Qualitative and Quantitative Mass Spectrometry Approaches for the Analysis of Phenolic Compounds in Complex Natural Matrices

Featured Application

This review maps the physicochemical properties of phenolic acids and flavonoids and explains their diagnostic LC–ESI–MS/MS fragmentation routes. It then details qualitative and quantitative MS analysis, covering solvent choice, control of ionization and adducts, chromatographic selectivity, matrix effects, and calibration. Finally, it synthesizes recent applications across complex natural matrices and offers a practical method development roadmap for robust identification and accurate quantification.

Abstract

Phenolic molecules represent one of the most prevalent and biologically important categories of secondary metabolites. Within this diverse group, phenolic acids and flavonoids are the most extensively studied categories, primarily due to their structural diversity and broad spectrum of reported bioactivities. We first provide an overview of the physicochemical characteristics of flavonoids and phenolic acids and discuss how these properties relate to mass spectrometry (MS) fragmentation patterns and chromatographic behavior, including retention characteristics and isomer resolution. Next, we systematically examine the utilization of MS-based procedures for the characterization of flavonoids and phenolic acids in complex natural matrices. We initially examine targeted liquid chromatography–tandem mass spectrometry (LC–MS/MS) utilizing triple-quadrupole (QQQ) platforms, focusing on selected/multiple reaction monitoring (SRM/MRM) and associated scanning techniques (product-ion and precursor-ion scans). We summarize validated methodologies and strategies for both absolute and relative quantification, including stable-isotope dilution, matrix-matched calibration or standard addition, and internal-standard normalization. We subsequently analyze untargeted high-resolution mass spectrometry methodologies (direct injection and coupled to liquid chromatography), highlighting recent progress in data acquisition while addressing ongoing challenges in computational processing. Finally, we present a brief evaluation of commonly used extraction and clean-up processes, highlighting their practical impact on phenolic recoveries.

1. Introduction

Phenols are a structurally diverse group of secondary metabolites in plants, encompassing phenolic acids, flavonoids, tannins, lignans, and stilbenes, which play key roles in defense against pathogens, oxidative stress, and UV radiation [1]. Over the last decade, a large number of studies have focused on the identification and quantification of phenols from plant material, as well as their evaluation in the context of human health [2,3].

Most plant phenolic compounds derive from the shikimate pathway that leads to phenylpropanoid metabolism [4]. In this pathway, the amino acid phenylalanine is transformed by phenylalanine ammonia-lyase (PAL), cinnamate 4-hydroxylase (C4H), and 4-coumarate-CoA ligase (4CL) into p-coumaroyl-CoA. This metabolite represents a key intermediate that channels the metabolic flux toward hydroxycinnamic acids, lignin precursors, and flavonoids. Flavonoid biosynthesis begins with the activity of chalcone synthase (CHS) and chalcone isomerase (CHI). Subsequent enzymes, including flavanone 3-hydroxylase (F3H), flavonol synthase (FLS), and anthocyanidin synthase (ANS), diversify the core structures. These intermediates are further modified by O-methyltransferases, acyltransferases, and UDP-glycosyltransferases (UGTs) to fine-tune solubility and stability. A conserved MYB–basic helix–loop–helix (bHLH)–WD40 transcriptional complex, known as the MBW complex, orchestrates the regulation of late biosynthetic genes and integrates developmental and environmental signals. This ensures that sample-specific phenolic profiles accurately reflect both pathway architecture and its transcriptional regulation [5]. By linking biosynthesis to measurement, this biochemical framework defines which congeners co-occur and how they vary across samples. The analytical challenge is to separate and identify these near-isobaric, chemically similar species, which remain the central difficulty in phenolic analysis [6]. Mass spectrometry (MS) has become essential in this context, as it enables structural characterization through specific fragmentation patterns [7]. Yet even MS alone is not sufficient, since many phenolic compounds share identical precursor masses (and/or fragments). This limitation makes chromatographic separation an essential step prior to MS analysis, ensuring that compounds with overlapping m/z values can be distinguished. The overall complexity of phenolic compounds analysis is therefore the result of the interplay between their structural diversity, physicochemical similarity, and the methodological requirements for reliable identification.

This review examines two principal categories of phenols: (i) phenolic acids and (ii) flavonoids, highlighting their physicochemical properties, structural characterization, and fragmentation patterns. The methodology section of the paper focuses on MS-based techniques, which are classified into targeted (quantitative) and untargeted (qualitative) approaches. In the targeted approach, we present the full range of quantification strategies, from the classical use of Liquid chromatography-tandem mass spectrometry (LC–MS/MS) with multiple reaction monitoring (MRM) and reference standards for known phenolics, to more advanced applications employing precursor ion, product ion, and neutral loss scans. In contrast, we discuss the untargeted approach in terms of profiling diverse plant matrices and enabling the detection and identification of novel phenolic compounds. Finally, we briefly address extraction procedures, focusing on their potential negative impact on phenolic yield and, depending on extract composition, on the ionization efficiency of samples in MS analysis.

2. Phenolic Compounds

2.1. Phenolic Acids

Phenolic acids are pervasive plant metabolites that enhance plants’ antioxidant capacity, aroma, and defense mechanisms [8]. Although phenolic acids are found in plants in significant total quantities, they are often bound as esters, amides, or glycosides rather than existing as free aglycones. One well-known example is chlorogenic acid, a highly abundant ester of caffeic and quinic acid that is present in coffee [9]. Another representative compound is rosmarinic acid, which is an ester of caffeic acid and 3,4-dihydroxyphenyllactic acid. It is frequently observed in members of the Lamiaceae plant family [10]. In addition to already mentioned esters and glycosides, plant defense mechanisms often involve amide conjugates [11]. These examples emphasize that phenolic acids are primarily stored and transported in conjugated forms, which facilitate their solubility, stability, and functions in plant metabolism [12]. This, in conjunction with their low abundance (µM–nM), presents a challenge for extraction and quantification in complex natural matrices [8].

Given the analytical focus of this review, we first summarize the structural classification of hydroxybenzoic and hydroxycinnamic acids. We then explain how substituent-driven electronic effects govern their reactivity and antioxidant behavior. Finally, we outline the characteristic MS/MS fragmentation patterns in electrospray ionization (ESI), which provide the basis for the targeted transitions used in the following sections.

