
Sesame	High	Not Specified
Overall Model	93.3%

Case Study 2: Geographic Origin Authentication of Grain Maize

2.1 Research Context and Challenge The global trade of grain maize makes it vulnerable to fraudulent misrepresentation of its geographical origin, which can impact quality and safety [65] [66]. This study aimed to develop an analytical method to trace the origin of grain maize samples, moving beyond reliance on shipping documents that can be falsified [65] [66].

2.2 Key Findings A non-targeted UHPLC-ESI-qToF metabolomics approach analyzed 151 grain maize samples from seven countries [65] [66]. Multivariate data analysis revealed that the non-polar metabolome (lipid fraction) was highly informative for origin discrimination [65]. Twenty selected lipid markers, identified as triglycerides, diglycerides, and phospholipids, were used to build a Random Forest classification model [65] [66]. The model achieved 90.5% accuracy in classifying samples based on geographical origin using repeated cross-validation [65] [66]. The marker set was also highly effective in one-vs-rest classification scenarios, yielding accuracies above 89% [65].

Table 2: Classification Performance for Grain Maize Geographic Origin

Statistical Model	Number of Marker Metabolites	Classification Accuracy	Key Metabolite Classes Identified
Random Forest	20	90.5% (100x repeated 10-fold cross-validation)	Triglycerides, Diglycerides, Phospholipids

Experimental Protocols

Protocol 1: Non-Targeted Metabolomics for Seed Authentication in Processed Food

1. Sample Preparation and Metabolite Extraction

• Reference Material Selection: Obtain authenticated, non-processed seeds from multiple independent vendors. For processed food, prepare the product (e.g., wheat cookies) with and without the target seed ingredient [60].
• Chemical Fractionation: Employ a broad fractionation scheme to cover diverse metabolites [60]:
  • Volatile Organic Compounds (VOCs): Analyze using headspace sampling.
  • Polar Metabolites (POL): Extract soluble polar compounds with aqueous methanol.
  • Solid Fraction (SOL): After exhaustive polar and lipophilic extraction, hydrolyze the remaining solid to release bound metabolites [60].
• Lipid Fraction: Omitted in the cited study due to abundant ubiquitous fatty acids that can be obscured by processing [60].

2. Instrumental Analysis – GC-MS Profiling

• Platform: Gas Chromatography-Mass Spectrometry (GC-MS).
• Chromatography: Use standard non-polar GC columns with retention index calibration for the POL and SOL fractions [60].
• Mass Spectrometry: Acquire data in full-scan mode to enable non-targeted profiling [60].

3. Data Processing and Marker Discovery

• Data Pre-processing: Use software to perform peak picking, alignment, and deconvolution, generating a data matrix of mass features (variables) [60].
• Marker Discovery: Apply Random Forest analysis for feature extraction and classification. For simpler cases, a rule-based "Min/Max ratio" approach can identify unique or highly enriched markers [60].

Protocol 2: Non-Targeted LC-MS for Geographic Origin Authentication

1. Sample Treatment and Lipid Extraction

• Milling and Homogenization: Grind grain maize samples to a fine powder using an ultra-centrifugal mill and homogenize thoroughly [65] [66].
• Lyophilization: Freeze-dry the ground samples for 24 hours to remove moisture [65] [66].
• Lipid Extraction: Weigh 50.0 mg of lyophilizate. Perform a modified Bligh and Dyer extraction using a ball mill with a chloroform/methanol/water mixture [65] [66]. Centrifuge to induce phase separation, collect the organic phase, dilute, and centrifuge again before LC-MS analysis [65] [66].

2. Instrumental Analysis – UHPLC-ESI-QToF-MS

• Chromatography:
  • System: UHPLC system.
  • Column: Reversed-phase column.
  • Mobile Phase: A) Water with 10 mmol/L ammonium formate; B) Isopropanol/acetonitrile (3/1 v/v) with 10 mmol/L ammonium formate [65] [66].
  • Gradient: Start at 55% B, ramp to 100% B over 18 minutes, hold, and re-equilibrate [65].
• Mass Spectrometry:
  • System: Electrospray Ionization Quadrupole Time-of-Flight (ESI-QToF).
  • Mode: Data-independent acquisition (DIA) or data-dependent acquisition (DDA) in positive and/or negative ion modes for broad metabolome coverage [65] [67].

3. Data Analysis and Model Building

• Feature Detection: Use non-targeted processing software for peak picking, alignment, and componentization.
• Multivariate Statistics: Apply unsupervised (PCA) and supervised (PLS-DA) methods to observe grouping and select significant features [65] [66].
• Marker Identification: Statistically select features (e.g., p-value, fold-change, VIP score) and identify them using MS/MS fragmentation and database matching [65].
• Classification Modeling: Use machine learning algorithms like Random Forest with repeated k-fold cross-validation to build and validate the classification model [65] [66].

The Scientist's Toolkit: Essential Research Reagents & Materials

Table 3: Key Reagents and Materials for Non-Targeted Metabolomics in Food Authentication

Item Name	Function / Application	Example from Case Studies
Ultra-Centrifugal Mill	Homogenization of solid samples into a fine, consistent powder.	Grinding of grain maize and seeds prior to extraction [65] [66].
Ball Mill (BeadRuptor)	Efficient mechanical disruption of cells for metabolite extraction.	Used in the Bligh & Dyer lipid extraction protocol for grain maize [65] [66].
Bligh & Dyer Reagents	Extraction of non-polar metabolites (lipids).	Chlorform/methanol/water mixture for lipidome analysis of grain maize [65] [66].
LC-MS Grade Solvents	High-purity solvents for LC-MS mobile phases and extractions to minimize background noise and ion suppression.	Acetonitrile, methanol, isopropanol, chloroform [65] [66].
Ammonium Formate	LC-MS mobile phase additive that improves ionization efficiency and chromatographic peak shape.	Added to both mobile phases in UHPLC analysis of maize lipids [65] [66].
UHPLC with C18 Column	High-resolution separation of complex metabolite mixtures prior to mass spectrometry.	Separation of lipid classes in grain maize using a reversed-phase column [65].
High-Resolution Mass Spectrometer	Accurate mass measurement for putative metabolite identification and MS/MS fragmentation for structural confirmation.	ESI-Quadrupole-Time-of-Flight (QToF) used in both case studies [65] [60].
Retention Index Standards	Calibration of retention times to account for analytical drift, crucial for GC-MS.	Used in GC-MS profiling of seed metabolites for reliable identification [60].

Navigating Challenges and Ensuring Analytical Robustness

Non-targeted metabolomics has emerged as a powerful hypothesis-free approach for food authentication, capable of verifying geographical origin, production methods, and detecting adulteration by providing a comprehensive molecular fingerprint of food samples [68]. The fundamental premise is that the metabolic network of a food product is highly sensitive to exogenous factors such as climate, soil composition, and anthropogenic influences, creating distinguishable chemical profiles even in closely related materials [68]. However, the analytical pathway from sample collection to biological interpretation is fraught with critical pitfalls that can compromise data quality, reproducibility, and ultimately, the validity of authentication claims. This application note details these pitfalls within the context of food authentication research and provides standardized protocols to enhance data reliability and cross-study comparability.

Critical Pitfalls in Sample Preparation and Experimental Design

Sample Variation and Representative Sampling

The first major pitfall occurs before any analytical instrumentation is touched: inadequate attention to sample variation and representative sampling. The chemical composition of plant-based foods varies significantly based on seed colour variants, cultivation conditions, and post-harvest processing [60]. For instance, a study aiming to distinguish chia, linseed, and sesame seeds took care to include multiple colour variants (e.g., brown and off-white chia seeds; golden and brown linseeds) from 15 different vendors within the European Union to account for natural metabolic variation [60]. Without this intentional capture of biological variation, discovered markers may reflect batch-specific artifacts rather than genuine authentication signatures.

Protocol 2.1: Representative Sample Collection for Food Authentication

• Source Authentication: Obtain reference materials from multiple independent, authenticated sources (minimum 3-5 independent sources per food type) [60].
• Biological Replicates: Include a minimum of 5-6 analytical replicates per sample type to account for biological variation.
• Color/Strain Variants: Deliberately include common colour variants or strains marketed for human consumption.
• Documentation: Meticulously record complete vendor information, geographical origin when available, and visual characteristics [60].

Chemical Fractionation and Metabolome Coverage

A second critical pitfall involves inadequate metabolome coverage. No single extraction method can capture the full chemical diversity of food matrices. Research has demonstrated that distinctive authentication markers may reside in different chemical fractions - volatile organic compounds (VOCs), polar metabolites, or even components of the solid food fraction after hydrolysis [60]. A study on seed authentication found that most unique metabolites were diluted or lost during food processing, emphasizing the need for comprehensive fractionation to retain marker detectability in processed foods [60].

Protocol 2.2: Comprehensive Metabolome Fractionation

• Volatile Organic Compounds (VOCs): Profile using headspace solid-phase microextraction (HS-SPME) coupled with GC-MS [60].
• Polar Metabolites: Extract using methanol/water mixtures followed by GC-MS analysis of derivatized extracts [60].
• Solid Food Fractions: Hydrolyse residual solid material after exhaustive extraction to access bound metabolites [60].
• Lipid Fractions: While often omitted from marker searches due to ubiquitous fatty acids, consider targeted lipid analysis when relevant to authentication claims [60].

Table 1: Critical Pitfalls in Sample Preparation and Experimental Design

Pitfall Category	Specific Risk	Impact on Data Quality	Mitigation Strategy
Sample Representation	Limited biological variation	Non-generalizable markers; batch-specific artifacts	Include multiple sources (≥3), colour variants, documented provenance
Metabolome Coverage	Single extraction method	Missed authentication markers in specific fractions	Implement multi-fraction approach: VOCs, polar, solid, and lipid fractions
Sample Documentation	Incomplete metadata	Limited utility for database building	Record vendor, origin, processing history, visual characteristics
Processing Effects	Ignoring food processing	Markers lost in processed foods	Analyze both raw ingredients and processed forms

Data Acquisition Pitfalls and Standardization Strategies

Analytical Variability and the Internal Standard Solution

The reproducibility of non-targeted metabolomics data across laboratories and instrumentation platforms represents a fundamental challenge that has limited the utility of these technologies for expanding food composition databases [38]. Liquid chromatography-mass spectrometry (LC-MS) platforms are particularly susceptible to retention time shifting across laboratories, instruments, and even consecutive days, complicating data alignment and comparison.

A groundbreaking solution to this pitfall is the implementation of a novel internal retention time standard (IRTS) mixture containing compounds non-endogenous to food, which enables robust chromatographic alignment of data across laboratories [38]. The PTFI Nontargeted Metabolomics Platform has pioneered this approach, incorporating a unique internal standard reagent comprising 33 compounds not found naturally in foods [25]. When researchers worldwide employ this standardized protocol and internal standard reagent, the resulting data can be harmonized and becomes comparable, enabling the construction of scalable data resources for food authentication [38] [25].

Protocol 3.1: Cross-Laboratory Standardized LC-MS Analysis

• Internal Standard: Incorporate IRTS mixture containing 33 non-endogenous compounds before extraction [38] [25].
• Chromatography: Utilize reverse phase liquid chromatography with standardized gradient elution.
• Mass Spectrometry: Employ high-resolution mass spectrometry (HRMS) for accurate mass detection [25].
• Ionization: Collect data in both positive and negative electrospray ionization (ESI) modes to maximize metabolite coverage [68].

Platform Selection and Method Optimization

The selection of analytical platforms presents another pitfall, as different technologies access complementary portions of the metabolome. While GC-MS platforms are excellent for volatile and derivatized polar compounds, they limit detectable metabolite coverage to vaporizable analytes [68]. LC-MS methods are more suitable for non-polar to medium-polar compounds, with ultra-high-performance liquid chromatography (UHPLC) systems providing superior separation efficiency [68]. High-resolution mass analyzers like Time-of-Flight (TOF) or Orbitrap instruments are particularly suitable for non-targeted metabolomics due to their high mass accuracy and scan rates [68].

Table 2: Analytical Platform Comparison for Food Authentication

Platform	Optimal Metabolite Classes	Key Advantages	Limitations for Food Authentication
GC-MS	VOCs, polar metabolites after derivatization	High chromatographic reproducibility; extensive spectral libraries	Limited to vaporizable compounds; derivatization required
LC-MS (RP)	Medium to non-polar metabolites	Broad coverage of secondary metabolites; no derivatization	Poor retention of highly polar compounds
HILIC-MS	Polar metabolites	Complementary to RP-LC; retains polar compounds	Less stable retention times than RP-LC
HRMS (Orbitrap, TOF)	Comprehensive screening	High mass accuracy; unknown identification capability	Higher cost; complex data handling

Data Preprocessing: From Raw Data to Biological Meaning

Peak Processing and Alignment Challenges

The conversion of raw instrument data into a quantitative feature table represents perhaps the most treacherous pitfall in non-targeted metabolomics. Inconsistent peak picking, alignment errors, and inadequate missing value handling can introduce technical artifacts that obscure true biological variation. This process involves multiple critical steps: noise filtering, peak picking and deconvolution, peak identification, peak alignment, and creation of a final data matrix for statistical processing [69]. The complexity of these steps necessitates robust, standardized processing workflows.

The integration of internal retention time standards (IRTS) directly addresses the alignment challenge by providing stable anchor points across datasets [38]. Research has demonstrated that this approach enables qualitative consensus of features across laboratories and/or instrumentation, establishing the foundation for comparable, non-targeted omics analysis to support the next generation of food composition data [38].

Protocol 4.1: Standardized Data Preprocessing Workflow

• Peak Picking: Use algorithms such as those in MetaboAnalystR 4.0 or the asari algorithm, optimizing parameters for signal-to-noise ratio and peak width [70].
• Retention Time Alignment: Leverage IRTS compounds as alignment anchors to correct for chromatographic shifts [38].
• Missing Value Handling: Apply advanced imputation methods such as quantile regression imputation of left-censored data (QRILC) or MissForest for bounded missing values [70].
• Data Filtering: Set group-wise thresholds to filter features based on missing values, removing features with >80% missingness in any group [70].

Normalization and Quality Control

Inadequate normalization represents a subtle but devastating pitfall that can introduce systematic biases, particularly in food authentication studies where sample matrices may vary significantly. Modern platforms like MetaboAnalyst offer multiple normalization options including Log2 transformation and variance stabilizing normalization, which should be selected based on data characteristics [70]. Additionally, implementing rigorous quality control procedures is essential for detecting technical biases before they propagate through downstream analysis.

Protocol 4.2: Quality Assurance and Normalization

• Quality Control: Inject pooled quality control samples (QC) every 4-6 injections to monitor instrument stability.
• Data Integrity Checks: Utilize diagnostic graphics for missing values and RSD distributions to identify potential technical biases [70].
• Normalization: Apply variance-stabilizing normalization followed by log transformation to address heteroscedasticity [70].
• Data Scaling: Employ mean-centering and Pareto scaling to balance the influence of high and low-abundance features in multivariate models.

Non-Targeted Metabolomics Workflow with Critical Pitfalls

The Authentication Toolbox: Statistical Analysis and Marker Discovery

Multivariate Statistics and Machine Learning Applications

The statistical analysis phase presents pitfalls related to both underutilization and overreliance on sophisticated algorithms. For simple authentication tasks where unique secondary metabolites exist (e.g., sesamol in chia), machine learning may be overdesigned [60]. However, for complex authentication challenges such as detecting dog meat adulteration in beef meatballs at 0.1% levels, advanced chemometric approaches like partial least squares-discriminant analysis (PLS-DA) become essential [71].

Random Forest (RF) machine learning with its inherent feature extraction capability has shown promise for non-targeted metabolic marker discovery, successfully classifying seed ingredients in processed cookies with 6.7% overall error despite dilution or loss of unique metabolites during processing [60]. However, RF-based feature extraction requires human supervision, and combination with alternative data analysis technologies is advised [60].

Protocol 5.1: Statistical Analysis Workflow for Food Authentication

• Exploratory Analysis: Begin with Principal Component Analysis (PCA) to identify natural clustering and detect outliers.
• Supervised Modeling: Apply PLS-DA or OPLS-DA to maximize separation between predefined classes (e.g., authentic vs adulterated) [71].
• Feature Selection: Utilize Random Forest with cross-validation to identify discriminative features, but verify selected markers with univariate statistics.
• Validation: Implement strict train-test splits or cross-validation to prevent overfitting; use permutation testing to assess model significance.