2.1.1. Structural Classification

Phenolic acids comprise a benzene ring bearing carboxyl and hydroxyl groups and are classified as (i) hydroxybenzoic or (ii) hydroxycinnamic acids [13]. Their antioxidant behavior follows simple Structure–Activity Relationship (SAR) rules: the carboxyl group is electron-withdrawing (ring-deactivating), whereas the phenolic OH group donates electron density and, after H-atom transfer, forms a resonance-stabilized phenoxy radical. Lower O–H bond dissociation enthalpy correlates with higher activity. Hydroxycinnamic derivatives (for example, p-coumaric and ferulic acid) generally show higher antioxidant activity than hydroxybenzoic analogs (such as p-hydroxybenzoic acid and 3,4-hydroxybenzoic acid). This difference arises because the ethylenic side chain separates the carboxyl group from the ring, facilitating hydrogen donation [14]. Within benzoic acids, meta mono-OH is more active than ortho/para mono-OH (deactivated by –COOH). In addition, activity increases with the number of OH groups (example: gallic acid with 3 OH groups have higher antioxidant activity than p-hydroxybenzoic acid (pHBA) with one OH group in para position). Methoxy groups and esterification typically reduce activity; methoxy raises lipophilicity (beneficial in nonpolar media) but lowers aqueous solubility, so ferulic acid (4-hydroxy-3-methoxycinnamic acid) tends to perform better in lipophilic environments, while caffeic acid is more effective in polar ones [12].

2.1.2. Phenolic Acids Fragmentation

Phenolic acids show a relatively simple tandem MS (MS/MS) pattern in ESI (−) dominated by decarboxylation (−44 Da) and a small set of class-diagnostic ions (Figure 1) [15].

Hydroxybenzoic acids behave similarly, but without the vinyl chain. Decarboxylation again dominates and the product m/z reflects ring substitution. Herein, we will give an example of 3,4-dihydroxybenzoic acid (3,4-DHBA) (Figure 1A). In ESI (−) MS/MS, the dominant pathway is decarboxylation of the ring-attached carboxylate, giving m/z 109 = [M–H–CO₂]⁻ (−44 Da), a catecholate-type anion stabilized by resonance; this is the most reliable quantifier transition (153 → 109). A concurrent radical channel may appear as m/z 108 (−45 Da, loss of •CO₂H from the precursor or H• from 109), which increases with higher collision energy (CE) and varies by instrument. Deeper fragmentation of the 109 ion gives m/z 81 (−28 Da, loss of CO) and then m/z 53 (another −28 Da, loss of CO), typically weak and matrix-dependent and thus best treated as auxiliary qualifiers rather than primary identifiers. Some DHBAs can also yield m/z 91 as a low-intensity ring-cleavage product. Overall: 153 → 109 (quantitative transition), with 153 → 108 (or 153 → 91) as qualifiers; 109 → 81 → 53 represent secondary, low-specificity steps.

Hydroxycinnamates are also primarily fragmented through decarboxylation (Figure 1B), resulting in the class-diagnostic and preferred quantifier (179 → 135), m/z 135 = [M–H–CO₂]⁻ (−44 Da) [16]. A robust qualifier (179 → 134) is provided by m/z 134 (−45 Da, •CO₂H), which is the result of a characteristic radical loss. m/z 161 (−18 Da, dehydration) and m/z 107 (−28 Da) are additional, method-dependent ions that can result from further CO loss from 135 (135 → 107). A signal at m/z 145 is not a primary fragment of caffeic acid; it is typically a minor, secondary channel (sequential loss from a dehydrated intermediate or rearrangement at higher CE) and should be used, at most, as a supplemental qualifier after verifying stability across matrices. Practice: 179 → 135 (quantitative transition) with 179 → 134 (qualitative transition); 179 → 161 or 135 → 107 may be incorporated as secondary qualifiers if they are consistently observed.

Because many isomers share elemental composition and the same dominant neutral losses, MS/MS alone cannot localize OH positions on the ring.

Table 1. Comparison of liquid chromatography–tandem mass spectrometry (LC–MS/MS) and physicochemical parameters used for the differentiation of positional isomers of hydroxybenzoic and coumaric acids. Source: [17].

Compound	pKa (Lowest)	Log S (at pH = 2)	Log D (at pH = 2)	Molecular Formula	Exact Mass/Da	Precursor Ion	Fragment Ions
3,4-DHBA (protocatechuic acid)	2.56	−0.579	1.570	C7H6O4	154.0266	153 (-)	109 (-), 108 (-), 81 (-), 53 (-)
2,5-DHBA (gentisic acid)	2.53	−0.572	1.561	C7H6O4	154.0266	153 (-)	109 (-), 108 (-), 81 (-), 53 (-)
o-coumaric acid	4.42	−1.675	1.831	C9H8O3	164.0473	163 (-)	145 (-), 119 (-), 93 (-)
m-coumaric acid	4.17	−1.674	1.830	C9H8O3	164.0473	163 (-)	145 (-), 119 (-), 93 (-)
p-coumaric acid	4.20	−1.674	1.829	C9H8O3	164.0473	163 (-)	145 (-), 119 (-), 93 (-)

shows an example of how to elucidate between 3,4 and 2,5-DHBA in LC-MS/MS analysis. In reversed-phase chromatography, retention is primarily governed by hydrophobic interactions, with more lipophilic compounds eluting later. Based on the calculated physicochemical parameters, 3,4-DHBA shows a slightly higher logD and lower aqueous solubility compared to 2,5-DHBA, indicating greater hydrophobicity and thus a longer retention time (RT). In addition, n 3,4-DHBA, the hydroxyl groups are positioned in the meta and para positions relative to the carboxyl group, which prevents strong intramolecular hydrogen bonding with –COOH. As a result, both hydroxyl groups remain more available for intermolecular interactions with the surrounding mobile phase and for hydrophobic interactions with the stationary phase.

Another representative example is coumaric acid, which occurs as ortho, meta, and para isomers (Table 1). All three share the same deprotonated ion at m/z 163 and fragment mainly by the loss of CO₂, forming ions at m/z 145 and 119. Because their fragmentation behavior is nearly identical, the distinction among isomers relies on physicochemical parameters and chromatographic retention. At pH 2, o-coumaric acid shows a slightly higher logD value (1.831) and lower solubility (−1.675) compared with the meta and para forms. This indicates greater hydrophobicity and explains its longer retention in reversed-phase chromatography. The ortho configuration enables intramolecular hydrogen bonding between the hydroxyl and carboxyl groups, reducing solvent exposure and enhancing hydrophobic interactions with the stationary phase. The meta and para isomers lack such internal stabilization, leading to slightly earlier elution, with the para form often retained just marginally longer than the meta due to more favorable aromatic stacking. Overall, the elution order under acidic reversed-phase conditions follows meta ≈ para < ortho coumaric acid.