Metabolite Identification and Pathway Analysis

A common pitfall in food authentication is the failure to progress from discriminative features to biologically meaningful markers. While unknown features can still serve as authentication markers, identified compounds provide stronger evidence and facilitate understanding of biological differences. Modern platforms like MetaboAnalyst support functional analysis of untargeted metabolomics data through approaches like mummichog or GSEA algorithms, which can identify activated pathways without complete compound identification [70].

Protocol 5.2: Marker Identification and Validation

• Database Searching: Query accurate mass and retention time against food-specific metabolite databases.
• MS/MS Confirmation: Perform targeted MS/MS fragmentation on potential markers and compare with standards or spectral libraries [70].
• Pathway Analysis: Input annotated markers into pathway analysis tools to identify affected biological processes [71].
• Validation: Confirm marker stability across multiple batches and analytical conditions; verify in independent sample sets.

Data Analysis Pathway for Food Authentication

Essential Research Reagent Solutions for Food Metabolomics

Table 3: Essential Research Reagent Solutions for Food Authentication Studies

Reagent Category	Specific Products/Composition	Function in Workflow	Performance Metrics
Internal Retention Time Standard	IRTS mixture of 33 compounds non-endogenous to food [38] [25]	Chromatographic alignment across laboratories and instruments	Enables qualitative consensus of features across platforms [38]
Chemical Derivatization Reagents	MSTFA (N-Methyl-N-(trimethylsilyl)trifluoroacetamide) for GC-MS [60]	Volatilization of polar metabolites for GC-MS analysis	Enables detection of amino acids, organic acids, sugars
Solid Phase Extraction Cartridges	C18, HILIC, mixed-mode sorbents [25]	Fractionation of complex food extracts; metabolite enrichment	Improves metabolome coverage; reduces ion suppression
Quality Control Materials	Pooled sample aliquots from all study samples	Monitoring of instrument performance; data normalization	RSD < 30% for QC pool features indicates stable performance
Authentication Standards	Reference compounds for suspected markers (e.g., sesamol for chia) [60]	Confirmation of marker identity; quantitative calibration	Enables transition from non-targeted to targeted analysis

Non-targeted metabolomics offers unprecedented potential for food authentication but requires meticulous attention to critical pitfalls throughout the analytical workflow. From intentional capture of biological variation during sample collection to implementation of standardized internal retention time standards for cross-laboratory comparability, each step demands careful execution and validation. The protocols and guidelines presented here provide a framework for minimizing technical variability while maximizing biological discovery, ultimately supporting the development of robust, validated authentication methods that can protect consumers and ensure fair trade practices in global food systems.

Impact of Food Processing on Metabolic Marker Stability

In the field of food authentication, non-targeted metabolomics has emerged as a powerful tool for assessing food quality, origin, and processing history. The stability of metabolic markers—small molecule metabolites that serve as chemical fingerprints for food identity—is critically influenced by food processing techniques and conditions. Thermal and non-thermal processing methods significantly alter the metabolic profile of food products, affecting the reliability of authentication markers. Understanding these impacts is essential for developing robust analytical methods that can accurately verify food authenticity despite processing-induced chemical changes. This Application Note examines the effects of various processing technologies on metabolic marker stability and provides detailed protocols for identifying and validating stable markers in processed food matrices, framed within the broader context of non-targeted metabolomics research for food authentication.

Effects of Processing Techniques on Metabolic Profiles

Food processing techniques induce significant changes in the metabolic profiles of food products, which can either degrade existing markers or generate new process-induced markers. The table below summarizes the effects of different processing methods on metabolic stability in various food matrices.

Table 1: Impact of Food Processing Techniques on Metabolic Marker Stability

Processing Technique	Food Matrix	Key Metabolic Changes	Identified Markers	Reference
Thermal Treatment (Standard)	Strawberry Puree	Temperature-dependent metabolic profile changes; apparent thermal effect	Pyroglutamic acid, Pteroyl-D-glutamic acid, 2-hydroxy-5-methoxy benzoic acid, 2-hydroxybenzoic acid β-d-glucoside	[72]
Thermal Treatment (Standard)	Apple Puree	Formation of new compounds; degradation of heat-sensitive metabolites	Di-hydroxycinnamic acid glucuronide, Caffeic acid, LysoPE(18:3(9Z,12Z,15Z)/0:0)	[72]
Vacuum Concentration	Strawberry Puree	Increased marker concentration; enhanced thermal degradation effects	Same as thermal treatment markers but with stronger intensity	[72]
High-Pressure Processing (HPP)	Apple Puree	Minimal changes to metabolic profile; preservation of fresh-like markers	Similar to fresh apple with minimal alterations	[72]
Baking Process	Seed-Enriched Cookies	Dilution or loss of unique secondary metabolites; formation of processing markers	4-hydroxybenzaldehyde (chia), Succinic acid monomethylester (linseed)	[60]

Multivariate analysis of processed fruit purees revealed that samples cluster according to processing type, demonstrating a clear temperature-dependent effect on metabolic profiles [72]. In strawberry purees, the discrimination models showed significant differences between fresh samples and those subjected to vacuum concentration, while samples undergoing cold-crushing with no heat treatment or mild heat treatment showed similar metabolic profiles [72]. This highlights the crucial role of thermal energy in modifying the food metabolome.

For apple purees, the application of high-pressure processing resulted in metabolic profiles more closely resembling fresh apples compared to thermally processed samples [72]. This demonstrates the potential of non-thermal technologies to better preserve the native metabolic signature of food products, which is particularly valuable for authentication purposes.

Detailed Experimental Protocols

Protocol 1: Untargeted Metabolomics for Processing Marker Discovery

This protocol describes the comprehensive workflow for identifying processing-specific metabolic markers in food products, based on established methodologies from the literature [72] [60].

Sample Preparation:

• Obtain authenticated reference materials from multiple independent sources (minimum 4-8 batches per food type) [60].
• For solid foods (seeds, fruits), implement a chemical fractionation scheme covering volatile organic compounds (VOC), polar soluble compounds (POL), and solid fraction components (SOL) [60].
• Homogenize samples under controlled conditions (temperature: 4°C, time: 2-5 minutes).
• For liquid samples (purees, juices), aliquot 100μL for analysis; for solid samples, use 100mg dry weight [73].
• Add internal standards prior to extraction: heptanoic methyl ester for fatty acid analysis or anthranilic acid C13 for other metabolites [18].

Metabolite Extraction:

• For comprehensive coverage, use dual extraction protocols: methanol-water (80:20 v/v) for polar metabolites and methyl-tert-butyl ether for lipophilic compounds [73].
• Perform extraction at 4°C with continuous shaking for 15 minutes.
• Centrifuge at 14,000 × g for 12 minutes at 4°C [74].
• Collect supernatant and filter through 0.22μm syringe filter [74].
• Dry under nitrogen stream and reconstitute in mobile phase compatible with LC-MS analysis.

LC-MS Analysis:

• Utilize ultra-performance liquid chromatography coupled to quadrupole time-of-flight mass spectrometry (UPLC-ESI-QTOF-MS) [72].
• Employ reverse-phase chromatography (C18 column) with gradient elution using solvent A (0.1% formic acid in water) and solvent B (0.1% formic acid in acetonitrile) [72] [73].
• Set flow rate to 0.25 mL/min with column temperature maintained at 21±2°C [74].
• Use electrospray ionization in both positive and negative modes.
• Set mass range to m/z 50-1200 with resolution >30,000.
• Include quality control samples: blanks, solvent samples, pooled QC samples from all samples, and reference standards [73] [18].

Data Processing:

• Perform peak picking, alignment, and integration using XCMS, MZmine, or similar platforms [75].
• Apply quality assessment including total ion current inspection, principal component analysis, correlation analysis, and coefficient of variation distribution [73] [18].
• Use TurboPutative web server for data handling and metabolite classification: execute Tagger, REname, RowMerger, and TPMetrics modules sequentially [75].
• Reduce data complexity by 80-90% through automated curation before manual inspection [75].

Multivariate Statistical Analysis:

• Perform Principal Component Analysis (PCA) to identify natural clustering of samples based on processing techniques.
• Apply Partial Least Squares-Discriminant Analysis (PLS-DA) to maximize separation between processing groups.
• Calculate Variable Importance in Projection (VIP) scores to identify metabolites contributing most to sample separation.
• Set VIP >1.0 as threshold for significant markers [72].
• Validate models through cross-validation (e.g., Q² value >0.6 indicates robust model) [72].

Protocol 2: Validation of Marker Stability During Processing

This protocol describes the procedure for validating the stability of identified metabolic markers across different processing conditions.

Controlled Processing:

• Apply different processing techniques to the same raw material: no heat treatment, mild heat treatment, standard thermal treatment, vacuum concentration, and high-pressure processing [72].
• For thermal treatments, use temperature range of 70-95°C for 1-10 minutes.
• For high-pressure processing, apply 400-600 MPa for 3-5 minutes at room temperature.
• Include reprocessing conditions (e.g., reprocessed mild heat treatment, reprocessed standard thermal treatment) to simulate industrial conditions [72].

Stability Assessment:

• Analyze processed samples using the untargeted metabolomics protocol above.
• Monitor intensity changes of candidate markers across processing conditions.
• Identify markers that consistently appear despite processing (robust authentication markers).
• Identify markers that specifically indicate processing history (processing degree markers).
• Calculate fold-changes and statistical significance (p-value <0.05) between different processing conditions.

Compound Identification:

• Confirm identity of significant markers using accurate mass (mass error <5 ppm) and MS/MS fragmentation pattern [72].
• Compare retention time and fragmentation spectrum with authentic standards when available [72].
• Search against metabolic databases (HMDB, KEGG, METLIN) with mass tolerance of 0.001 Da.

Analytical Workflow and Data Processing Strategies

The following diagram illustrates the comprehensive workflow for evaluating metabolic marker stability in processed foods:

Diagram 1: Comprehensive Workflow for Evaluating Metabolic Marker Stability in Processed Foods

Data Processing Strategies: Effective data processing is essential for obtaining meaningful results from metabolomics data. The following strategies have been validated for processing data related to food processing markers:

• Normalization: Apply cross-contribution compensating multiple standard normalization (ccmn) using internal standards to remove unwanted systematic variation while preserving biological information [18]. For well-controlled studies, ccmn followed by square root transformation produces optimal results [18].

• Transformation: Implement data transformation methods to reduce skewness and correct heteroscedasticity. Square root (sqrt) transformation is particularly effective for food metabolomics data, followed by generalized log (glog) and cube root transformations [18].

• Scaling: Apply scaling methods to reduce fold differences between metabolite concentrations. Pareto scaling, range scaling, and vast scaling have shown effectiveness for food metabolomics datasets [18].

The combination of ccmn normalization followed by square root transformation has been demonstrated to produce processed data most similar to absolute quantified data, enabling more accurate biological interpretation [18].

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 2: Essential Research Reagents and Materials for Metabolic Marker Stability Studies

Category	Item	Specification	Application/Function
Chromatography	UPLC-ESI-QTOF-MS System	High-resolution (>30,000) with electrospray ionization	Comprehensive metabolite profiling [72] [73]
Chromatography	C18 Reverse-Phase Column	50-100mm × 2.1mm, 1.8μm particle size	Metabolite separation [74]
Internal Standards	Heptanoic Methyl Ester	GC-MS grade, >99% purity	Fatty acid analysis normalization [18]
Internal Standards	Anthranilic Acid C13	Isotopically labeled, >98% purity	General metabolomics normalization [18]
Solvents	LC-MS Grade Acetonitrile	LC-MS grade, with 0.1% formic acid	Mobile phase for LC-MS [74]
Solvents	LC-MS Grade Water	LC-MS grade, with 0.1% formic acid	Mobile phase for LC-MS [74]
Solvents	LC-MS Grade Methanol	LC-MS grade, >99.9% purity	Metabolite extraction [73]
Software	TurboPutative Platform	Web server with Tagger, REname, RowMerger, TPMetrics modules	Data handling and metabolite classification [75]
Software	Metabox 2.0	R package for metabolomic data analysis	Data processing, biomarker analysis, integrative analysis [18]
Reference Materials	Authentic Metabolite Standards	>95% purity, certified reference materials	Compound identification and validation [72]

Application to Food Authentication

The stability assessment of metabolic markers across processing conditions enables several practical applications in food authentication:

1. Processing History Verification: Metabolic markers can objectively determine the degree and type of processing applied to food products. For instance, the ratio of pyroglutamic acid to native glutamic acid can indicate thermal processing intensity in fruit products [72]. Similarly, the presence of specific compounds like di-hydroxycinnamic acid glucuronide in apple products indicates thermal treatment history [72].

2. Authenticity Despite Processing: Identifying markers that persist through processing allows authentication of premium ingredients in final products. For example, 4-hydroxybenzaldehyde remains detectable in chia-containing cookies despite the baking process, enabling verification of this high-value ingredient [60].

3. Quality Control: Monitoring processing-induced markers helps manufacturers maintain consistent product quality by ensuring uniform processing conditions across batches. The metabolic profile consistency correlates with sensory characteristics and nutritional quality [72] [11].

4. Fraud Detection: Unexpected metabolic profiles can reveal adulteration or mislabeling, even in processed products. Deviation from expected processing markers may indicate use of inferior ingredients or unauthorized processing methods [76] [60].

The following diagram illustrates the relationship between processing conditions and marker stability in food authentication:

Diagram 2: Relationship Between Processing Conditions and Marker Stability in Food Authentication

The stability of metabolic markers during food processing is a critical consideration for developing robust authentication methods. Thermal processing induces the most significant changes in metabolic profiles, both through degradation of native compounds and formation of process-induced markers. Non-thermal technologies such as high-pressure processing better preserve the native metabolic signature of foods. Through carefully designed untargeted metabolomics workflows incorporating appropriate data processing strategies, researchers can identify both stable authentication markers that persist through processing and specific markers that indicate processing history. This dual approach enables comprehensive food authentication that remains effective despite the modifications introduced by various processing techniques, supporting efforts to ensure food integrity, quality, and authenticity in complex food supply chains.

Strategies for Harmonizing Methods Across Laboratories and Instruments

In food authentication research, non-targeted metabolomics provides a powerful tool for detecting food fraud and verifying origin by comprehensively analyzing the small-molecule composition of food products [8]. However, the major challenge facing this technique is the lack of harmonization across different laboratories and instrument platforms, which can compromise the comparability and reproducibility of results [77] [78]. Methodological variations in pre-analytical, analytical, and post-analytical processes create significant bottlenecks for implementing non-targeted metabolomics in routine food control [77]. This application note outlines standardized protocols and strategic frameworks designed to overcome these harmonization challenges, enabling reliable cross-laboratory data comparison in food authentication studies.

Core Principles of Harmonization

Successful harmonization of non-targeted metabolomics data across multiple laboratories relies on three fundamental pillars: reference standardization, consistent data processing, and quality assurance [78] [79]. Reference standardization involves using calibrated, matrix-matched reference materials to correct for systematic technical errors and enable quantitative comparisons [78]. Consistent data processing requires standardized workflows for feature extraction, compound identification, and multivariate statistical analysis to minimize technical variability [47]. Quality assurance incorporates quality control samples and standardized reporting to ensure data reliability throughout the analytical workflow [77] [79]. Together, these approaches form an integrated system for producing comparable metabolomic data regardless of the laboratory or instrument platform used.

Protocol 1: Reference Standardization for Quantitative Harmonization

Experimental Principle

Reference standardization addresses quantification challenges in high-resolution metabolomics (HRM) by normalizing metabolite spectral peak intensities from experimental samples to metabolite concentrations in calibrated reference materials analyzed concurrently [78]. This approach corrects for systematic technical errors and enables harmonization of metabolomics data collected across different studies, laboratories, and analytical methods [78]. For food authentication, this principle allows for direct comparison of metabolite profiles generated in different laboratories, which is essential for building cumulative databases of authentic food materials.