Unambiguous identification, therefore, requires adequate chromatographic separation, supported by quantifier/qualifier ion ratios from the transitions.

2.2. Flavonoids

Flavonoids are a group of plant secondary metabolites that can be found in a wide range of natural sources, including fruits, vegetables [18], cereals [19], tea [20], wine [21], etc. They are crucial for the regulation of plant growth, defense against pathogens, UV protection, and pigmentation [22]. Flavonoids exert diverse beneficial effects on human health [23]. However, the specific molecular pathways underlying the activity of individual phenolic compounds have not yet been fully elucidated.

There are six main subclasses of flavonoids: flavones (luteolin, apigenin), flavonols (quercetin, kaempferol, myricetin), flavanones (hesperetin, naringenin), flavanols or catechins (catechin, epicatechin, epigallocatechin gallate), isoflavones (genistein, daidzein), and anthocyanidins (cyanidin, delphinidin, pelargonidin) [22]. Despite the fact that flavonoids are frequently categorized into structural subclasses, their high degree of structural similarity continues to present substantial analytical challenges. A variety of factors contribute to these challenges, such as their frequent occurrence as glycosides rather than aglycones [24] and their susceptibility to processing and extraction conditions [25]. In particular, flavonoids exhibit greater antioxidant activity than phenolic acid, which implies that they are more susceptible to degradation and less stable.

Additionally, in the context of MS analysis, numerous flavonoids possess identical precursor masses and produce overlapping fragments as a result of similar fragmentation reactions, which impedes unambiguous identification and quantification. Their physicochemical properties are also very similar, resulting in minor differences in RTs, particularly when a unified method is devised for both phenolic acids and flavonoids [26]. Flavonoids are generally more strongly retained on chromatographic columns than phenolic acids, resulting in longer RTs. Consequently, the use of advanced separation strategies in conjunction with MS/MS and more meticulous optimization of chromatographic conditions is required.

2.2.1. Structural Classification

Flavonoids are phenolic compounds with a C6–C3–C6 backbone [27], consisting of two benzene rings and a heterocyclic C-ring (Figure 2). Their structural diversity arises from differences in C-ring saturation, hydroxylation patterns, and glycosylation. Hydroxyl groups are commonly located at C-5 and C-7 in the A-ring, while variation occurs mainly in the B-ring, often forming a catechol structure (3′,4′-dihydroxy). Unsaturation between C-2 and C-3, combined with a 4-oxo group, enhances antioxidant activity by promoting electron delocalization. This is well illustrated by the comparison of quercetin [28], which contains a catechol group (3′,4′-dihydroxy) in the B-ring, with myricetin [29], which has an additional hydroxyl substituent (3′,4′,5′-trihydroxy), resulting in an even higher antioxidant capacity.

In addition, glycosylation, particularly at C-3, reduces activity, and aglycones are generally stronger antioxidants. Antioxidant capacity is further diminished by methoxylation at C-3. Conversely, structural features such as ortho-dihydroxy substitution in the B-ring, a 3-OH group, and molecular planarity strongly contribute to radical scavenging and stabilization of phenoxyl radicals [30,31,32].

2.2.2. Flavonoid Fragmentation

Figure 3A illustrates the nomenclature for the fragmentation of phenolic compounds in negative ion mode. The notations _i,jA⁰ and _i,jB⁰ denote fragments that contain intact A- and B-rings, respectively. The i and j denote cleavages in the C-ring. In flavonoids, the retro-Diels–Alder (RDA) reaction is the most significant fragmentation pathway (Figure 3B). This reaction produces diagnostic ions that preserve intact aromatic rings and a portion of the C-ring. Flavones and flavonols generate robust ^1,3A/^1,3B fragments as a result of the presence of a C2=C3 double bond and a 4-oxo group. Conversely, flavanones and flavanols exhibit weaker RDA cleavage and favor neutral losses (H₂O, CO) in the absence of this conjugation. These ions are highly informative for structural elucidation. The shifted attachment of the B-ring in isoflavones results in the production of distinctive ^2,4A/^2,4B fragments. Beyond the RDA pathway, each flavonoid subclass exhibits characteristic fragmentation routes that differ between positive and negative ionization modes.

2.2.3. Flavonoid Glycosides Fragmentation

LC-MSn is a powerful tool for characterizing flavonoid glycosides, enabling determination of glycosylation type (O-, C-, or mixed), sugar units (hexoses, deoxyhexoses, pentoses), sequence and linkage of residues, position of glycosylation, and aglycone identity [33].

Fragments for flavonoid glycosides are usually annotated according to Mabry et al. [34] and Domon & Costello [35]. According to their rules, aglycone fragments are denoted _i,jA₀ and _i,jB₀; glycosidic fragments A_i, B_i, C_i retain the sugar charge; and X_j, Y_j, Z_j ions retain the aglycone. Index 0 indicates the glycosidic bond directly linked to the aglycone. For C,O-glycosides, numbering starts at the C-linked sugar; Y₀/Z₀ ions indicate loss of the C-glycosyl unit, and C–C cleavage gives the characteristic k,lA_i/k,lX_i ions [36]. Sugar type and sequence can be deduced from characteristic fragment ions (A_i, B_i, C_i, X_i, Y_i, Z_i), even if not all are directly visible in spectra [33]. Typically, flavonoid glycosides contain one or two sugars, and diglycosides can differ by attachment site or linkage type. The presence or absence of Z_i ions, together with characteristic fragments for hexoses, deoxyhexoses, and pentoses, aids in distinguishing mono- from diglycosides and in determining sugar distribution. In diglycosides, cleavage of the sugar directly attached to the aglycone may occur prior to the loss of the second sugar, with the latter simultaneously forming a new linkage with the aglycone. This process results in complex MS spectra and complicates structural assignment [33].

In this review, we will not go into detail on fragmentation mechanisms for each sugar type. Instead,

Table 2. Diagnostic fragments of C- and O-glycosides (hexoses, deoxyhexoses, pentoses), adapted from Vukics & Guttman (2010) [33].

Specific Fragment	Hexoses	Deoxyhexoses	Pentoses
	Mass/Da
0,1X	150	134	120
0,2X	120	104	90
0,3X	90	74	60
1,5X	134	120	104
2,3X-2H2O	66	66
0,4X-2H2O	96	80	66
0,2X-H2O	138	122	108
0,2X-2H2O	156	140	126
2,3X-3H2O	84	84
Yi	162	146	132
Xi	180	164	150

summarizes the key diagnostic neutral losses/cross-ring ions used in practice to recognize the common sugar classes (hexose, deoxyhexose, pentose) and to distinguish mono- from di-glycosides (the informative presence/absence of Z_i ions). For simplicity, we will assign the sugar class primarily by total neutral loss (≈162, 146, 132 Da), while noting that full structure confirmation should consider additional X_i/Y_i/Z_i fragments and comparison with the literature spectra. Importantly, the CE heavily influences the MS/MS spectrum: low energies favor sugar-retaining ions, whereas higher energies increasingly yield aglycone-dominant patterns (RDA-type fragments), which can mask sugar fragments and misperceive assignment.

3. MS Analysis

3.1. Targeted Analysis

Targeted analysis of phenolic compounds is most commonly carried out using liquid chromatography coupled with triple quadrupole MS (LC–QQQ). The reliability of this approach strongly depends on establishing a robust chromatographic separation prior to MS detection. Given the broad spectrum of physicochemical properties within the phenolic family, particular attention must be paid to the selection of mobile phases and their pH values.

In most targeted LC–QQQ workflows, mobile phase A consists of acidified water and mobile phase B of an organic solvent such as acetonitrile or methanol. The choice of solvent and pH adjustment depends on the structural features of the analytes under investigation. Since phenolic acids generally have lower pKa values than flavonoids, maintaining the pH below the (lowest) pKa of the compounds ensures that they remain in their neutral form, which improves retention on reversed-phase columns (

Table 3. Example of pKa values of two phenolic acids and flavonoids. Source: [17].

Phenolic Compound	Class	pKa (Lowest)	Molecular Mass/Da
p-hydroxybenzoic acid	Phenolic acid; hydroxybenzoic acid	4.38	138.0317	100% in neutral form when pH below 2
p-coumaric acid	Phenolic Acid; hydroxycinnamic acid	4.20	164.0473
Quercetin	Flavonoid; flavonol	7.58	302.0426	100% in neutral form when pH below 6
Rutin	Glycosylated flavonoid; Quercetin 3-rutinoside	6.37	610.1534	100% in neutral form when pH below 4

).

This is typically achieved by adding 0.1% formic acid to the aqueous phase and/or organic phase, which suppresses ionization of acidic groups, enhances binding of analytes to the hydrophobic C18 stationary phase, and thereby improves peak shape and resolution. Under such chromatographic conditions, phenolic acids elute earlier in the gradient, followed by flavonoids and their more hydrophobic derivatives. Examples of different mobile phases and columns used for quantitative analysis of phenolic compounds are given in

Table 4. Representative chromatographic conditions (mobile phases, column, gradient, and flow rate) for quantitative analysis of phenolic compounds by liquid chromatography–triple quadrupole (LC-QQQ) mass spectrometry across complex natural matrices.

Matrix	Mobile Phase A	Mobile Phase B	Column	Gradient	Flow Rate/mL/min	Ref.
Propolis	0.1% formic acid in Milli-Q water	0.1% formic acid in acetonitrile	Zorbax SB-C18, (2.1 mm × 50 mm I.D, 1.8)	0.00–0.90 min: 1% → 10% B 0.90–3.00 min: 10% → 20% B 3.00–4.50 min: 20% → 25% B 4.50–6.00 min: 25% → 30% B 6.00–7.50 min: 30% B (isocratic) 7.50–9.00 min: 30% → 90% B 9.00–9.30 min: 90% B (isocratic) 9.30–9.60 min: 90% → 10% B 9.60–15.00 min: 10% → 1% B 15.00–17.00 min (re-equilibration)	0.33	[37]
Plants	0.1% formic acid in Milli-Q water	0.1% formic acid in acetonitrile	C18 column (InfinityLab Poroshell 120 EC-C18, 2.1 × 150 mm, 2.7 μm)	0.00–6.00 min: 20% B (isocratic) 6.00–16.00 min: 20% → 80% B 16.00–20.00 min: 80% B (isocratic) 20.00–25.00 min: 80% → 20% B (return to initial) 25.00–30.00 min: 20% B (re-equilibration)	0.40	[38]
Coffee pulp	0.2% formic acid in Milli-Q water	Acetonitrile	C18 reversed-phase Avantor® ACE® Excel® C18-PFP (100 mm × 2.1 mm, 1.7 μm)	0.00–0.30 min: 10% B (isocratic) 0.30–2.40 min: 10% → 15% B 2.40–3.25 min: 15% → 20% B 3.25–3.60 min: 20% B (isocratic) 3.60–6.20 min: 20% → 95% B 6.20–7.00 min: 95% B (isocratic) 7.00–7.50 min: 95% → 10% B 7.50–11.00 min: 10% B (re-equilibration)	0.30	[39]
Ceylon black tea	0.1% formic acid in Milli-Q water	Acetonitrile	InfinityLab Poroshell 120 EC-C18 column (2.1 mm ×150 mm, 1.9 μm)	0.00–1.50 min: 5% B (isocratic) 1.50–11.00 min: 5% → 15% B 11.00–18.00 min: 15% → 35% B 18.00–25.00 min: 35% → 95% B 25.00–27.00 min: 95% B (isocratic) 27.00–27.10 min: 95% → 5% B 27.10–29.00 min: 5% B (re-equilibration) 29.00+ min: column washing and re-equilibration	0.28	[40]
Rosé Wines	0.1% formic acid in Milli-Q water	0.1% formic acid in methanol	reversed-phase Acquity HSS T3 1.8 µm 1.0 × 100 mm	0.00–2.00 min: 1% B (isocratic) 2.00–2.10 min: 1% → 5% B 2.10–8.00 min: 5% → 10% B 8.00–12.00 min: 10% → 28% B 12.00–18.00 min: 28% B (isocratic) 18.00–22.00 min: 28% → 45% B 22.00–23.50 min: 45% → 99% B 23.50–26.50 min: 99% B (isocratic) 26.50–27.00 min: 99% → 1% B 27.00–30.00 min: 1% B (re-equilibration)	0.17	[41]
Acer negundo tree	0.1% formic acid in Milli-Q water	0.1% formic acid in acetonitrile	Waters, BEH, 2.1 mm × 50 mm, 1.7 Microns	0.00–30.00 min: 1% → 50% B 30.00–35.00 min: 50% → 99% B 35.00–39.00 min: 99% B (isocratic) 39.00–40.00 min: 99% → 1% B 40.00–45.00 min: 1% B (re-equilibration)	0.30	[42]
Rocket-Salad Leaves	0.1% formic acid in Milli-Q water	0.1% formic acid in methanol	Synergi Polar–RP C18 (250 mm × 4.6 mm, 4 µm)	0.00–1.00 min: isocratic (initial composition; %B not specified) 1.00–25.00 min: 20% B (isocratic) 25.00–26.00 min: 20% → 85% B (linear ramp) (brief) isocratic hold at 85% B—duration not reported 26.00–32.00 min: 85% → 20% B (linear return)	0.20	[43]
Solid Residues from the Essential Oil Industry	0.1% formic acid in Milli-Q water	Acetonitrile	Poroshell 120 EC-C18 (4.6 × 150 mm, 4 μm)	0.00–5.00 min: 15% → 25% B (linear) 5.00–10.00 min: 25% → 35% B 10.00–28.00 min: 35% → 60% B 28.00–28.01 min: 60% → 15% B (fast return) 28.01–35.00 min: 15% B (isocratic/re-equilibration)	0.50	[44]