Materials and Equipment

• Reference Materials: NIST SRM 1950 (Metabolites in Frozen Human Plasma) for clinical studies; matrix-matched food reference materials (e.g., pooled authentic food samples) for food authentication [78] [80].
• LC-HRMS System: Liquid chromatography coupled to high-resolution mass spectrometry (e.g., Q-Exactive Orbitrap, Thermo Scientific) [78].
• Chromatography Columns: HILIC column (e.g., Waters XBridge BEH Amide XP, 2.1 × 50 mm, 2.6 µm) for polar metabolites; C18 column (e.g., Higgins Targa C18, 2.1 × 50 mm, 3 µm) for lipophilic metabolites [78].
• Solvents: LC-MS grade water, acetonitrile, methanol, formic acid, ammonium acetate [78].
• Internal Standards: Stable isotope-labeled internal standards for quantification [33].

Step-by-Step Procedure

• Preparation of Reference Materials:

  • Acquire or prepare matrix-matched reference pools (e.g., pooled authentic olive oil for olive oil authentication studies).
  • Characterize approximately 200 metabolites in the reference material using authentic chemical standards [78].
  • Aliquot and store reference materials at -80°C to maintain stability [78].

• Sample Preparation:

  • Prepare food samples using a standardized extraction protocol. For plant materials, use 200.00 ± 0.01 mg of sample with 4 mL of ethyl acetate [47].
  • Perform ultrasound-assisted extraction for 30 minutes at room temperature.
  • Centrifuge at 4400 × g for 10 minutes and filter through 0.45 µm nylon filters [47].

• Instrumental Analysis:

  • Analyze reference materials and food samples in alternating sequences (e.g., one reference sample for every 10 study samples) [78].
  • For HILIC-ESI+ analysis: Use mobile phase A (water), B (acetonitrile), C (2% formic acid) with gradient elution [78].
  • For C18-ESI- analysis: Use mobile phase A (water), B (acetonitrile), C (10 mM ammonium acetate) with gradient elution [78].
  • Operate HRMS at 120,000 resolution with mass range 85-1275 m/z [78].

• Data Processing:

  • Integrate peak areas for detected metabolites in both reference and study samples.
  • Calculate relative metabolite concentrations in study samples using the formula: [Metabolite]study = (Peak Astudy / Peak Aref) × [Metabolite]ref where Peak A is the integrated peak area and [Metabolite]ref is the known concentration in the reference material [78].

Data Interpretation

The reference standardization protocol enables quantitative harmonization of metabolomic data across platforms. The performance of this approach can be evaluated by calculating inter-laboratory coefficients of variation (CVs) for key metabolite markers. Successful harmonization should yield inter-laboratory CVs below 20-30% for the majority of metabolites, significantly improving upon the typical variability observed in non-harmonized studies [78].

Table 1: Performance Metrics of Reference Standardization for Cross-Laboratory Harmonization

Metric	Pre-Harmonization	Post-Harmonization	Assessment Method
Inter-lab CV for Amino Acids	25-50%	10-15%	Coefficient of variation across 17 studies [78]
Inter-lab CV for Lipids	30-60%	12-18%	Coefficient of variation across 17 studies [78]
Number of Quantifiable Metabolites	~50-100	~200	Quantitative measures in reference materials [78]
Data Harmonization Period	N/A	17 months	Long-term reproducibility assessment [78]

Protocol 2: Harmonized Data Processing for Food Authentication

Experimental Principle

Data processing harmonization addresses the challenge of inconsistent feature detection, annotation, and marker selection across different software platforms and laboratories [47]. This protocol provides a standardized workflow for processing non-targeted metabolomics data in food authentication applications, specifically comparing open-source and commercial software options to ensure consistent identification of discriminant markers for geographical origin, adulteration, or authenticity [47].

Materials and Equipment

• Data Processing Software: Commercial Compound Discoverer (Thermo Fisher Scientific) and/or open-source MS-DIAL [47].
• Spectral Databases: FoodDB (www.foodb.ca), Human Metabolome Database (www.hmdb.ca), PhytoHub (www.phytohub.eu), NIST MS libraries, Fiehn libraries [8] [15].
• Multivariate Analysis Software: SIMCA-P+ (Umetrics) or R-based packages for statistical analysis [47].

Step-by-Step Procedure

• Raw Data Conversion:

  • Convert raw HRMS data to open formats (e.g., mzML, mzXML) using conversion tools like MSConvert [47].

• Feature Detection and Alignment:

  • MS-DIAL Parameters: Set mass accuracy to 5 ppm for GC-Orbitrap-HRMS or 10 ppm for LC-Orbitrap-HRMS; set retention time tolerance to 0.1-0.5 min [47].
  • Compound Discoverer Parameters: Use analogous settings with "Unknown Detection" node for feature detection [47].
  • Perform peak alignment using statistical algorithms (e.g, LOWESS) to correct for retention time drift [33].

• Compound Annotation:

  • Match accurate mass (typically < 5 ppm error) and retention time/index against authentic standards when available [47].
  • Utilize MS/MS spectral matching with a similarity score threshold > 70% for confident annotations [47].
  • For food authentication, prioritize food-specific databases like FoodDB and PhytoHub [8].

• Multivariate Data Analysis:

  • Perform Principal Component Analysis (PCA) to identify natural clustering and outliers.
  • Use Partial Least Squares-Discriminant Analysis (PLS-DA) to identify metabolites discriminating sample groups [47].
  • Apply false discovery rate (FDR) correction to statistical tests with q-value < 0.05 considered significant.

• Marker Identification:

  • Select features with Variable Importance in Projection (VIP) scores > 1.5 from PLS-DA models.
  • Verify significance with univariate statistics (e.g., p-value < 0.05 after FDR correction) [47].

Data Interpretation

The harmonized data processing workflow enables consistent identification of authentication markers across different laboratories. Performance can be evaluated by comparing the number of consistently identified features and discriminant markers between software platforms and across laboratories. Successful implementation should yield a core set of authentication markers that are consistently identified regardless of the processing software or laboratory.

Table 2: Comparison of Data Processing Software for Food Authentication Metabolomics

Parameter	Compound Discoverer	MS-DIAL	Implication for Food Authentication
Features Detected	Moderate	High	MS-DIAL may detect more potential markers [47]
Level 2 Annotations	52 compounds	115 compounds	MS-DIAL provides more putative identifications [47]
Duplicate Features	Moderate	Moderate	Both require careful curation [47]
Background Signals	Moderate	Moderate	Both effectively remove background with proper blanks [47]
Database Dependency	High	High	Marker identification heavily depends on used databases [47]

Protocol 3: Quality Assurance and Quality Control Framework

Experimental Principle

A comprehensive Quality Assurance and Quality Control (QA/QC) framework is essential for maintaining data quality throughout the metabolomics workflow and ensuring long-term reproducibility of food authentication methods [77] [79]. This protocol incorporates system suitability testing, pooled quality control samples, and standardized reporting to monitor and maintain analytical performance across multiple laboratories and over extended time periods [79].

Materials and Equipment

• System Suitability Test Mix: Commercial metabolite standards covering key chemical classes relevant to food analysis (e.g., amino acids, organic acids, phenolics, lipids) [79].
• Pooled Quality Control (QC) Sample: Representative pool of all study samples or matrix-matched reference material [78] [80].
• Solvents and Buffers: LC-MS grade solvents and additives for mobile phase preparation [78].

Step-by-Step Procedure

• System Suitability Testing:

  • Prior to sample analysis, inject system suitability test mix to verify instrument performance.
  • Ensure retention time stability (CV < 2%), mass accuracy (< 5 ppm error), and peak intensity stability (CV < 15%) for all standards [79].

• Pooled QC Implementation:

  • Prepare a large aliquot of pooled QC sample from all study samples or use matrix-matched reference material.
  • Inject pooled QC samples at the beginning of the sequence for system equilibration (3-5 injections).
  • Analyze pooled QC repeatedly throughout the sequence (every 6-10 study samples) to monitor system stability [78].

• Data Quality Monitoring:

  • Monitor retention time drift throughout the sequence; apply correction algorithms if necessary.
  • Track peak area CV for features detected in pooled QC samples; features with QC CV > 30% should be flagged or excluded [33].
  • Use principal component analysis of pooled QC samples to identify significant drift or outliers.

• Standardized Reporting:

  • Document all sample preparation parameters, instrument methods, and data processing settings.
  • Report metabolite identifications according to confidence levels (Level 1: confirmed with standard, Level 2: putative annotation, etc.) [77].
  • Include all QA/QC results in study reports and publications.

Data Interpretation

Effective QA/QC implementation results in stable analytical performance throughout the sequence and across laboratories. Key performance indicators include >70% of detected features with QC CV < 30%, retention time stability with CV < 2%, and tight clustering of pooled QC samples in PCA space. These metrics ensure that observed differences in food metabolite profiles reflect true biological variation rather than technical artifacts.

The Scientist's Toolkit

Table 3: Essential Research Reagents and Materials for Harmonized Non-Targeted Metabolomics

Item	Function	Examples & Specifications
Reference Materials	Cross-laboratory calibration and quality control	NIST SRM 1950, matrix-matched food reference materials [78] [80]
Stable Isotope Standards	Internal standards for quantification	13C-, 15N-, or 2H-labeled metabolites [33]
Chromatography Columns	Metabolite separation	HILIC (polar metabolites), C18 (lipophilic metabolites) [78]
Data Processing Software	Feature detection, annotation, and analysis	Compound Discoverer (commercial), MS-DIAL (open-source) [47]
Metabolite Databases	Compound identification and annotation	FoodDB, HMDB, PhytoHub, NIST, Fiehn libraries [8] [15]

Workflow Visualization

Harmonized Metabolomics Workflow

The harmonization strategies presented in this application note provide a comprehensive framework for generating comparable and reproducible non-targeted metabolomics data in food authentication research. By implementing reference standardization, consistent data processing protocols, and rigorous QA/QC measures, laboratories can overcome the significant challenges associated with cross-platform and cross-laboratory metabolite analysis. These approaches enable the construction of robust, cumulative databases of authentic food metabolite profiles, which are essential for combating food fraud and protecting consumers. As the field advances, continued development of matrix-matched reference materials, open-source computational tools, and community-wide standardization efforts will further enhance the reliability and implementation of non-targeted metabolomics in food authentication.

Quality Assurance and Quality Control (QA/QC) in Non-Targeted Workflows

Quality Assurance (QA) and Quality Control (QC) are fundamental processes for generating reliable, reproducible data in non-targeted metabolomics, especially in food authentication research. According to ISO9000 standards, QA encompasses all planned and systematic activities implemented before and during data acquisition to provide confidence that quality requirements will be fulfilled. In contrast, QC describes the operational techniques and activities used to measure and report these quality requirements during and after data acquisition [81] [82]. In the context of food authentication, where verifying label claims and detecting adulteration are critical, implementing robust QA/QC protocols ensures that metabolic markers used for discrimination are analytically sound and biologically meaningful rather than technical artifacts [8] [83].

The fundamental distinction between targeted and untargeted analyses necessitates specialized QA/QC approaches. While targeted assays focus on measuring predefined analytes with known identities, untargeted metabolomics aims to comprehensively measure as many metabolites as feasible without prior knowledge of their identities [81]. This inherent uncertainty means QA/QC cannot be optimized for specific metabolites beforehand but must instead demonstrate that the analytical platform reliably measures the overall metabolic profile [84]. This guidance outlines comprehensive QA/QC practices tailored to non-targeted workflows, with specific application to food authentication research.

QA/QC Framework and Core Concepts

Quality Management Principles

Effective quality management in non-targeted metabolomics integrates both QA and QC components throughout the entire experimental workflow. QA activities include formal design of experiment, staff training, standard operating procedures, preventative instrument maintenance, and standardized computational workflows [81]. QC measurements objectively demonstrate that quality management processes have been fulfilled and include analysis of QC samples such as reference standards, pooled samples, and blanks [82].

The "fit-for-purpose" principle is imperative for QA/QC guidance, making it inclusive, flexible, and non-prescriptive to reach the largest contingent of practitioners [84]. This principle acknowledges that different applications require differing levels of QA/QC based on the research objectives. For food authentication studies, where identifying subtle differences between authentic and adulterated products is crucial, rigorous QA/QC protocols are essential to detect meaningful biological variations amidst technical noise.

Seven-Stage QA/QC Framework

The Metabolomics Quality Assurance and Quality Control Consortium (mQACC) has prioritized seven principal QC stages that form a comprehensive framework for untargeted metabolomics [84]:

• Study Design including sample collection, storage, and tracking
• Sample Preparation following standardized protocols
• Instrument Performance Monitoring using system suitability tests
• Data Acquisition with interleaved QC samples
• Data Pre-processing and quality assessment
• Metabolite Identification and annotation
• Data Reporting and documentation

This framework ensures quality is maintained at each step of the analytical process, from initial sample collection through final data interpretation, which is particularly important in food authentication where chain of custody and sample integrity are crucial.

Experimental Protocols for QA/QC in Food Authentication

Protocol 1: System Suitability Testing

Purpose: To verify that the analytical platform is "fit-for-purpose" before analyzing valuable study samples, particularly important when analyzing high-value food products prone to adulteration such as olive oil or specialty seeds [81] [83].

Materials:

• System suitability test mixture (5-10 authentic chemical standards)
• Mobile phases (MS-grade)
• Blank solvent (matching extraction solvent)

Procedure:

• Blank Analysis: Run a blank gradient with no sample to identify impurities from solvents or LC system contamination [81].
• SST Sample Preparation: Prepare a solution containing 5-10 authentic chemical standards dissolved in chromatographically suitable diluent. Select analytes distributed across the m/z and retention time range to assess the full analytical window [81].
• SST Analysis: Inject the SST sample and acquire data using the same method as for experimental samples.
• Performance Assessment: Evaluate the acquired data for:
  • Mass accuracy: m/z error ≤ 5 ppm compared to theoretical mass
  • Retention time stability: RT error < 2% compared to defined retention time
  • Peak area: Predefined acceptable peak area ± 10%
  • Peak shape: Symmetrical peaks with no evidence of splitting [81]
• Corrective Action: If acceptance criteria are not met, perform corrective maintenance and reanalyze the SST sample before proceeding with study samples.

Application Note: For food authentication studies, include at least one chemical standard representative of the food matrix being analyzed (e.g., a characteristic fatty acid for oil authentication or amino acid for dairy products) to verify performance for relevant compound classes.

Protocol 2: Preparation and Use of Pooled QC Samples

Purpose: To monitor analytical stability throughout the batch sequence, perform intra-study reproducibility measurements, and correct for systematic errors [81].

Materials:

• Aliquots from all study samples (minimum 10-20 μL each)
• Internal standard mixture

Procedure:

• QC Pool Preparation: After all study samples are prepared, combine equal aliquots (10-20 μL) from each study sample to create a pooled QC sample.
• Sample Vial Preparation: Transfer the pooled QC to multiple injection vials (enough for entire batch).
• Batch Sequence Design: Incorporate pooled QC samples throughout the analytical batch:
  • Conditioning: Inject pooled QC 5-10 times at beginning of sequence to condition system
  • Monitoring: Analyze pooled QC every 4-8 study samples throughout sequence
  • Dilution Series: Include a pooled QC dilution series to assess linearity and range [84]
• Data Evaluation: Monitor key parameters in pooled QC samples across the batch:
  • Retention time drift
  • Peak area variance
  • Signal intensity
  • Mass accuracy
• Data Correction: Apply batch correction algorithms (if needed) using pooled QC data to normalize systematic errors.

Application Note: In food authentication studies, ensure the pooled QC represents all sample types being compared (e.g., authentic and potential adulterants) to provide comprehensive quality assessment across the entire metabolic profile.

Protocol 3: Identity Marker Discovery with Quality Controls

Purpose: To discover and validate metabolic markers for food authentication while ensuring analytical quality throughout the process [83].