. It can be seen from the table, where we listed only a few examples of developed LC methods (exclusively on the LC-QQQ instrument), that they all share similar conditions. Each paper had satisfactory separation of phenolic compounds, so it can be concluded that the basic rule for successful separation is an acidified mobile phase that consists of (A) water and (B) acetonitrile or methanol, a C18 column and gradient elution.

3.1.1. MRM and SRM

Selected Reaction Monitoring (SRM) and its more advanced form, Multiple Reaction Monitoring (MRM), represent the core acquisition modes in QQQ-MS for the targeted analysis of phenolic compounds. Both rely on monitoring specific precursor-to-product ion transitions, which ensures high selectivity and sensitivity. In SRM, a single transition is monitored per analyte, while MRM enables the simultaneous monitoring of multiple transitions within a single chromatographic run, thereby allowing the analysis of complex phenolic mixtures in natural matrices.

One MRM method consists of defining the precursor ion, ionization mode, product ions, and the collision energy for each product ion. In the case of phenols, it is desirable to list more than one fragment per analyte [8]. This selectivity is particularly important because many compounds from different structural classes generate recurring and partially overlapping fragment ions. For example, isomeric phenolic acids 2,5-dihydroxybenzoic acid (gentisic acid) and 3,4-dihydroxybenzoic acid (protocatechuic acid) share the same deprotonated precursor [M–H]⁻ at m/z 153.0; therefore we monitor 153 → 109 (quantifier; CO₂ loss) together with 153 → 91 (qualifier) in ESI (−), using compound-specific collision energies and predefined ion-ratio/retention-time windows to discriminate the isomers. Likewise, flavonoids such as luteolin and kaempferol share the same precursor [M–H]⁻ at m/z 285.0 and both yield a highly abundant m/z 151 product ion; because 151 is not unique, we set 285 → 151 as the quantifier and include diagnostic qualifiers (e.g., luteolin 285 → 133, kaempferol 285 → 117) with optimized collision energies, confirming identity by the quantifier/qualifier ion ratio and RT.

As an extension of conventional MRM workflows, dynamic MRM (dMRM) has emerged as a preferred strategy for large-scale phenolic profiling, time-scheduling transitions within analyte-specific retention windows to preserve sensitivity and throughput. This optimization significantly reduces dwell time competition, thereby maintaining high sensitivity even when monitoring dozens of transitions [37,42,43]. Barrales-Cureño et al. used dynamic MRM to quantify a panel of 67 phenolic compounds in Acer negundo extracts [42]. Each analyte was scheduled within a 2 min retention-time window, which kept the number of concurrent transitions low despite polarity switching (positive/negative). Product ions were optimized per compound (typical CE 5–40 eV, fragmentor 80–170 V, fixed cell-accelerator 7 V), and quantification used quadratic calibration over 0.25–18 µM with R² ≈ 0.99 for nearly all analytes (chrysophanol R² = 0.90). The dMRM scheduling preserved sensitivity and stable cycle times across the broad panel, enabling robust quantification under the reported LC conditions (shown in Table 4).

3.1.2. Precursor and Product Ion Scan

In QQQ-MS, precursor and product ion scanning are essential modes for the characterization of structurally related compounds. In a precursor ion scan, the third quadrupole (Q3) is fixed to monitor a specific fragment ion, while the first quadrupole (Q1) is scanned across a defined m/z range. This allows for the detection of all precursor ions that generate the selected fragment, providing a powerful means of identifying compounds sharing a common structural motif. Conversely, in a product ion scan, Q1 is fixed to transmit a selected precursor ion, while Q3 is scanned to record all fragment ions produced upon collision-induced dissociation. This mode enables detailed structural elucidation of a specific compound by mapping its fragmentation behavior.

These approaches are particularly valuable in phenolic compounds analysis, where groups of compounds often share core structural features and produce characteristic diagnostic fragments. Precursor ion scanning facilitates the discovery of entire classes of compounds by exploiting these shared fragments, while product ion scanning provides confirmatory structural information for individual analytes. Together, they form a complementary strategy for both targeted and untargeted analysis of phenolic compounds in complex matrices.

Li et al. employed precursor ion scanning in a UHPLC–QTrap workflow to enhance phenolic characterization in Danhong injection, enabling sensitive detection of ninety compounds, including forty-six salvianolic acids [45]. This strategy, by filtering precursors via diagnostic fragment ions, improved specificity beyond full-scan MS and exemplifies the utility of precursor ion scans for resolving complex phenolic profiles. In the work by Saftić et al., product-ion and precursor-ion scanning were used to profile hydroxycinnamic-acid derivatives, with emphasis on caffeic- and p-coumaric acid congeners that are known allergens in European (poplar-type) propolis [37]. Product-ion scans of authentic standards established diagnostic fragments—caffeic acid: m/z 178, 179, 135, 134; p-coumaric acid: m/z 163, 119. Subsequent precursor-ion scans enabled screening for all chromatographic features generating these ions, revealing multiple esters and novel derivatives. For example, in a QQQ precursor-ion scan targeted to the caffeate fragment (Q3 fixed at m/z 179, Q1 scanned across the LC dimension), several precursors >179 Da were observed, consistent with esterified caffeic-acid species. Among these, a peak at [M–H]⁻ m/z 283 was detected. Next, a targeted product-ion scan with Q1 = 283 produced a dominant caffeate fragment at m/z 179 together with the CO₂-loss ions at m/z 135/134, i.e., the expected caffeate series, thereby identifying the compound as caffeic acid phenethyl ester (CAPE). This combined use of precursor and product ion scans enabled comprehensive profiling of hydroxycinnamic acid derivatives, many of which are not commercially available as standards but are critical due to their allergenic potential. By leveraging diagnostic MS/MS fragments derived from authentic standards, they established a workflow for the detection and tentative quantification of structurally related compounds. This approach was essential to reveal the high abundance of caffeic and p-coumaric acid derivatives in propolis extracts and provided insight into their role as major contributors to allergenicity.

3.1.3. Neutral Loss

In MS, neutral loss scanning is a way to obtain data in which precursor and product ions are watched for a constant mass difference that corresponds to the loss of a neutral fragment during collision-induced dissociation. Although the neutral species has not been discovered, the approach finds all precursor ions that exhibit the same characteristic neutral loss. In contrast to product ion scanning, where specific fragment ions (charged molecules) are monitored, neutral loss scanning requires the rejection of a defined neutral molecule. Therefore, its applicability strongly depends on the structural characteristics of the analytes, making it essential to clearly define the analytical objective before employing this approach

Neutral loss scanning is particularly useful in the analysis of flavonoid glycosides, as these compounds frequently undergo the loss of sugar moieties, as already described in previous sections. For example, a neutral loss of 162 Da corresponds to the cleavage of a hexose, typically observed in flavonoid-O-glucosides such as quercetin-3-O-glucoside. A neutral loss of 146 Da indicates the detachment of a deoxyhexose, commonly found in quercetin-3-O-rhamnoside. Finally, the 176 Da neutral loss is diagnostic of glucuronidated flavonoids, such as quercetin-3-O-glucuronide. These neutral losses provide highly specific signatures that enable the rapid detection, classification, and tentative structural assignment of flavonoid conjugates within complex natural matrices.

As a concrete example, Goh et. al. combined neutral loss scanning with high-resolution MS/MS to profile flavonoid and limonoid glycosides in citrus tissues [46]. By exploiting characteristic neutral losses of 120, 162 and 308 Da for hexose-, rutinose- and neohesperidose-substituted flavonoids, and 197 Da for limonoid glycosides, the method selectively enriched precursor ions bearing glycosyl moieties. Subsequent high-resolution MS/MS provided accurate mass and diagnostic fragments that resolved closely related sugar substituents and differentiated flavonoid subclasses, ultimately enabling the detection of 19 flavonoid and 6 limonoid glycosides. A further illustration comes from the work of Cheng et al., who combined MCF-7 cell biospecific extraction with neutral loss and diagnostic ion filtering in an HPLC–QTOF–MS/MS framework to uncover antitumor polymethoxylated flavonoids (PMFs) in Citri Reticulatae Pericarpium Viride [47]. The use of neutral loss and diagnostic ion criteria enabled systematic annotation of PMFs following their selective enrichment by tumor cell binding, leading to the identification of sixteen compounds, including minor constituents with potent bioactivity. This study exemplifies how targeted MS scanning modes, when coupled with biologically guided extraction, can accelerate the discovery of bioactive phenolic compounds, providing both mechanistic insight and translational relevance. Lastly, Shang et al. combined two-dimensional LC–MS with a mass defect filter to profile prenylated phenolic compounds in Glycyrrhiza uralensis [48]. By exploiting diagnostic neutral losses (a neutral loss of 42 Da indicated annular-type prenylation, whereas losses of 56 and 69 Da were diagnostic for chain-type prenyls and A- or B-ring substituents), they rapidly annotated 320 compounds, including three novel dihydrostilbenes confirmed by isolation. This demonstrates how neutral loss filters, coupled with advanced separation, can resolve complex prenylation patterns within phenolic subclasses.

Taken together, new applications demonstrate the adaptability of neutral loss techniques in phenolic chemistry. These examples demonstrate how neutral loss scan, long thought to be narrowly focused, is increasingly used in both targeted and untargeted workflows to expedite structural annotation, broaden chemical coverage, and connect phenolic variety to biological function.

3.1.4. Absolute vs. Relative Concentration

The quantification of phenolic compounds in LC–QQQ can be conducted in either absolute concentration or relative abundance.

Absolute quantification is the most rigorous and reproducible method for reporting quantitative data, as it is designed to ascertain the precise concentration of an analyte in a sample. It is typically accomplished by preparing calibration curves with authentic standards across multiple concentration levels, typically using linear or weighted regression models to cover both low and high concentration ranges. In pharmaceutical preparations, external calibration in solvent (typically methanol or aqueous–organic mixtures) is frequently adequate due to the high reproducibility and minimal matrix-derived ion suppression or enhancement. This is particularly true when the matrix composition is relatively straightforward and well-regulated. Conversely, the examination of intricate natural matrices, including plant extracts, fermented products, and bee products, is considerably more difficult. In ESI sources, these matrices contain high concentrations of carbohydrates, proteins, organic acids, or pigments that co-elute with phenolic compounds and significantly affect ionization efficiency. Matrix effects may result in the suppression or enhancement of analyte signals, resulting in deviations from the true concentrations.

In such instances, matrix-matched calibration, which involves the direct preparation of calibration curves in the sample matrix or a closely related blank matrix, yields more precise results by accounting for matrix-induced variability.

The precision of absolute quantification can be further enhanced by employing internal standards (IS) that are isotopically labeled, with deuterated (d-) analogs being the most frequently used. These IS exhibit nearly identical behavior to their native counterparts during sample preparation, chromatographic separation, and ionization, thereby compensating for sample-to-sample variation in ionization and recovery. They may be employed as class surrogates or applied individually to each analyte, such as using caffeic acid-d3 as a reference for a group of caffeic acid derivatives. Nevertheless, chemical modifications, such as methylation, glycosylation, or acylation, can significantly affect the polarity, solubility, stability, and ionization efficacy of structurally related groups. Extrapolating quantitative data from a single isotopically labeled standard to structurally diverse derivatives necessitates meticulous consideration of these distinctions.