Materials:

• Authenticated reference materials
• Internal standard reagent (e.g., 33 compounds nonendogenous to food) [25]
• Quality control samples

Procedure:

• Reference Material Selection: Acquire authenticated reference materials from multiple independent sources [83].
• Chemical Fractionation: Implement comprehensive fractionation covering volatile, polar, and solid fractions [83].
• Non-targeted Analysis: Analyze all samples using LC-MS or GC-MS with interleaved QC samples.
• Data Pre-processing: Convert raw data to organized, tabulated formats with quality assessment [85].
• Marker Discovery: Apply random forest machine learning with feature extraction for marker identification [83].
• Statistical Validation: Validate markers using appropriate statistical methods.
• Compound Annotation: Annotate significant markers using accurate mass, MS/MS, and database searching.

Application Note: For processed foods, recognize that many unique metabolites may be diluted or lost during processing, requiring sensitive detection and rigorous QC to identify residual markers [83].

QA/QC Visualization Workflows

QA/QC in Non-Targeted Metabolomics Workflow

Food Authentication Marker Discovery Workflow

Research Reagent Solutions for QA/QC

Table 1: Essential Research Reagents for QA/QC in Non-Targeted Metabolomics

Reagent Type	Specific Examples	Function in QA/QC	Application in Food Authentication
System Suitability Test Mixtures	5-10 authentic chemical standards across m/z and RT range [81]	Verifies instrument performance before sample analysis	Ensure detection capability for food-specific metabolites
Internal Standards	Isotopically-labelled compounds; PTFI mixture of 33 nonendogenous compounds [25]	Monitors system stability for each sample; enables data harmonization	Corrects for matrix effects in diverse food samples
Pooled QC Samples	Aliquots pooled from all study samples [81] [84]	Assesses intra-study reproducibility; enables batch correction	Monitors analytical stability across authentic and test samples
Process Blanks	Solvent blanks without biological matrix [81]	Identifies contamination from solvents, containers, or processing	Detects background interference in complex food matrices
Reference Materials	Authenticated food samples; certified reference materials [83]	Provides benchmark for method validation and quality assessment	Serves as ground truth for authentication models

Data Quality Assessment and Reporting

Quality Assessment Metrics

Table 2: Key QC Metrics and Acceptance Criteria for Non-Targeted Metabolomics

QC Metric	Assessment Method	Acceptance Criteria	Frequency of Assessment
Retention Time Stability	Relative standard deviation (RSD) in pooled QCs [81]	< 2% drift across batch	Throughout analytical batch
Mass Accuracy	Deviation from theoretical mass in ppm [81]	≤ 5 ppm error	Each sample and QC injection
Signal Intensity	RSD of internal standards in study samples [82]	< 20-30% RSD	Throughout analytical batch
Peak Shape	Symmetry factor, peak width [81]	No splitting; symmetrical shape	System suitability testing
Feature Detection	Number of features in pooled QCs [84]	Consistent count (±15%)	Each pooled QC injection
Missing Values	Percentage in QC and study samples [84]	< 20% in study samples	After data pre-processing

Reporting Standards

Comprehensive reporting of QA/QC procedures is essential for building confidence in food authentication findings. The Metabolomics Standards Initiative (MSI) has developed minimum reporting standards covering sample preparation, experimental analysis, quality control, metabolite identification, and data pre-processing [85]. Key reporting elements include:

• Sample preparation metadata: Tissue harvesting method, extraction solvents, storage conditions [85]
• Chromatography instrumentation: Manufacturer, column details, separation parameters [85]
• MS instrumentation: Manufacturer, model, ionization source, mass analyzer [85]
• QC sample results: System suitability outcomes, pooled QC stability, blank contamination assessment [82]
• Data quality metrics: Retention time stability, mass accuracy, signal intensity variance [82]

For food authentication studies, additional reporting should include details of reference material authentication, processing methods for food samples, and validation of marker compounds against certified standards when available.

Application to Food Authentication Research

In food authentication, QA/QC practices enable reliable detection of metabolic markers that differentiate authentic products from adulterated counterparts. For example, in distinguishing chia, linseed, and sesame seeds—high-value ingredients often subject to adulteration—robust QA/QC ensures that detected markers like sesamol in chia or specific fatty acid profiles are biologically meaningful rather than technical artifacts [83]. The PTFI Non-targeted Metabolomics Platform demonstrates how standardized protocols incorporating unique internal standard reagents enable data harmonization across laboratories, creating scalable resources for food authentication [25].

Implementing these QA/QC protocols allows researchers to:

• Confidently identify metabolic markers for food origin, processing, and adulteration
• Generate reproducible data across batches and laboratories
• Support regulatory decisions with analytically sound evidence
• Build comprehensive databases of authentic food metabolomes

As non-targeted metabolomics continues to evolve as a powerful tool for food authentication, adherence to these QA/QC practices will strengthen the field by ensuring data quality, enhancing reproducibility, and building stakeholder confidence in the resulting authentication models.

In non-targeted metabolomics for food authentication, researchers face the dual challenge of data overload and model overfitting. Modern high-resolution liquid chromatography-mass spectrometry (LC-MS) platforms generate extremely complex datasets from biological samples, capturing thousands of metabolic features in a single analysis [86]. This high-dimensional data space, where the number of variables (p) vastly exceeds the number of observations (n), creates ideal conditions for overfitting—where models perform well on training data but fail to generalize to new samples [87]. The resulting multivariate models may appear statistically significant while being biologically meaningless, compromising their utility for authenticating food origin, processing, and adulteration.

The data overload problem stems from the analytical sensitivity of modern platforms. Ultra-high-performance liquid chromatography (UHPLC) systems coupled to time-of-flight (TOF) or quadrupole-time-of-flight (QTOF) mass spectrometers can detect hundreds to thousands of metabolites in a single food sample run [86]. Without proper safeguards, overfitting occurs when models capture not only the genuine biological signal but also analytical noise and irreproducible variations, ultimately failing when applied to new sample batches or slightly different analytical conditions. This review establishes structured protocols to navigate these challenges while building robust, validated models for food authentication research.

Foundational Principles for Model Robustness

Strategic Experimental Design

A carefully constructed experimental design forms the first defense against overfitting by ensuring biological relevance and statistical power:

• Incorporate Biological Variability: Include sufficient biological replicates that represent the expected natural variation in food samples (e.g., different growing regions, seasons, varieties). For food authentication studies, sample size should be determined by power analysis whenever possible [87].

• Implement Cross-Over Designs: When investigating dietary interventions or processing effects, cross-over designs where the same unit receives multiple treatments are preferred over parallel designs as they minimize inter-individual variation that can obscure true effects [87].

• Control Pre-Analytical Variables: Standardize sample collection, storage, and preparation protocols to minimize unwanted technical variance. For plant-based foods, control harvesting conditions, post-harvest treatments, and storage parameters [86].

• Include Quality Control Samples: Prepare pooled quality control (QC) samples by combining equal aliquots from all study samples and analyze these repeatedly throughout the acquisition sequence to monitor instrumental stability [87].

Data Acquisition Considerations

Standardized LC-MS data acquisition parameters significantly enhance data quality and comparability:

Table 1: Recommended LC-MS Parameters for Non-Targeted Metabolomics

Parameter	Recommended Setting	Rationale
Mass Range	50-1500 m/z	Covers most metabolites while limiting file size [86]
Chromatography	Reversed-phase UHPLC	Superior separation efficiency for complex food extracts [86]
Ionization Mode	Both positive and negative ESI	Increases metabolite coverage [86]
Mass Accuracy	< 5 ppm	Enishes reliable molecular formula assignment [86]
Quality Control	Pooled QC every 6-10 samples	Monitors instrument stability [87]

Computational Workflow for Overfitting Prevention

The data analysis workflow must incorporate multiple checkpoints to prevent overfitting while extracting biologically meaningful information from food metabolomics data.

Data Preprocessing Protocol

Proper data preprocessing is essential before multivariate analysis to minimize technical variance while preserving biological signal:

Protocol 3.1: LC-MS Data Preprocessing for Food Authentication

• Peak Detection and Alignment

  • Use automated peak detection algorithms (e.g., XCMS, MS-DIAL) with consistent parameter settings across all samples
  • Apply retention time alignment to correct for chromatographic shifts
  • Set intensity threshold to 3-5 times the baseline noise level
  • Time requirement: 4-8 hours for a typical dataset of 100 samples

• Missing Value Imputation

  • For values missing completely at random, apply k-nearest neighbor imputation (k=10)
  • For values missing due to low abundance, use one-fifth of the minimum positive value for each variable
  • Critical Note: Document the percentage of missing values per sample; exclude samples with >30% missing values

• Normalization and Scaling

  • Apply probabilistic quotient normalization to correct for overall concentration differences
  • Follow with unit variance scaling (autoscaling) to give equal weight to all metabolites
  • Validation Step: Check QC samples after normalization - relative standard deviation should be <30% for most metabolites

Multivariate Analysis with Validation Safeguards

Protocol 3.2: Supervised Analysis with Cross-Validation

• Initial Exploratory Analysis

  • Perform Principal Component Analysis (PCA) on autoscaled data to identify outliers and natural clustering
  • Exclude outliers exceeding 95% confidence limits in the Hotelling's T² distribution
  • Documentation: Record percentage of variance explained by first 5 principal components

• Supervised Modeling with PLS-DA/OPLS-DA

  • Apply Partial Least Squares-Discriminant Analysis (PLS-DA) or Orthogonal PLS-DA (OPLS-DA) for class separation
  • Determine optimal number of components by 7-fold cross-validation
  • Use permutation testing (n=100-200) to assess model significance
  • Acceptance Criterion: Permuted models should have significantly lower R² and Q² values than the original model

• Model Validation Metrics

  • Calculate R²X (variance in X explained), R²Y (variance in Y explained), and Q² (predictive ability)
  • Quality Thresholds: Q² > 0.5 indicates good predictive ability; Q² > 0.9 suggests overfitting
  • Compute discrimination accuracy on independent test set (minimum 20% of total samples)

Table 2: Multivariate Model Validation Parameters and Acceptance Criteria

Validation Parameter	Calculation Method	Acceptance Criterion	Implementation in Food Authentication
Q² (Predictive Ability)	7-fold cross-validation	Q² > 0.5	Indicates model can reliably classify unknown food samples
R²Y (Goodness of Fit)	Model performance on training data	R²Y - Q² < 0.3	Large differences indicate overfitting
Permutation Testing	Random Y-shuffling (n=100)	p-value < 0.05	Confirms model significance beyond chance
Component Number	Cross-validation error	Minimum CV-error	Prevents fitting to noise
External Validation	Blind prediction on test set	Accuracy > 80%	Ensures real-world applicability

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 3: Research Reagent Solutions for LC-MS Metabolomics

Reagent/Material	Specification	Function in Workflow	Quality Control
LC-MS Grade Solvents	Water, methanol, acetonitrile with < 5 ppb impurities	Mobile phase preparation; minimizes background noise	Lot-to-lot consistency testing
Internal Standards	Stable isotope-labeled compounds (e.g., ¹³C, ²H)	Retention time alignment; quantification reference	Purity > 95% by certificate of analysis
Quality Control Pool	Pooled aliquot from all study samples	Monitoring instrumental performance; data normalization	Homogeneity assessment before analysis
Retention Index Markers	C8-C30 fatty acid methyl esters	Retention time calibration across sequences	Fresh preparation for each batch
Blank Solvent	LC-MS grade solvents without biological matrix	System contamination monitoring; background subtraction	Analyze repeatedly throughout sequence
Standard Reference Material	Certified reference materials (NIST, LGC)	Method validation; cross-laboratory comparability	Use established reference values

Advanced Statistical Approaches for Complex Designs

Nutritional metabolomics often involves complex experimental designs with repeated measurements, requiring specialized statistical approaches:

Protocol 5.1: ANOVA-Simultaneous Component Analysis (ASCA) for Complex Food Studies

ASCA is particularly valuable for food authentication studies investigating multiple factors (e.g., geographical origin, processing method, storage conditions):

• Experimental Design Matrix

  • Create a design matrix encoding all experimental factors and their interactions
  • Include batch effects and analytical sequence as additional factors
  • Ensure balanced design whenever possible

• Variance Decomposition

  • Decompose the data matrix into effect matrices for each experimental factor
  • Apply PCA to each effect matrix to visualize factor-specific metabolic responses
  • Output Interpretation: The variance percentage explains how much of the metabolic variation is attributable to each factor

• Statistical Validation

  • Use permutation tests (n=1000) to assess significance of each factor
  • Apply cross-validation to determine optimal number of components
  • Authentication Application: The geographical origin factor should explain the largest variance component for successful authentication

Validation Framework and Reporting Standards

Protocol 6.1: Comprehensive Model Validation

• Internal Validation

  • Perform 7-10 fold cross-validation with multiple iterations
  • Calculate Q² values for each cross-validation round
  • Apply permutation testing with 100-200 permutations
  • Exclusion Criteria: Models with Q² < 0.4 or permutation p-value > 0.05 should be rejected

• External Validation

  • Reserve 20-30% of samples as completely independent validation set
  • Collect new samples from different batches or seasons
  • Apply the model to predict class membership in validation samples
  • Performance Metrics: Report accuracy, sensitivity, specificity, and AUC-ROC

• Biological Validation

  • Identify discriminant metabolites using Variable Importance in Projection (VIP) scores
  • Confirm chemical identity of key markers using MS/MS and authentic standards
  • Interpret biological plausibility of metabolic pathways affected
  • Authentication Specific: Marker metabolites should relate to known food composition differences

Table 4: Minimum Reporting Standards for Multivariate Models in Food Metabolomics

Reporting Category	Essential Elements	Purpose
Sample Preparation	Extraction method, quenching procedure, normalization approach	Enables protocol replication
Instrumental Analysis	LC column, gradient, MS ionization, resolution, mass accuracy	Facilitates cross-platform comparisons
Data Preprocessing	Software, parameters, missing value handling, normalization	Allows reprocessing and validation
Multivariate Analysis	Software, algorithm, scaling method, validation method	Ensures statistical rigor
Model Performance	R²X, R²Y, Q² values, permutation results, ROC curves	Demonstrates predictive ability
Marker Metabolites	VIP scores, identification confidence level, effect sizes	Supports biological interpretation

Effective management of overfitting and data overload in multivariate model building requires a systematic approach spanning experimental design, data acquisition, preprocessing, analysis, and validation. By implementing the protocols and validation frameworks outlined herein, researchers in food authentication can develop robust models that reliably classify unknown samples and identify meaningful metabolic markers. The integration of multiple validation strategies—from cross-validation and permutation testing to independent external validation—ensures that models generalize beyond the specific dataset used for their development. As non-targeted metabolomics continues to evolve toward larger datasets and more complex analytical platforms, these foundational practices will remain essential for extracting biologically meaningful insights from the data-rich landscape of food metabolomics.

Validation Frameworks and Integration with Other Omics

Establishing Validation Criteria for Non-Targeted Fingerprinting Methods

Non-targeted fingerprinting approaches, primarily using liquid chromatography coupled with high-resolution mass spectrometry (LC-HRMS), have emerged as a powerful strategy for food authentication. These methods aim to comprehensively characterize complex food matrices without prior selection of target analytes, enabling the detection of origin, species variety, and adulteration [77]. Despite their demonstrated potential in research settings, the implementation of these approaches in routine analysis and official food control remains limited. This gap largely stems from the absence of standardized validation protocols that ensure method reliability, reproducibility, and fitness for purpose [77]. This document establishes comprehensive validation criteria and detailed experimental protocols to guide the adoption of non-targeted fingerprinting within food authentication research. By providing a framework for assessing method performance, these criteria aim to enhance data quality, facilitate inter-laboratory comparisons, and build confidence in non-targeted results for food provenance and authenticity decisions.

Core Validation Parameters for Non-Targeted Fingerprinting

The validation of non-targeted methods requires a paradigm shift from traditional targeted analysis. It must address not only classical analytical performance metrics but also the stability of the instrumental system, the quality of the acquired data, and the predictive ability of the multivariate statistical models [77]. The following parameters are essential for establishing a validated non-targeted fingerprinting workflow.

Table 1: Essential Validation Parameters for Non-Targeted Fingerprinting

Validation Parameter	Description	Acceptance Criteria Example
Specificity	The ability to distinguish between different sample classes (e.g., geographical origin, species) based on their metabolic fingerprints.	Clear separation of classes in multivariate models (e.g., PCA, PLS-DA).
Robustness	The capacity of the method to remain unaffected by small, deliberate variations in analytical conditions (e.g., column temperature, mobile phase gradient).	Consistent classification accuracy despite parameter variations.
System Suitability	Daily verification of instrument performance (sensitivity, mass accuracy, chromatographic retention).	Analysis of a quality control (QC) sample; metrics include retention time shift < 2%, mass accuracy < 3 ppm.
Signal Stability	Assessment of instrumental drift over the sequence via repeated analysis of a pooled QC sample.	Relative Standard Deviation (RSD) of key features in QC samples < 15-30%.
Repeatability	Precision under the same operating conditions over a short interval.	RSD of feature intensities from technical replicates < 20%.
Intermediate Precision	Precision under different conditions (different days, different analysts).	RSD of feature intensities from samples analyzed across different days < 30%.
Predictive Ability	The performance of the statistical model in classifying new, unknown samples.	Cross-validation accuracy > 80%; performance on an external validation set.

Beyond these parameters, a rigorous quality assurance (QA) framework is critical. The Metabolomics Quality Assurance and Quality Control Consortium (mQACC) provides guidelines for implementing QA and Quality Control (QC) protocols to ensure data reliability and reproducibility [88]. This includes the use of internal standards to correct for variability during sample preparation and analysis, and the consistent use of pooled QC samples throughout the analytical run to monitor system stability [88].

Experimental Protocols for Method Validation

Protocol for Sample Preparation and Quality Control

This protocol ensures the integrity of the metabolome from collection to analysis, which is foundational for any reliable non-targeted study [88].

1. Sample Collection and Quenching:

• Collect samples in a sterile and consistent manner to minimize biological variability.
• Immediately quench metabolic activity to preserve the in-vivo metabolic state. This is critical for tissues and cells. Methods include:
  • Flash freezing in liquid nitrogen.
  • Using chilled organic solvents like methanol (-20°C to -80°C) [88].
• For biofluids, rapid freezing at -80°C is often sufficient.

2. Metabolite Extraction:

• Employ a solvent system that provides broad coverage of the metabolome. For comprehensive food analysis, a biphasic extraction is often suitable to separate polar and non-polar metabolites [89] [88].
• Biphasic Extraction (e.g., Methanol/Chloroform/Water):
  • Homogenize the quenched sample with a pre-cooled solvent mixture (e.g., 2:1:1 ratio of Methanol:Chloroform:Water).
  • Vortex vigorously and incubate on ice.
  • Centrifuge to separate phases. The upper aqueous phase contains polar metabolites, and the lower organic phase contains lipids and non-polar metabolites.
  • Carefully collect both phases into separate vials.
  • Evaporate solvents under a gentle stream of nitrogen or using a vacuum concentrator.
  • Reconstitute the dried extracts in solvents compatible with the subsequent chromatographic method (e.g., water for HILIC, methanol for RPLC) [90].
• Add a mixture of internal standards at the beginning of the extraction process to correct for losses and variability [88].

3. Quality Control (QC) Sample Preparation:

• Create a pooled QC sample by combining equal aliquots from all samples in the study.
• This pooled QC is analyzed repeatedly throughout the analytical batch—at the beginning for system conditioning, and then at regular intervals (e.g., every 4-10 samples) to monitor instrument stability and data quality [89].

Protocol for Instrumental Analysis and Data Acquisition

Liquid Chromatography-Mass Spectrometry (LC-MS) is the cornerstone of non-targeted fingerprinting.

1. Liquid Chromatography:

• Utilize complementary chromatographic modes to increase metabolome coverage.
  • Reversed-Phase (RPLC): Ideal for non-polar to semi-polar metabolites (e.g., lipids, many secondary plant metabolites).
  • Hydrophilic Interaction (HILIC): Ideal for polar metabolites (e.g., sugars, amino acids, organic acids) not retained by RPLC [89].
• Use a UPLC system with a stable, reproducible gradient to minimize retention time drift.

2. Mass Spectrometry:

• Employ a high-resolution mass spectrometer (e.g., Q-TOF, Orbitrap) for accurate mass measurement.
• Acquire data in both positive and negative ionization modes to maximize feature detection [89].
• Use data-dependent acquisition (DDA) or data-independent acquisition (DIA) modes to collect fragmentation (MS/MS) data for metabolite annotation.

3. Analytical Sequence:

• Sequence samples in a randomized order to avoid bias from instrument drift.
• Intersperse with the pooled QC samples to track performance.

Protocol for Data Processing and Model Validation

1. Data Pre-processing:

• Convert raw data files into a data matrix of features (defined by m/z and retention time) and their intensities.
• Use open-source tools like XCMS (often integrated into platforms like patRoon) or MS-DIAL for peak picking, alignment, and retention time correction [90] [89].
• Perform blank subtraction to remove background signals.

2. Data Quality Assessment:

• Filter the data matrix based on QC reproducibility. Features with a high RSD (e.g., >30%) in the pooled QC samples should be removed as they represent unstable or noisy signals [77].

3. Statistical Modeling and Validation:

• Use multivariate statistical methods like Principal Component Analysis (PCA) for unsupervised exploration and Partial Least Squares-Discriminant Analysis (PLS-DA) for supervised classification.
• Critical Step - Model Validation: Avoid overfitting by not relying solely on the model's fit to the training data.
  • Internal Validation: Use cross-validation (e.g., 7-fold cross-validation) to assess the model's predictive performance.
  • External Validation: The gold standard is to test the model on a completely independent set of samples that were not used in model building [77].
• The model's predictive ability is the primary validation metric, reported as the classification accuracy for the external validation set.

Workflow Visualization

The following diagram summarizes the comprehensive validation workflow for non-targeted fingerprinting, from initial sample handling to final model reporting.

The Scientist's Toolkit: Essential Research Reagents and Materials

A successful non-targeted fingerprinting study relies on a suite of carefully selected reagents, solvents, and software tools.

Table 2: Essential Research Reagents and Solutions for Non-Targeted Fingerprinting

Item	Function / Purpose	Examples / Notes
Solvents (HPLC/MS Grade)	Sample extraction, reconstitution, and mobile phase preparation.	Methanol, Acetonitrile, Chloroform, Water, Methyl tert-butyl ether (MTBE). Low contaminants are critical for background signal reduction [88].
Internal Standards	Correction for variability during sample preparation and analysis.	Stable isotope-labeled compounds (e.g., ¹³C, ²H labeled amino acids, fatty acids). Added prior to extraction [88].
Chromatography Columns	Separation of complex metabolite mixtures.	C18 (for RPLC), HILIC (e.g., Amide, ZIC-pHILIC for polar metabolites). Using multiple column chemistries increases coverage [89].
Quality Control (QC) Material	Monitoring instrument stability and data quality throughout the run.	Pooled sample from all study specimens; commercial standard reference materials if available [89] [88].
Mass Calibration Solution	Ensuring mass accuracy of the mass spectrometer.	Vendor-specific solution providing known ions across a wide m/z range.
Data Processing Software	Converting raw data into a structured feature table and performing statistical analysis.	Open Source: XCMS, patRoon, MS-DIAL, MetFrag [89]. Commercial: Vendor-specific software.
Chemical Databases	Annotating detected features by matching against known compounds.	PubChem, PubChemLite (exposomics-focused), organism-specific databases (e.g., WormJam for C. elegans), in-silico fragmentation tools (CFM-ID) [89].

Inter-laboratory Studies and Statistical Equivalence of Spectral Data

In food authentication research, the verification of metabolic biomarkers and spectral profiles across different laboratories presents a significant scientific challenge. Non-targeted metabolomics has emerged as a powerful tool for detecting food fraud, verifying geographical origin, and identifying ingredient substitution [91] [60]. However, the reproducibility and equivalence of spectral data generated across different research facilities remain critical concerns that must be addressed through standardized protocols and rigorous statistical validation [92] [93].

The fundamental challenge in inter-laboratory metabolomic studies lies in the multitude of analytical and computational variables that can introduce systematic biases. These include differences in instrumentation, sample preparation methodologies, data processing algorithms, and statistical approaches [93] [47]. Without proper standardization, these variables compromise the comparability of data, potentially leading to inconsistent biomarker identification and unreliable authentication models [92]. This application note establishes a framework for achieving statistical equivalence of spectral data in non-targeted metabolomics for food authentication research.

Key Challenges in Inter-Laboratory Metabolomic Studies

Analytical Variations Across Platforms

Inter-laboratory studies consistently reveal substantial variation in metabolomic data due to differences in analytical platforms and procedures. A comparative study of lettuce cultivars using two different UPLC-QTOF-MS platforms (Waters and Agilent) found that the number of shared candidate biomarkers varied significantly depending on the data pre-processing software used [92]. When the same software (Progenesis QI) was applied to both datasets, only 26 candidate biomarkers were shared between sample groups. In contrast, when each dataset was processed using the manufacturers' embedded software, 101 shared candidates were identified, with only 13 metabolites common to both processing approaches [92].

Data Processing and Annotation Discrepancies

The choice of data processing tools significantly influences metabolite annotation and subsequent biomarker selection. A comparative study of GC-Orbitrap-HRMS data processing strategies for thyme geographical differentiation demonstrated that open-source MS-DIAL and commercial Compound Discoverer software yielded markedly different results [47]. Compound Discoverer putatively annotated 52 compounds at Level 2 confidence, while MS-DIAL annotated 115 compounds at the same confidence level. This discrepancy was attributed to differences in feature detection algorithms, peak alignment parameters, and database matching strategies [47].

Table 1: Methodological Variations Affecting Inter-Laboratory Reproducibility

Variation Source	Impact on Data	Reported Magnitude of Effect
Instrumentation Platform	Differential sensitivity & metabolite coverage	26 vs. 101 shared biomarkers in lettuce cultivars [92]
Data Processing Software	Feature detection & annotation rates	52 vs. 115 annotated compounds in thyme [47]
Sample Preparation	Ionization efficiency & matrix effects	Median CV 15-30% in GC-MS plasma studies [93]
Chromatographic Separation	Retention time alignment & peak resolution	Addressed by internal retention time standards [38]

Standardized Experimental Protocols

Cross-Laboratory Metabolomic Profiling Protocol

The following protocol provides a standardized framework for inter-laboratory non-targeted metabolomics in food authentication research, with specific applications for geographical origin discrimination and ingredient verification.

Sample Preparation and Extraction

• Sample Commutation: Reduce particle size to 0.2 mm using an ultra-centrifugal mill (e.g., ZM200, Retsch GmbH) for 10 minutes at 8000 rpm to ensure homogeneous representation [47].
• Extraction Protocol: Weigh 200.00 ± 0.01 mg of sample into a 15 mL polypropylene tube. Add 4 mL of GC-MS grade ethyl acetate and subject to ultrasound-assisted extraction for 30 minutes at 37 kHz and room temperature [47].
• Post-Extraction Processing: Centrifuge at 4400 × g for 10 minutes, filter through 0.45 µm nylon filters, and store at -21°C until analysis. Include procedure blanks to identify background signals [47].

Instrumental Analysis and Quality Control

• Chromatographic Separation: Utilize a BP5MS capillary column (30 m × 0.25 mm i.d., 0.25 µm particle size) with a consistent temperature gradient optimized for the food matrix [47].
• Internal Standardization: Implement a novel internal retention time standard (IRTS) mixture containing compounds non-endogenous to food samples to enable robust chromatographic alignment across laboratories [38].
• Quality Control Samples: Analyze quality control pools created by combining aliquots of all samples to monitor instrument performance and correct for signal drift [94].

Data Acquisition Parameters

• Mass Spectrometry: Operate in both positive and negative electrospray ionization modes to maximize metabolome coverage. For GC-MS applications, use electron ionization mode at 70 eV [94] [47].
• Resolution Settings: For high-resolution mass spectrometry, maintain resolution of at least 70,000 full width at half maximum to ensure accurate mass measurements for elemental composition determination [47].
• Data Acquisition Range: Collect data in the m/z range of 50-1000 to cover most known food metabolites, including primary and secondary metabolites relevant to authentication [60].

Data Processing and Statistical Analysis Workflow

The following workflow diagram illustrates the standardized data processing protocol for achieving comparable results across laboratories:

Data Pre-processing and Normalization

• Peak Detection and Alignment: Use consistent parameters for peak picking, noise filtering, and retention time alignment across all participating laboratories. The IRTS mixture enables robust chromatographic alignment of data across laboratories [38].
• Signal Drift Correction: Apply batch correction algorithms (e.g., batchCorr) to correct for analytical variances occurring during sequence runs [94].
• Data Normalization: Implement multiple normalization strategies including probabilistic quotient normalization, total area normalization, and internal standard-based normalization to enhance cross-laboratory comparability [93].

Multivariate Statistical Analysis

• Feature Selection: Apply random forest machine learning with inherent feature extraction for non-targeted metabolic marker discovery [60].
• Pattern Recognition: Utilize unsupervised (PCA, HCA) and supervised (PLS-DA, OPLS-DA) methods to identify discriminative features between sample classes.
• Validation: Perform permutation testing (n > 100) and cross-validation (e.g., 7-fold) to assess model robustness and prevent overfitting [60].

Quantitative Assessment of Inter-Laboratory Reproducibility

Rigorous assessment of reproducibility metrics is essential for establishing statistical equivalence of spectral data across laboratories. The following table summarizes key performance indicators from published inter-laboratory studies:

Table 2: Quantitative Reproducibility Metrics in Inter-Laboratory Metabolomic Studies

Study Focus	Analytical Platform	Reproducibility Metric	Performance Outcome	Reference
Human Plasma Profiling	GC-MS (Two Laboratories)	Annotation Repeatability	55 metabolites commonly annotated	[93]
NIST SRM1950 Plasma	GC-MS (Two Laboratories)	Compound Identification	26/30 overlapped metabolites	[93]
Food Metabolite Profiling	LC-MS (Standardized Method)	Feature Consistency	Qualitative consensus achieved	[38]
Thyme Authentication	GC-Orbitrap-HRMS	Platform Precision	CV < 15% with high-resolution MS	[47]
Fish Tissue Metabolomics	UPLC-MS/MS & FI-MS/MS	Precision Comparison	Targeted > Non-targeted precision	[94]

The Scientist's Toolkit: Essential Research Reagents and Materials

Successful inter-laboratory studies require careful selection and standardization of research reagents and materials. The following table outlines essential solutions for non-targeted metabolomics in food authentication:

Table 3: Essential Research Reagent Solutions for Inter-Laboratory Metabolomics

Reagent/Material	Technical Function	Application Example	Technical Specification
Internal Retention Time Standard (IRTS)	Chromatographic alignment across laboratories	Cross-laboratory LC-MS studies	Mixture of compounds non-endogenous to food [38]
Fatty Acid Methyl Esters (FAMEs)	Retention index calibration for GC-MS	GC-MS metabolomic profiling	C8-C30 linear chain length mixture [93]
Stable Isotope Standards	Quality control & quantification	Correction of matrix effects	13C-, 15N-, or 2H-labeled analogs [94]
Reference Materials (NIST SRM1950)	Method validation & performance tracking	Inter-laboratory precision assessment	Fortified with known metabolites [93]
Silylation Derivatization Reagents	Volatilization for GC-MS analysis	Chemical derivatization of polar metabolites	N-Methyl-N-(trimethylsilyl)trifluoroacetamide [93]

Achieving statistical equivalence of spectral data in inter-laboratory studies requires meticulous attention to analytical protocols, data processing strategies, and validation procedures. The implementation of standardized methods, such as the internal retention time standard (IRTS) approach, enables qualitative consensus of features across laboratories and instrumentation [38]. Additionally, the selection of appropriate data processing software and parameters significantly influences the detection and annotation of discriminative markers in food authentication studies [47].

Food authentication researchers should prioritize protocol harmonization across participating laboratories, including standardized sample preparation, instrumental analysis, and data processing workflows. Furthermore, the implementation of robust quality control measures, including reference materials and internal standards, is essential for monitoring and correcting technical variations [93]. Future methodological developments should focus on advanced computational approaches for data integration and validation, including machine learning algorithms for pattern recognition and biomarker selection [60] [95]. Through the adoption of these standardized protocols and statistical frameworks, the food authentication research community can enhance the reliability and interoperability of non-targeted metabolomic data, ultimately strengthening the scientific foundation for food integrity systems.