Still, there are rare studies employing this method of quantification of phenols from complex matrices. For example, Irakli et al. used salicylic acid as the IS and added it at a fixed concentration to all calibration standards and sample extracts immediately prior to injection [44]. This approach is advantageous because a constant IS level compensates for injection-volume variability, short-term source fluctuations, and moderate matrix-induced ionization changes, thereby improving precision and run-to-run reproducibility of the LC–QQQ data. However, a single, non-isotopically labeled IS that is not closely matched to each phenolic class has limitations: when spiked post-extraction, it does not correct for differential extraction recoveries; it may not track class-specific matrix effects or polarity-switching behavior; potential endogenous salicylic acid in botanical matrices can introduce background; and co-elution or differing retention relative to late-eluting targets can reduce normalization accuracy. For these reasons, while the IS improves robustness, it is not ideal; superior accuracy would be obtained with isotopically labeled, compound-specific (or class-matched) internal standards, preferably spiked pre-extraction to correct both recovery and ionization effects.

Relative quantification, on the other hand, is based on comparing signal intensities, typically obtained from extracted ion chromatograms (EIC) or from the total ion current (TIC), without conversion into absolute concentrations. Instead of reporting the exact amount of a compound in the sample, this approach focuses on relative abundance, enabling comparative profiling of samples. Relative quantification is particularly valuable when the primary aim is to evaluate differences in phenolic composition between groups of samples within a single study, for example, when comparing botanical origins, seasonal variations, or processing effects. However, the main limitation of relative quantification lies in its reduced reliability for inter-study comparisons. Signal intensities are highly dependent on instrumental parameters, ionization efficiency, and extraction procedures, all of which may differ considerably between laboratories or even between runs on the same instrument. Variations in ion source conditions, collision energies, and chromatographic resolution further complicate direct comparison. Nevertheless, relative concentration approaches can also be extended to semi-quantification by normalizing unknown metabolites against a common precursor or structurally related standard. This allows estimation of their levels even in the absence of specific calibration curves. A well-recognized example is extra virgin olive oil (EVOO) analysis, where many secoiridoid derivatives lack commercially available calibrants. In such cases, oleuropein is frequently used as a reference standard due to its structural similarity. For instance, Peršurić and Saftić used this approach for the quantification of several oleuropein derivatives in EVOO, including 3,4-dihydroxyphenylethanol-elenolic acid (3,4-DHPEA-EA), p-hydroxyphenylethanol-elenolic acid (p-HPEA-EA), etc. [49]. These compounds were quantified semi-quantitatively using the calibration curve of oleuropein, thereby converting all phenolic derivatives into oleuropein equivalents. This methodological adjustment enabled the estimation of metabolite levels despite the absence of authentic standards, ensuring consistent comparative profiling within the study. In addition, Saftić et al. used the same approach to caffeic acid derivatives in propolis, identified on the basis of caffeic acid, with recalculation using a molecular-weight correction factor (Molecular weight of derivative/Molecular weight of standard) [37]. However, this approach assumes equivalent ionization/fragmentation efficiencies and neglects matrix and isomer-specific effects, so results remain semi-quantitative until verified with authentic standards.

One valuable compromise between absolute and semi-relative approaches is the standard addition method (Figure 4). This technique involves spiking known amounts of standard directly into the sample matrix and constructing a calibration curve by plotting the measured signal against the added concentration. Extrapolation of the calibration curve to the x-axis then reveals the native concentration of the analyte in the unspiked sample.

Because the calibration is performed within the actual matrix, standard addition directly accounts for matrix effects such as ion suppression or enhancement, making it particularly useful in highly complex samples where co-eluting constituents cannot be fully eliminated. In two studies, this method was successfully applied for the analysis of phenolic compounds in honey, a matrix dominated by high sugar content that strongly affects ionization efficiency in electrospray ionization [50,51]. By adding standards directly into the honey samples, the authors minimized the impact of ion suppression and obtained more reliable concentration estimates compared to external calibration in solvent. Although the use of isotopically labeled standards represents the most accurate approach to absolute quantification, the lack of such standards made it essential to adopt an alternative method. Furthermore, experiments with solid-phase extraction (SPE), aimed at reducing sugar interference, revealed substantial analyte losses. When eluting with acidified water, a significant fraction of phenolic acids was lost, with recovery for some analytes falling below 20% [50,51]. These findings highlighted the risk of severe underestimation if conventional SPE-based clean-up was applied. For this reason, the standard addition method performed directly in the untreated honey matrix provided the most suitable compromise: while inherently labor-intensive, it corrected for matrix effects without introducing further analyte loss. Thus, although complex and time-consuming, this approach ensured a higher degree of accuracy and confidence in quantifying phenolic compounds in honey than either external calibration or extraction-based methods.

3.2. Untargeted Analysis

The objective of untargeted MS analysis of phenolic compounds is to capture the broadest possible spectrum of molecular features, ranging from simple phenylpropanoids to flavonoid glycosides, without prior specification of analytes. In practice, this involves workflows with and without chromatographic separation. Chromatography-based platforms offer RT as an additional dimension for structural resolution, improved MS/MS quality, and more reliable isomer discrimination. A systematic comparison of Quadrupole Time-of-Flight MS (QTOF) and Orbitrap instruments in plant metabolomics showed that both platforms offer comparable sensitivity and mass accuracy, with QTOF providing slightly better spectral accuracy, while Orbitrap offered lower detection limits for selected compounds [52].

Methodologically, both platforms deliver high-resolution, accurate-mass data but optimize different trade-offs. QTOF couples flexible quadrupole isolation with very fast MS/MS, which makes it well suited for High-resolution mass spectrometry (HRMS) data-independent acquisition (DIA) surveys and broad untargeted discovery of phenolic compounds in complex matrices [53]. Orbitrap offers higher resolving power and sub-ppm mass accuracy, sharpening isobaric discrimination and stabilizing long-gradient identifications, particularly useful for closely related glycosylated flavonoids [54]. The literature on plant and food phenolics predominantly reports the use of LC–QTOF for untargeted HRMS–DIA profiling, with LC–Orbitrap increasingly adopted when maximal resolving power or parallel reaction monitoring confirmation is required.