In the face of increasing global food authenticity challenges, the demand for robust analytical techniques to verify food origin, processing, and composition has never been greater [3]. Food fraud—including mislabeling, adulteration, and false claims of geographical origin—has significant economic and health implications, driving the need for sophisticated authentication methods [3] [96]. While traditional analytical techniques have limitations in detecting sophisticated fraud, omics technologies have emerged as powerful tools for comprehensive food analysis [3] [97]. Among these, metabolomics, genomics, and proteomics each offer unique capabilities and markers for authentication purposes. This application note provides a comparative analysis of these three omics approaches, focusing on their applications, workflows, and complementary value in food authentication research, particularly within the framework of non-targeted metabolomics studies.

Fundamental Principles and Markers

Metabolomics

Metabolomics involves the systematic study of small molecules (typically <1-2 kDa) within a biological system, providing a snapshot of the metabolic state closest to the phenotype [98] [23]. The metabolome represents the final downstream product of gene expression and is highly sensitive to exogenous factors such as climate, soil composition, and food processing methods [86]. In food authentication, metabolites serve as excellent markers for geographical origin, production methods, freshness, and processing history [86] [98]. Non-targeted metabolomics aims to capture as many metabolites as possible without prior hypothesis, making it ideal for discovering novel authentication markers [86] [6]. Key analytical platforms include liquid chromatography-mass spectrometry (LC-MS), gas chromatography-mass spectrometry (GC-MS), and nuclear magnetic resonance (NMR) spectroscopy, each with distinct advantages for different metabolite classes [86] [23].

Genomics

Genomics focuses on the analysis of DNA sequences, gene locations, and genome mapping [3]. DNA itself serves as the primary marker in genomic authentication, with specific sequences providing unique fingerprints for species identification and origin verification [3]. The stability of DNA makes it particularly suitable for analyzing deeply processed foods where other biomarkers may degrade [3]. Polymerase chain reaction (PCR) technologies, including real-time PCR and droplet digital PCR (ddPCR), form the cornerstone of DNA-based authentication, enabling precise amplification and detection of species-specific sequences even in complex food matrices [3].

Proteomics

Proteomics encompasses the large-scale analysis of proteins expressed by a genome, cell, or tissue [97] [96]. Proteins act as functional markers that can indicate species, tissue origin, and processing history of food products [97]. Protein profiles are influenced by both genetic information and environmental factors, providing intermediate information between genomics and metabolomics [97]. Mass spectrometry, particularly matrix-assisted laser desorption ionization (MALDI)-MS and electrospray ionization (ESI)-MS, are widely used in proteomic analysis of foods, enabling protein identification, quantification, and post-translational modification analysis [97] [96].

Table 1: Comparative Analysis of Omics Markers for Food Authentication

Feature	Metabolomics	Genomics	Proteomics
Primary Marker	Metabolites (small molecules <1-2 kDa)	DNA sequences	Proteins and peptides
Stability	Sensitive to processing and storage	Highly stable, resistant to processing	Moderate stability, can denature
Information Provided	Phenotypic snapshot, processing history, freshness	Species identity, genetic lineage	Functional activity, processing effects
Key Analytical Platforms	LC-MS, GC-MS, NMR	PCR, qPCR, ddPCR, DNA sequencing	MALDI-TOF, ESI-MS/MS, 2D electrophoresis
Throughput	High	High	Moderate to High
Cost	Moderate to High	Low to Moderate	Moderate to High
Sample Preparation	Moderate complexity	Relatively simple	Can be complex
Ideal For	Geographical origin, processing verification, freshness	Species identification, GMO detection, adulteration	Varietal identification, thermal processing verification

Experimental Protocols and Workflows

Non-Targeted Metabolomics Protocol for Food Authentication

Sample Preparation:

• Homogenization: Grind food samples (e.g., seeds, meat, plant materials) to a fine powder using liquid nitrogen to prevent metabolite degradation [60].
• Extraction: Weigh 50-100 mg of homogenized material and add 1 mL of extraction solvent (acetonitrile:methanol:formic acid, 74.9:24.9:0.2, v/v/v) [6]. For comprehensive coverage, multiple extraction methods may be necessary for different metabolite classes [60].
• Internal Standards: Add stable isotope-labeled internal standards (e.g., l-Phenylalanine-d8 and l-Valine-d8) for quality control and normalization [6].
• Extraction Procedure: Vortex vigorously for 30 seconds, sonicate for 15 minutes at 4°C, then centrifuge at 14,000 × g for 10 minutes [6].
• Storage: Transfer supernatant to MS vials and store at -80°C until analysis [6].

LC-MS Analysis:

• Chromatography: Utilize HILIC or reverse-phase chromatography depending on metabolite polarity. For HILIC: Mobile phase A: 0.1% formic acid, 10 mM ammonium formate in water; Mobile phase B: 0.1% formic acid in acetonitrile [6].
• Gradient: Apply linear gradient from 100% B to 60% B over 15 minutes for HILIC separation [6].
• Mass Spectrometry: Use high-resolution mass spectrometer (e.g., Q-TOF, Orbitrap) with both positive and negative ionization modes [86] [6]. Typical settings: resolution >35,000, mass range 50-1500 m/z, scan rate 1-10 Hz [86].

Data Processing:

• Peak Detection: Use software like XCMS, MZmine, or Compound Discoverer for peak picking, alignment, and integration [23] [6].
• Compound Identification: Compare mass features to databases (HMDB, MetLin) using accurate mass (±5 ppm) and MS/MS fragmentation when available [23].
• Multivariate Analysis: Apply PCA and OPLS-DA to identify discriminatory metabolites [86] [60].

Non-Targeted Metabolomics Workflow for Food Authentication

Genomic Authentication Protocol

DNA Extraction:

• Cell Lysis: Digest 20 mg of food sample in lysis buffer with proteinase K at 56°C for 2 hours [3].
• DNA Purification: Use silica-based columns or magnetic beads to isolate DNA [3].
• Quality Assessment: Check DNA purity and concentration using spectrophotometry [3].

PCR Amplification:

• Primer Design: Design species-specific primers for target DNA sequences [3].
• Amplification: Set up PCR reactions with 10-100 ng DNA template, primers, dNTPs, and DNA polymerase [3].
• Detection: Use gel electrophoresis, real-time PCR, or digital PCR for amplification detection [3].

Proteomic Authentication Protocol

Protein Extraction:

• Extraction: Homogenize food sample in lysis buffer containing urea, thiourea, and protease inhibitors [97].
• Cleanup: Perform protein precipitation using acetone or TCA/acetone [97].
• Digestion: Digest proteins with trypsin (1:50 enzyme-to-protein ratio) at 37°C overnight [97].

LC-MS/MS Analysis:

• Separation: Use reverse-phase nanoLC with C18 column [97].
• Mass Spectrometry: Perform data-dependent acquisition on high-resolution mass spectrometer [97].
• Database Search: Identify proteins using search engines (Mascot, MaxQuant) against species-specific databases [97].

Applications in Food Authentication

Meat and Seafood Authentication

Genomics provides unambiguous species identification in meat products through DNA barcoding, effectively detecting adulteration of premium meats with cheaper alternatives [3]. Proteomics complements this by characterizing tissue-specific protein profiles and detecting processing-induced modifications [97]. Metabolomics excels in assessing meat quality, freshness, and storage history by tracking metabolite changes resulting from degradation and microbial activity [99]. For example, in poultry, metabolomics can differentiate processing methods and storage conditions by monitoring metabolite profiles [99].

High-Value Oil and Seed Authentication

The authentication of high-value oils like olive oil presents unique challenges. Genomics can identify the botanical origin through DNA analysis, though DNA degradation in oil matrices can limit effectiveness [3]. Metabolomics has proven highly effective for geographical origin verification of olive oil by detecting metabolite patterns influenced by growing conditions [3] [60]. In seed authentication, non-targeted metabolomics successfully distinguishes between chia, linseed, and sesame seeds, even in processed products like cookies, by identifying unique metabolic markers such as sesamol in chia and succinic acid monomethylester in linseed [60].

Dairy and Fermented Products

Proteomics enables the verification of dairy product authenticity by characterizing milk protein profiles and detecting adulteration with milk from different species [96]. Metabolomics tracks fermentation processes and identifies metabolite markers indicative of product quality and maturation state [98]. Genomics can detect microbial contaminants and verify starter cultures in fermented products [96].

Table 2: Application-Specific Performance of Omics Technologies

Food Category	Authentication Challenge	Metabolomics Performance	Genomics Performance	Proteomics Performance
Meat Products	Species substitution	Moderate (monitors quality)	Excellent (definitive ID)	Good (species markers)
Seafood	Species mislabeling	Good (freshness indicators)	Excellent (definitive ID)	Good (protein profiling)
Olive Oil	Geographical origin	Excellent (chemical terroir)	Moderate (DNA degradation issues)	Limited
Grains & Seeds	Varietal authentication	Excellent (chemical profile)	Good (genetic markers)	Good (storage protein patterns)
Dairy Products	Adulteration	Good (metabolic profile)	Good (microbial content)	Excellent (milk protein variants)
Processed Foods	Ingredient verification	Good (processing markers)	Moderate (DNA stability)	Moderate (protein denaturation)

Integrated Approaches and Complementary Applications

While each omics approach has distinct strengths, their integration provides the most comprehensive solution for food authentication [3] [96]. The combination of genomic species identification with metabolomic quality assessment and proteomic processing verification creates a powerful multi-dimensional authentication system [96]. For instance, genomics can confirm the species origin of meat, proteomics can verify the tissue type and processing history, and metabolomics can assess freshness and storage conditions [3] [97] [99].

Multi-omics integration is particularly valuable for addressing complex authentication challenges such as geographical origin verification, where genetic factors, environmental influences, and processing methods collectively contribute to the food's characteristics [3]. Studies have demonstrated that combining proteomic and metabolomic analyses provides superior discrimination of products from different regions compared to single-omics approaches [96].

Multi-Omics Integration for Comprehensive Food Authentication

The Scientist's Toolkit: Essential Research Reagents and Materials

Table 3: Essential Research Reagents and Materials for Food Authentication Studies

Category	Item	Specification/Example	Application
Sample Preparation	Extraction Solvent	Acetonitrile:methanol:formic acid (74.9:24.9:0.2)	Metabolite extraction [6]
	Lysis Buffer	Proteinase K containing buffer	DNA extraction for genomic analysis [3]
	Protein Digestion Buffer	Urea/thiourea with trypsin	Protein extraction and digestion [97]
Internal Standards	Isotope-labeled metabolites	l-Phenylalanine-d8, l-Valine-d8	Quality control and normalization in metabolomics [6]
Separation	HILIC Column	Waters Atlantis HILIC Silica	Polar metabolite separation [6]
	Reverse Phase Column	C18 column	Lipophilic compound separation [86]
Analysis	High-Resolution Mass Spectrometer	Q-TOF, Orbitrap systems	Metabolite and protein identification [86] [23]
	PCR Thermocycler	Real-time PCR systems	DNA amplification and detection [3]
Data Analysis	Processing Software	XCMS, MZmine, Compound Discoverer	Metabolomics data preprocessing [23] [6]
	Statistical Packages	R packages (metabolomics, mixOmics)	Multivariate data analysis [23] [60]

Metabolomic, genomic, and proteomic markers each offer unique advantages for food authentication, with their relative effectiveness depending on the specific authentication challenge. Metabolomics provides unparalleled insight into food quality, processing history, and geographical origin through comprehensive chemical profiling. Genomics delivers definitive species identification with high specificity and sensitivity. Proteomics bridges the gap between genotype and phenotype, offering information on functional properties and processing effects. The future of food authentication lies in integrated multi-omics approaches that leverage the complementary strengths of these technologies, providing comprehensive solutions to increasingly sophisticated food fraud challenges. Non-targeted metabolomics serves as a particularly powerful discovery tool within this framework, enabling the identification of novel markers without prior hypothesis and adapting to emerging authentication needs.

Food authenticity assessment is undergoing a transformative shift, moving from targeted analysis of single compounds to a holistic, systems-level approach. Non-targeted metabolomics has emerged as a powerful tool for detecting food fraud and verifying origin by providing a comprehensive molecular fingerprint of food products [8]. However, the metabolome alone represents just the final output of complex biological processes. The true future of this field lies in multi-omic data integration, where metabolomics is synergistically combined with genomics, transcriptomics, and proteomics. This integrated "foodomics" approach provides a deeper, mechanistic understanding of the biochemical pathways that define a food's identity, moving beyond correlation to establish causation and significantly enhancing the robustness of authentication models [100] [101] [102]. By leveraging artificial intelligence (AI) and advanced computational tools, researchers can now integrate these vast, complex datasets to build predictive models for food quality, safety, and traceability, ultimately ensuring a more transparent and secure global food supply chain [100] [103].

The Multi-Omic Toolkit for Food Authentication

A multi-omics approach leverages several analytical layers to build a complete biological story. The table below summarizes the key omics disciplines and their roles in authenticating food products.

• Table 1: Core Omics Technologies in Food Authentication

  Omics Discipline	Analytical Focus	Key Authentication Application	Common Analytical Platforms
  Genomics	DNA sequence and structure [101]	Species identification, verification of botanical or animal origin, detection of adulteration with non-declared species [101].	PCR, MLVA, PFGE, 16S rRNA sequencing [101].
  Transcriptomics	Complete set of RNA transcripts [101]	Understanding gene expression in response to origin, stress, or fraud; rarely used directly on finished food products.	Microarrays, RNA-Seq.
  Proteomics	Protein abundance and post-translational modifications [101]	Detection of protein markers for species, geographic origin, and processing conditions (e.g., heat treatment) [101].	LC-MS/MS, Q-TOF, Orbitrap, Ion Mobility MS (timsTOF) [101].
  Metabolomics	Small-molecule metabolites (<1000 Da) [101]	Geographic origin discrimination, detection of adulterants, assessment of freshness, and verification of production methods (e.g., organic vs. conventional) [86] [8] [104].	LC-MS (Q-TOF, Orbitrap), GC-MS, NMR spectroscopy [86] [101] [105].
  Lipidomics	Lipids and lipophilic compounds [101]	Authentication of fats and oils, detection of unauthorized oil blending, verification of dairy fat purity [101].	LC-MS (Q-TOF, Orbitrap).

The power of multi-omics is unlocked not by using these technologies in isolation, but by integrating them. For instance, genomic analysis can confirm the species of a meat sample, while proteomic and metabolomic profiles can reveal its geographical origin and whether it has been frozen and thawed [8] [101]. This layered evidence creates a far more defensible authentication model.

Computational Frameworks for Data Integration

The primary challenge in multi-omics is the computational integration of heterogeneous, high-dimensional datasets. Several strategies and tools have been developed to address this.

• Visual Integration Tools: Tools like the Pathway Tools (PTools) Cellular Overview enable simultaneous visualization of up to four omics datasets (e.g., transcriptomics, proteomics, metabolomics) on organism-scale metabolic network diagrams [106]. Different data types can be painted onto different visual channels—such as reaction arrow color or thickness, and metabolite node color or thickness—allowing researchers to visually correlate changes across molecular layers within their functional metabolic context [106].
• AI and Machine Learning-Driven Integration: Artificial Intelligence (AI) and machine learning algorithms are critical for analyzing complex, integrated omics data to predict outcomes and guide decisions [100]. These computational methods can identify key genes, proteins, and metabolites that contribute to nutritional traits and stress resilience in crops, and by extension, to authentication markers [100]. AI models can integrate multi-omics data to optimize crop traits and pinpoint the most informative biomarkers for distinguishing, for example, different geographical origins of a food product [100].
• Data Fusion and Chemometrics: Statistical and chemometric methods remain foundational. Techniques like Principal Component Analysis (PCA) and Orthogonal Partial Least Squares (OPLS) are used to extract meaningful information from complex spectral data from NMR or MS [86] [104] [102]. When applied to fused multi-omics datasets, these methods can identify the combined variables that most effectively differentiate authentic from adulterated samples.

The following diagram illustrates the conceptual workflow for integrating multi-omics data, from acquisition to biological insight.

Multi-Omic Data Integration Workflow

Application Notes & Experimental Protocols

This section provides a detailed, actionable protocol for a food authentication study using a multi-omics approach, with a focus on non-targeted metabolomics.

Application Note: Authenticating the Geographical Origin of Honey

Background: High-value honey is often targeted for fraudulent mislabeling of its geographical origin. A single-omics approach (e.g., metabolomics) can discriminate origins but may lack the robustness for regulatory enforcement. Multi-Omic Solution: Integrating NMR-based metabolomics with LC-MS proteomics (focusing on pollen and honeybee proteins) to create a definitive origin signature. Outcome: The integrated model significantly improves classification accuracy compared to metabolomics alone, as it combines environmental metabolite influences (metabolomics) with direct biological evidence of the foraging region (proteomics from pollen).

Detailed Protocol: An Integrated LC-MS Metabolomics and Proteomics Workflow

Objective: To distinguish between authentic and adulterated olive oil samples.

I. Sample Preparation

• Metabolite Extraction: Weigh 100 ± 1 mg of homogenized sample. Add 1 mL of cold methanol:water (80:20, v/v) containing internal standards. Vortex vigorously for 1 min, sonicate in an ice bath for 10 min, and incubate at -20°C for 1 hour to precipitate proteins. Centrifuge at 14,000 × g for 15 min at 4°C. Transfer the supernatant to a new vial for LC-MS analysis [86] [8].
• Protein Extraction and Digestion: For the same sample, resuspend the protein pellet in a denaturing buffer. Reduce with dithiothreitol, alkylate with iodoacetamide, and digest with trypsin overnight at 37°C. Desalt the resulting peptides using C18 solid-phase extraction tips prior to LC-MS/MS analysis [101].

II. Data Acquisition

• LC-MS Metabolomics (Non-targeted):
  • Chromatography: UHPLC system with a C18 column (e.g., 2.1 x 100 mm, 1.7 µm). Mobile phase A: 0.1% formic acid in water; B: 0.1% formic acid in acetonitrile. Gradient: 5% B to 100% B over 20 min.
  • Mass Spectrometry: Q-TOF mass analyzer operating in both positive and negative ESI mode. Data acquired in data-independent acquisition (DIA) or MS^E mode to fragment all ions, collecting both precursor and fragment ion data [86] [105].
• LC-MS/MS Proteomics (Data-Dependent Acquisition):
  • Chromatography: Nano-UHPLC system with a C18 nano-column.
  • Mass Spectrometry: High-resolution mass spectrometer (e.g., Orbitrap Eclipse Tribrid) coupled with an ion mobility cell (FA-IMS). Full MS scan followed by DDA of the top N most intense ions for MS/MS fragmentation [101].

III. Data Processing and Integration

• Metabolomics Data: Preprocess raw data using software like XCMS for peak picking, alignment, and retention time correction. Generate a peak table with metabolite features (m/z, RT, intensity) [86].
• Proteomics Data: Process raw files using search engines (e.g., MaxQuant) against a relevant protein database for identification and quantification.
• Data Integration: Use multi-omics visualization tools like PTools [106] or statistical fusion techniques in R/Python. Map significantly altered metabolites and proteins onto KEGG metabolic pathways to identify disrupted biological modules that serve as powerful authentication markers.

The following workflow diagram outlines the key experimental and computational steps.

Integrated Metabolomics & Proteomics Workflow

The Scientist's Toolkit: Essential Research Reagents & Materials

Successful multi-omics studies rely on a suite of reliable reagents, analytical columns, and bioinformatics tools. The following table details key components of the integrated workflow described in the protocol.

• Table 2: Research Reagent Solutions for Multi-Omic Food Authentication

  Item Name	Function/Application	Brief Rationale
  C18 UHPLC Column (e.g., 2.1 x 100 mm, 1.7 µm)	Chromatographic separation of metabolites in non-targeted metabolomics [86].	Provides high-resolution separation of complex metabolite mixtures, reducing ion suppression and increasing metabolite coverage [86].
  C18 Nano-UHPLC Column	Chromatographic separation of peptides in proteomics [101].	Essential for sensitive nano-LC-MS/MS systems, enabling high-efficiency separation and identification of low-abundance proteins [101].
  Trypsin (Sequencing Grade)	Enzymatic digestion of extracted proteins into peptides for bottom-up proteomics [101].	Highly specific protease that cleaves at lysine and arginine, generating peptides suitable for MS analysis and database searching [101].
  Mass Spectrometry Internal Standards (e.g., stable isotope-labeled metabolites/amino acids)	Quality control and normalization of MS data for both metabolomics and proteomics.	Corrects for instrument variability and matrix effects, improving quantitative accuracy and reproducibility across batches [86].
  Pathway Tools (PTools) Software	Visual integration and painting of multiple omics datasets onto metabolic pathway diagrams [106].	Allows for the simultaneous visualization of up to four omics data types (e.g., transcript, protein, metabolite levels) in a functional metabolic context, facilitating biological interpretation [106].

The future of food authentication is inextricably linked to the advancement of multi-omic data integration. While non-targeted metabolomics provides a powerful fingerprint, its combination with genomics, proteomics, and transcriptomics creates a multi-layered, defensible body of evidence that is far more resistant to fraud. The convergence of these advanced analytical technologies with AI-driven bioinformatics and intuitive visualization tools is creating a new paradigm [100] [106] [103]. This paradigm moves the field from simply detecting adulteration to deeply understanding the biological basis of food origin, quality, and safety. As these integrated workflows become more standardized and accessible, they will empower researchers and regulatory bodies to ensure greater transparency and integrity within the global food system.

In the evolving field of food authentication, non-targeted metabolomics has emerged as a powerful strategy to combat fraudulent practices and verify food integrity. Unlike targeted analyses that focus on specific known compounds, non-targeted fingerprinting approaches comprehensively capture the complex metabolic profile of a sample, providing a unique chemical descriptor that can be used for authentication purposes. The fundamental premise is that a food's metabolome reflects not only its genetic makeup but also environmental factors, production practices, and processing methods, creating a distinguishable signature that can be detected through appropriate analytical and chemometric techniques [107].

The transition of these non-targeted methods from proof-of-concept research to reliable tools for routine control laboratories hinges on rigorous validation and a clear demonstration of fitness-for-purpose. This protocol outlines a structured approach for developing, validating, and implementing non-targeted metabolomic methods, with a specific focus on High-Performance Liquid Chromatography with Ultraviolet detection (HPLC-UV) fingerprinting for meat authentication as a representative application.

Experimental Design and Workflow

A fit-for-purpose non-targeted metabolomics study follows a systematic workflow from sample preparation to data interpretation. The key stages are outlined in the diagram below, which provides a visual guide to the entire process.

Sample Preparation and Data Acquisition

The initial phase focuses on generating robust chemical fingerprints from samples.

1. Sample Collection and Extraction:

• Sample Set: A diverse and representative set of samples is crucial. For meat authentication, this includes samples from different species (e.g., lamb, beef, pork, poultry) and those with different attributes (e.g., Protected Geographical Indication (PGI), organic, Halal, Kosher) [107].
• Extraction: A simple, reproducible extraction protocol is employed. For the meat authentication example, a water extraction procedure is used to solubilize polar metabolites, followed by processing with solvents like methanol (Chromosolv for HPLC) to precipitate proteins and extract a broad range of compounds [107].

2. HPLC-UV Fingerprinting:

• Principle: This approach uses a standard HPLC-UV system to generate a chromatographic profile ("fingerprint") of the extract without targeting specific metabolites. The entire chromatogram serves as a unique pattern for each sample type [107].
• Method: An optimized gradient elution method is developed to separate as many compounds as possible within a reasonable runtime. The UV-Vis detector captures absorbance across a range of wavelengths, generating a rich, multi-dimensional data point for each sample.

Data Processing and Statistical Analysis

The acquired fingerprints are processed and analyzed to build a predictive model.

1. Data Pre-processing: Raw chromatographic data are processed to correct for baseline drift, align retention times, and reduce noise. The data is often normalized to correct for variations in sample concentration or instrument response [40] [108].

2. Chemometric Analysis: Multivariate statistical techniques are applied to the processed data matrix.

• Unsupervised Learning: Principal Component Analysis (PCA) is used for exploratory data analysis to identify natural clustering and detect outliers [107].
• Supervised Learning: Partial Least Squares-Discriminant Analysis (PLS-DA) is used to build a classification model. This technique maximizes the separation between pre-defined sample classes (e.g., beef vs. pork, organic vs. conventional) [107].

Table 1: Performance Metrics for a PLS-DA Model Discriminating Meat Species

Performance Metric	Calibration/Cross-Validation Value	Prediction Value (Decision Tree)
Sensitivity	> 100%	100%
Specificity	> 99.3%	100%
Classification Error	< 0.4%	0%

The model performance, as demonstrated in a meat speciation study, can achieve excellent sensitivity and specificity, with classification errors below 0.4% for calibration and 100% accuracy in prediction when using a hierarchical decision tree model [107].

Validation Framework for Non-Targeted Methods

For a method to be deemed fit-for-purpose, it must undergo a thorough validation process that goes beyond single-laboratory studies. The following diagram illustrates the multi-faceted validation strategy required.

The validation framework should be comprehensive, assessing the method's analytical and statistical robustness [77].

1. Analytical Method Validation: This ensures the reliability of the chemical analysis itself.

• Precision/Repeatability: The method should demonstrate low variation when analyzing the same sample multiple times (repeatability) and over different days (intermediate precision).
• Robustness: The method's performance should be stable against small, deliberate variations in analytical parameters (e.g., mobile phase pH, column temperature).
• Specificity: The fingerprint, combined with the chemometric model, must be able to distinguish between the defined classes without confusion.

2. Statistical Model Validation: This assesses the performance and predictive power of the chemometric model.

• Cross-validation: Techniques like leave-one-out or k-fold cross-validation are used to estimate how the model will generalize to an independent data set.
• Use of an Independent Test Set: The model's performance is ultimately tested on a set of samples that were not used in model building or cross-validation.
• Figures of Merit: Metrics such as sensitivity, specificity, and classification error are calculated, as shown in Table 1 [107].

3. System Challenge: The method should be tested with samples that represent real-world challenges, including:

• Adulteration Detection: The method should identify and, if possible, quantify adulteration. For instance, PLS regression has been used to detect and quantify adulteration levels of non-PGI, non-organic, or non-Halal meats in authentic products in the range of 15-85% with low prediction errors (<6.6%) [107].
• Long-Term Stability: The model's performance should be monitored over time and re-calibrated if significant drift is detected.

Table 2: Performance of HPLC-UV Fingerprinting for Various Meat Authentication Tasks

Authentication Task	Overall Sensitivity	Overall Specificity	Classification Error
Geographical Origin (PGI)	> 91.2%	> 91.2%	< 6.9%
Organic Production	> 91.2%	> 91.2%	< 6.9%
Halal and Kosher Products	> 91.2%	> 91.2%	< 6.9%

Implementation for Routine Control

The Scientist's Toolkit: Essential Research Reagent Solutions

Implementing a non-targeted metabolomics workflow requires specific materials and computational tools. The following table details key components.

Table 3: Essential Research Reagent Solutions and Computational Tools

Item	Function/Description	Application in Workflow
Methanol (HPLC grade)	Organic solvent for protein precipitation and metabolite extraction from solid food matrices.	Sample Preparation [107]
Water (HPLC grade)	Aqueous solvent for extraction and as a mobile phase component in HPLC.	Sample Preparation, Data Acquisition [107]
Chemical Standards	Authentic metabolite standards for instrument calibration and method development.	System Suitability, Method Development
HPLC-UV System	Instrumentation comprising pumps, autosampler, column oven, and UV-Vis detector for generating chromatographic fingerprints.	Data Acquisition [107]
Post-column Derivatization Reagents (e.g., for MCheM)	Reagents targeting specific functional groups (e.g., amines, carboxylic acids) to add a layer of chemical reactivity data for improved metabolite annotation.	Advanced Data Acquisition [109]
MetaboAnalyst	A web-based platform for comprehensive metabolomic data processing, normalization, and statistical analysis (PCA, PLS-DA, etc.).	Data Processing & Statistical Analysis [108]
GNPS (Global Natural Products Social Molecular Networking)	An online platform for MS/MS data sharing, molecular networking, and annotation.	Metabolite Annotation & Data Integration [40] [109]

A Protocol for Meat Authentication via HPLC-UV Fingerprinting and PLS-DA

Objective: To authenticate meat samples based on species, geographical origin, or production attributes using a validated HPLC-UV fingerprinting method.

Materials and Equipment:

• HPLC system equipped with a UV-Vis or photodiode array (PDA) detector
• Reverse-phase C18 column
• HPLC-grade methanol and water
• Meat samples (authentic and test samples)
• Centrifuge and vortex mixer

Step-by-Step Procedure:

• Sample Extraction:

  • Homogenize 1.0 g of meat sample with 10 mL of ultrapure water.
  • Vortex vigorously for 2 minutes, then centrifuge at 10,000 × g for 15 minutes.
  • Transfer the supernatant and mix with 20 mL of methanol to precipitate proteins.
  • Centrifuge again at 10,000 × g for 15 minutes and filter the final supernatant through a 0.22 µm membrane before HPLC analysis [107].

• HPLC-UV Analysis:

  • Inject 10 µL of the filtered extract onto the HPLC system.
  • Use a binary gradient with mobile phase A (water with 0.1% formic acid) and B (methanol with 0.1% formic acid).
  • Run a linear gradient from 5% B to 95% B over 25 minutes, with a flow rate of 1.0 mL/min.
  • Monitor the effluent with the UV detector, collecting full spectra in the 200-400 nm range [107].

• Data Pre-processing and Model Building:

  • Export the chromatographic data as a peak intensity table (features defined by retention time and absorbance).
  • Import the data into a statistical software package (e.g., MetaboAnalyst [108]).
  • Apply normalization (e.g., probabilistic quotient normalization) and Pareto scaling.
  • Perform PCA to visualize overall data structure.
  • Develop a PLS-DA model using the known class labels of the training set samples.

• Model Validation and Deployment:

  • Validate the model using cross-validation and an independent test set of authentic samples.
  • Establish classification rules and performance metrics (sensitivity, specificity).
  • Apply the validated model to classify unknown test samples.

Non-targeted metabolomics, exemplified by the HPLC-UV fingerprinting approach, provides a powerful and fit-for-purpose solution for complex authentication challenges in food control. Its strength lies in its ability to detect patterns resulting from a multitude of factors that are invisible to genetic tools. By adhering to a rigorous workflow that encompasses robust analytical techniques, comprehensive chemometric analysis, and a thorough validation framework, these methods can transition from promising research concepts to reliable tools for routine surveillance. This ensures the integrity of the food supply chain, protects consumers from economic fraud and health risks, and upholds the value of authentic, high-quality products.

Conclusion

Non-targeted metabolomics has firmly established itself as a powerful, data-driven tool for food authentication, capable of detecting unforeseen fraud and verifying complex claims like geographical origin. The integration of advanced analytical platforms like high-resolution MS and NMR with sophisticated machine learning algorithms has enabled the discovery of robust biomarker patterns, even in processed foods. However, the transition from promising research to routine application hinges on overcoming key challenges: the development of standardized, harmonized, and fully validated methods, along with strategies to manage the effects of processing and natural variability. Future progress will be driven by the wider adoption of 'Foodomics'—the integration of metabolomics with genomics and proteomics—to build a more comprehensive understanding of food identity. For biomedical and clinical research, the rigorous validation frameworks and data analysis pipelines pioneered in food science offer a valuable template for applications in biomarker discovery, nutrient profiling, and understanding the intricate links between diet and health. The ongoing development of portable analytical equipment and mobile data analysis software promises to further decentralize and accelerate food authenticity testing, enhancing safety and trust across global supply chains.

References

• 1. Food Fraud Costs the Global Food Industry $10-15 Billion Annually [https://www.ifis.org/blog/global-food-fraud]
• 2. Food Authentication: Combating Fraud in Global Supply Chains | Technology Networks [https://www.technologynetworks.com/applied-sciences/articles/authentic-foods-and-where-to-find-them-challenges-and-solutions-in-food-fraud-testing-394030]
• 3. Foodomics as a Tool for Evaluating Food Authenticity and Safety from Field to Table: A Review - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC11719641/]
• 4. Food authentication and traceability: An Asian and Australian perspective - ScienceDirect [https://www.sciencedirect.com/science/article/abs/pii/S0956713516300299]
• 5. Untargeted metabolomic approaches in food authenticity: A review... [https://pubmed.ncbi.nlm.nih.gov/35986993/]
• 6. A Protocol for Untargeted Metabolomic Analysis: From Sample Preparation to Data Processing - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC9284939/]
• 7. Promoting LC-QToF based non - targeted fingerprinting and biomarker... [https://pure.qub.ac.uk/en/publications/promoting-lc-qtof-based-non-targeted-fingerprinting-and-biomarker]
• 8. Application of the Metabolomics Approach in Food Authentication - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC8706891/]
• 9. Opportunities and Challenges Using Non-targeted Methods for Food Fraud Detection - PubMed [https://pubmed.ncbi.nlm.nih.gov/31322874/]
• 10. of data processing strategies using commercial vs.... Comparison [https://pmc.ncbi.nlm.nih.gov/articles/PMC11249438/]
• 11. Recent Advancements in the Application of Metabolomics in Food ... [https://link.springer.com/article/10.1007/s43555-025-00073-9]
• 12. Uncovers Untargeted Risks: Polystyrene... Metabolomics Food Safety [https://pubmed.ncbi.nlm.nih.gov/40870694/]
• 13. analysis reveals Untargeted meat quality... metabolomics comparative [https://bmcgenomics.biomedcentral.com/articles/10.1186/s12864-025-12008-8]
• 14. Uncovers Untargeted Risks: Polystyrene... Metabolomics Food Safety [https://www.mdpi.com/2304-8158/14/16/2781]
• 15. Emerging applications of metabolomics in food science and future trends - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC9743159/]
• 16. Global food surges in fraud 2025 [https://www.foodnavigator.com/Article/2025/06/16/global-food-fraud-surges-in-2025/]
• 17. A guide to reverse metabolomics—a framework for big data discovery strategy | Nature Protocols [https://www.nature.com/articles/s41596-024-01136-2]
• 18. Data processing solutions to render metabolomics more quantitative: case studies in food and clinical metabolomics using Metabox 2.0 - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC10941642/]
• 19. Fingerprinting and Origin | SGS [https://www.sgs.com/en/services/fingerprinting-and-origin]
• 20. Current Progress and Future Trends of Genomics-Based Techniques for Food Adulteration Identification [https://www.mdpi.com/2304-8158/14/7/1116]
• 21. Frontiers | Plant microbiomes as contributors to agricultural terroir [https://www.frontiersin.org/journals/agronomy/articles/10.3389/fagro.2023.1216520/full]
• 22. Towards a scientific interpretation of the terroir concept: plasticity of... [https://bmcplantbiol.biomedcentral.com/articles/10.1186/s12870-015-0584-4]
• 23. Guide to Metabolomics Analysis: A Bioinformatics Workflow - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC9032224/]
• 24. UNL | Powers Group | Differential Metabolomics [https://bionmr.unl.edu/metabolomics.php]
• 25. Nontargeted metabolomics : PTFI Platform | Alliance Bioversity... [https://alliancebioversityciat.org/publications-data/nontargeted-metabolomics-ptfi-platform]
• 26. Functional Analysis of Global Metabolomics [https://www.metaboanalyst.ca/resources/vignettes/Functional_Analysis_global_metabolomics.html]
• 27. analysis, visualization, and integration Metabolomics data [https://pubmed.ncbi.nlm.nih.gov/18287705/]
• 28. - Non Authentication Approach for Extra Virgin Targeted - PMC Olive Oil [https://pmc.ncbi.nlm.nih.gov/articles/PMC7073519/]
• 29. From Targeted Quantification to Untargeted Metabolomics | IntechOpen [https://www.intechopen.com/chapters/75755]
• 30. Assessment of metabolome diversity in black and white pepper in response to autoclaving using MS- and NMR-based metabolomics and in relation to its remote and direct antimicrobial effects against food-borne pathogens - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC10989240/]
• 31. Metabolomics based inferences to unravel phenolic compound diversity in cereals and its implications for human gut health - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC9449372/]
• 32. LC-HRMS-Based Non - Targeted for the Assessment of... Metabolomics [https://pmc.ncbi.nlm.nih.gov/articles/PMC9607529/]
• 33. Development, characterization and comparisons of targeted and non-targeted metabolomics methods | PLOS One [https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0207082]
• 34. - Non Metabolite Profiles and Sensory Properties Elucidate... Targeted [https://pmc.ncbi.nlm.nih.gov/articles/PMC7344679/]
• 35. studies by Olive and nontarget LC–QTOF-MS... oil authenticity target [https://link.springer.com/article/10.1007/s00216-016-9891-3]
• 36. - Non Targeted Metabolomics [https://www.news-medical.net/life-sciences/Non-Targeted-Metabolomics.aspx]
• 37. Application of the Metabolomics Approach in Food Authentication - PMC [https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8706891/]
• 38. A standardized nontargeted metabolomics method for cross-laboratory comparison of food profiles - PubMed [https://pubmed.ncbi.nlm.nih.gov/40829463/]
• 39. Frontiers | Editorial: Current trends in targeted and non - targeted ... [https://www.frontiersin.org/journals/molecular-biosciences/articles/10.3389/fmolb.2025.1614399/full]
• 40. Statistical analysis of feature-based molecular networking results from... [https://orbit.dtu.dk/en/publications/statistical-analysis-of-feature-based-molecular-networking-result/]
• 41. Chromatography hyphenated to high resolution mass spectrometry in untargeted metabolomics for investigation of food (bio)markers - ScienceDirect [https://www.sciencedirect.com/science/article/abs/pii/S0165993620303903]
• 42. Emerging Techniques: GCMS Food Safety and Food Quality... [https://www.thermofisher.com/blog/food/emerging-techniques-gas-chromatography-mass-spectrometry-for-food-applications/]
• 43. High resolution mass spectrometry for structural identification of metabolites in metabolomics | Metabolomics [https://link.springer.com/article/10.1007/s11306-015-0882-8]
• 44. - Non – targeted -based differences in Indica rice... GC MS metabolomics [https://pmc.ncbi.nlm.nih.gov/articles/PMC11162567/]
• 45. - Non - targeted -based... | Research Square GC MS metabolomics [https://www.researchsquare.com/article/rs-4158043/v1]
• 46. The Role of Liquid Chromatography-Mass Spectrometry in Food ... [https://www.intechopen.com/chapters/66149]
• 47. of data processing strategies using commercial vs.... Comparison [https://link.springer.com/article/10.1007/s00216-024-05347-0]
• 48. - Non – targeted /MS LC profiling of human plasma... MS metabolomic [https://pmc.ncbi.nlm.nih.gov/articles/PMC10709258/]
• 49. - Non – targeted /MS LC profiling of human plasma... MS metabolomic [https://pure.qub.ac.uk/en/publications/non-targeted-lcmsms-metabolomic-profiling-of-human-plasma-uncover]
• 50. Chemical Profiling and Quality Assessment of Food Products Employing Magnetic Resonance Technologies - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC12294631/]
• 51. Applications In Nuclear Magnetic Resonance Spectroscopy Foods [https://www.academia.edu/118335706/Nuclear_Magnetic_Resonance_Spectroscopy_Applications_In_Foods]
• 52. Non-Targeted Nuclear Analysis for Magnetic ... Resonance Food [https://pmc.ncbi.nlm.nih.gov/articles/PMC11434360/]
• 53. -based NMR approach for the assessment of traceability... metabolomic [https://iris.unitn.it/handle/11572/410566]
• 54. Spectroscopy NMR for Protocols Food Metabolomics Applications [https://link.springer.com/protocol/10.1007/978-1-4939-7643-0_14]
• 55. The Science Behind Food Authentication [https://www.numberanalytics.com/blog/science-behind-food-authentication]
• 56. Trust Your Food : Authenticity Matters - Food Safety Works [https://foodsafetyworks.com/insights/trust-your-food-authenticity-matters/]
• 57. Synergistic approaches of AI and NMR in enhancing food component analysis: A comprehensive review - ScienceDirect [https://www.sciencedirect.com/science/article/abs/pii/S0924224424005284]
• 58. Low-field NMR-based deep learning for non-destructive quality assessment of frozen model foods - ScienceDirect [https://www.sciencedirect.com/science/article/abs/pii/S030881462502432X]
• 59. Use of NMR relaxometry for determination of meat properties a brief... [https://maxapress.com/article/doi/10.48130/FMR-2022-0008]
• 60. of Discovery identity food by markers and machine... metabolomics [https://www.nature.com/articles/s41598-019-46113-y?error=cookies_not_supported&code=eee42127-b2ed-4c8e-918d-310a76046c22]
• 61. in Machine Learning - GeeksforGeeks Random Forest Algorithm [https://www.geeksforgeeks.org/machine-learning/random-forest-algorithm-in-machine-learning/]
• 62. Understanding Random Forest | Towards Data Science [https://towardsdatascience.com/understanding-random-forest-58381e0602d2/]
• 63. Medical Records » Submission » Machine Learning-Driven... [https://dergipark.org.tr/en/pub/medr/issue/92094/1718952]
• 64. - Non and targeted reveal metabolic... metabolomics machine learning [https://pubmed.ncbi.nlm.nih.gov/40695070/]
• 65. - Non LC-MS Targeted Approach towards an... Metabolomics [https://pmc.ncbi.nlm.nih.gov/articles/PMC8466891/]
• 66. Non-Targeted LC-MS Metabolomics Approach towards an Authentication of the Geographical Origin of Grain Maize (Zea mays L.) Samples [https://www.mdpi.com/2304-8158/10/9/2160]
• 67. Untargeted metabolomics by liquid chromatography-mass... [https://pubmed.ncbi.nlm.nih.gov/35347871/]
• 68. Food Phenotyping: Recording and Processing of Non-Targeted Liquid Chromatography Mass Spectrometry Data for Verifying Food Authenticity - PMC [https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7504784/]
• 69. Metabolomics Data Analysis, Visualization, and Integration | SpringerLink [https://link.springer.com/protocol/10.1007/978-1-59745-535-0_20]
• 70. MetaboAnalyst [https://www.metaboanalyst.ca/]
• 71. Analysis of dog meat adulteration in beef meatballs using non - targeted ... [https://link.springer.com/article/10.1007/s44211-023-00470-x]
• 72. Untargeted Metabolomics Reveals New Markers of Food ... Processing [https://pmc.ncbi.nlm.nih.gov/articles/PMC9653871/]
• 73. Untargeted Metabolomics Services | LC‑MS/MS Global Profiling & Biomarker Discovery - MetwareBio [https://www.metwarebio.com/untargeted-metabolomics/]
• 74. Validated LC-MS/MS Method for the Quantification of... | PLOS One [https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0164967]
• 75. Frontiers | TurboPutative: A web server for data handling and metabolite classification in untargeted metabolomics [https://www.frontiersin.org/journals/molecular-biosciences/articles/10.3389/fmolb.2022.952149/full]
• 76. profiling goes green: Sustainable analysis strategies for Food ... food [https://pubmed.ncbi.nlm.nih.gov/39140227/]
• 77. Review of validation and reporting of non - targeted fingerprinting... [https://pubmed.ncbi.nlm.nih.gov/26231890/]
• 78. Reference Standardization for Quantification and Harmonization of Large-Scale Metabolomics - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC7887762/]
• 79. Progress in Metabolomics Standardisation and its Significance in Future Clinical Laboratory Medicine - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC5282916/]
• 80. Development of NIST Reference Materials for Untargeted Metabolomics, Lipidomics and Proteomics | NIST [https://www.nist.gov/programs-projects/development-nist-reference-materials-untargeted-metabolomics-lipidomics-and]
• 81. Guidelines and considerations for the use of system suitability and quality control samples in mass spectrometry assays applied in untargeted clinical metabolomic studies | Metabolomics [https://link.springer.com/article/10.1007/s11306-018-1367-3]
• 82. Quality assurance and quality control reporting in untargeted metabolic phenotyping: mQACC recommendations for analytical quality management - PMC [https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9420093/]
• 83. Discovery of food identity markers by metabolomics and machine learning technology | Scientific Reports [https://www.nature.com/articles/s41598-019-46113-y]
• 84. Establishing a framework for best practices for quality assurance and quality control in untargeted metabolomics | Metabolomics [https://link.springer.com/article/10.1007/s11306-023-02080-0]
• 85. Proposed minimum reporting standards for chemical analysis Chemical Analysis Working Group (CAWG) Metabolomics Standards Initiative (MSI) - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC3772505/]
• 86. Phenotyping: Recording and Processing of Food - Non Liquid... Targeted [https://pmc.ncbi.nlm.nih.gov/articles/PMC7504784/]
• 87. Mass-spectrometry-based metabolomics : limitations and... [https://pmc.ncbi.nlm.nih.gov/articles/PMC2794347/]
• 88. Mass-spectrometry based metabolomics: an overview of workflows, strategies, data analysis and applications | Proteome Science | Full Text [https://proteomesci.biomedcentral.com/articles/10.1186/s12953-025-00241-8]
• 89. Leveraging open cheminformatics tools for non - targeted ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12583387/]
• 90. A Protocol to Compare Methods for Untargeted Metabolomics [https://link.springer.com/protocol/10.1007/978-1-4939-8769-6_1]
• 91. Techniques for Food Authentication: Trends and Emerging Approaches - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC10048066/]
• 92. Can we trust biomarkers identified using different non - targeted ... [https://pubmed.ncbi.nlm.nih.gov/32737683/]
• 93. Inter-laboratory reproducibility of an untargeted metabolomics GC–MS assay for analysis of human plasma | Scientific Reports [https://www.nature.com/articles/s41598-020-67939-x]
• 94. Development, characterization and comparisons of targeted and non-targeted metabolomics methods - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC6237353/]
• 95. Deep Learning-Enhanced Spectroscopic Technologies for Food Quality Assessment: Convergence and Emerging Frontiers - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC12248972/]
• 96. Multi-omics in food safety and authenticity in terms of food components - ScienceDirect [https://www.sciencedirect.com/science/article/abs/pii/S030881462302561X]
• 97. Review on proteomics for food authentication - PubMed [https://pubmed.ncbi.nlm.nih.gov/27389853/]
• 98. Metabolites as Tools for Food , Processing, and... Authentication [https://pmc.ncbi.nlm.nih.gov/articles/PMC8465241/]
• 99. Metabolomics workflow for quality control of differently-processed pre-cooked chicken fillets - ScienceDirect [https://www.sciencedirect.com/science/article/pii/S0308814621020124]
• 100. - Integrative and Artificial Intelligence (AI)-driven... Multi omics [https://www.frontiersin.org/research-topics/65534/integrative-multi-omics-and-artificial-intelligence-ai-driven-approaches-for-superior-nutritional-quality-and-stress-resilience-in-crops/magazine]
• 101. Recent Trends in Food Quality and Authentication : The Role of... [https://pmc.ncbi.nlm.nih.gov/articles/PMC12072367/]
• 102. Advances in Food Metabolomics [https://www.mdpi.com/2076-3417/15/15/8283]
• 103. The need for an integrated - multi approach in... | PNNL OMICs [https://www.pnnl.gov/publications/need-integrated-multi-omics-approach-microbiome-science-food-system]
• 104. A Contribution to the Harmonization of Non - targeted NMR Methods ... [https://link.springer.com/article/10.1007/s12161-019-01664-8]
• 105. Advances in high-resolution mass spectrometry based on metabolomics studies for food--a review - PubMed [https://pubmed.ncbi.nlm.nih.gov/26365870/]
• 106. Frontiers | Visual analysis of multi-omics data [https://www.frontiersin.org/journals/bioinformatics/articles/10.3389/fbinf.2024.1395981/full]
• 107. Meat Authentication Based on Animal Species and Other Quality... [https://link.springer.com/article/10.1007/s12161-025-02840-9]
• 108. MetaboAnalyst: a web server for metabolomic data analysis and interpretation - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC2703878/]
• 109. Expanding Non - Targeted with Multiplexed Chemical... Metabolomics [https://communities.springernature.com/posts/behind-the-paper-expanding-non-targeted-metabolomics-with-multiplexed-chemical-labeling]