Non-separated strategies such as direct injection (DI) emphasize speed and throughput, generating high-dimensional fingerprints for chemometric classification at the expense of compound-level confidence. Ion mobility mass spectrometry (IM-MS) is increasingly integrated into both approaches, adding collisional cross section (CCS) as an orthogonal descriptor that separates isomeric and conformational variants in the gas phase and strengthens annotation confidence. Regardless of modality, untargeted outputs must be processed into aligned feature tables, followed by normalization, batch correction, and statistical modeling to enable class discrimination and marker discovery. Compound annotation is followed through accurate mass and isotopic pattern, diagnostic neutral losses, spectral library and in silico MS/MS matching, as well as network-based contextualization. The following sections examine direct-injection and LC–MS/MS modalities, their trade-offs, and the role of ion mobility in bridging throughput and structural resolution.

3.2.1. Acquisition Modes and Reproducibility

Recent benchmarking on an Orbitrap platform elucidated the trade-offs among data-dependent (DDA), data-independent (DIA), and Thermo’s AcquireX workflows [55]. In a complex lipid matrix enriched with low-abundance eicosanoids, DIA detected the most features (≈1036 on average across weeks) and showed the highest week-to-week consistency (pairwise overlap up to 61%), owing to its fixed, highly reproducible MS² windowing [55]. In contrast, the proportion of DIA features that obtained confident MS² identifications was constrained, not by the instrument itself, but by downstream software limitations (such as deconvolution and library coverage) [55].

In contrast, DDA generated consistently high-quality spectra but experienced stochastic MS² triggering, resulting in inconsistent re-fragmentation of low-intensity precursors across different days. This variability led to inconsistent identifications despite the use of identical samples and settings [55,56]. AcquireX enhanced MS² coverage in pooled identification runs. However, performance was significantly influenced by retention-time stability and intensity thresholds. At trace levels, numerous precursors did not activate MS² despite multiple iterations [55,57]. This reliance on accurate RT alignment and list management is a recognized limitation of list-driven DDA solutions [57].

Sensitivity analyses revealed a general cut-off around 0.1–0.01 ng mL⁻¹ for all three modes; none reliably detected eicosanoids at physiological levels, underscoring that “not detected” in untargeted surveys rarely implies absence [55]. Methodological practices underline that matrix effects, signal drift, and unequal chromatographic complexity might amplify disparities across acquisition modes if appropriate quality control is not implemented. The impacts are mitigated when system suitability testing (SST) and pooled quality control samples are used throughout the sequence, along with suitable signal normalization. Such measures ensure that observed variations between DDA, DIA and AcquireX reflect true analytical performance rather than artifacts of the measurement process [58,59,60].

Implications for Phenolic Chemistries

The selection of acquisition mode significantly influences both coverage and repeatability in the analysis of flavonoids and phenolic acids. DIA offers the most comprehensive representation of these compound classes, capturing consistent MS² fragmentation patterns even among complex glycosylated flavonoid series, contingent upon the utilization of advanced deconvolution and spectral-matching algorithms validated for DIA data [61,62,63]. Such approaches are particularly advantageous for abundant subclasses such as flavonol and flavone glycosides, where consistent fragmentation of diagnostic neutral losses enables confident substructure annotation across runs.

When precise structural confirmation of individual phenolics is required, DDA offers clear advantages because it isolates a single precursor ion in a narrow mass window and produces clean, compound-specific MS² spectra [56,64]. This level of specificity is particularly valuable for isomeric phenolic acids, such as the various caffeoylquinic or ferulic acid isomers, whose diagnostic fragment ions and neutral losses differ only subtly [65,66]. This method incorporates stochasticity, implying that low-abundance or co-eluting chemicals may not consistently be chosen for fragmentation, leading to variable MS² coverage among replicate runs. To address this limitation, DDA methods can be enhanced using planned inclusion lists that specify m/z and retention-time periods for target molecules, assuring constant fragmentation at the chromatographic apex. Other enhancements include using replicate-based acquisition logic, which prioritizes ions missed in earlier runs, and using stepped collision energies to generate richer fragmentation patterns [64,65]. Together, these solutions preserve DDA’s structural precision while decreasing its random sampling behavior, resulting in a reliable method for validating the identities of closely related flavonoid and phenolic acid isomers.

Several issues persist across all modalities, including structural isomer coelution, ambiguity caused by adduct and neutral-loss formation, and a lack of reference spectra for phenolic compounds [65,66]. These variables commonly impede the accurate annotation of closely related phenolic acids and flavonoids in untargeted datasets. To address this, current best practices emphasize the integration of orthogonal analytical dimensions, such as retention-time constraints, in silico MS² prediction, and ion-mobility separation, with targeted confirmation using authentic standards.

3.2.2. Direct Injection (DI) MS

Without chromatographic separation (direct infusion or flow-injection), phenolic extracts are commonly interrogated on high-resolution MS platforms such as Electrospray Ionization–Time-of-Flight MS (ESI-TOF) or Electrospray Ionization–Quadrupole Time-of-Flight MS (ESI-QTOF), and less frequently by Matrix-Assisted Laser Desorption/Ionization–Time-of-Flight (MALDI-TOF). These approaches provide rapid, high-mass-accuracy MS or MS/MS fingerprints, but the simultaneous introduction of all analytes inevitably leads to co-ionization, ion suppression, heterogeneous adduct formation, and overlapping isobaric or isomeric signals. Such effects restrict confident compound-level identification and robust quantification. Consequently, DI MS is often more suited to comparative chemometric phenotyping of complex samples than to comprehensive annotation of individual metabolites.

Saftić et al. [26] analyzed identical propolis extracts by three orthogonal MS workflows: LC–QQQ for targeted quantification, Liquid Chromatography–Quadrupole Time-of-Flight MS (LC–QTOF) for untargeted profiling, and DI ESI–QTOF for rapid classification. DI–QTOF achieved accurate geographic discrimination in ~2 min acquisitions (processing window 0.1–2.0 min; PLS-DA: RMSE = 0.0345, R² = 0.9947, cross validation: RMSE = 0.1455, R² = 0.9099), highlighting its strength as a high-throughput screening tool that circumvents both chromatographic separation and compound identification. However, this gain in speed comes at the cost of resolution: isomeric and co-eluting features collapse into single signals, matrix effects are amplified, and minor between-sample differences can be masked. Indeed, Peršurić et al. used DI–QTOF to profile pomegranate peel phenolic compounds and successfully identified major constituents such as punicalagin I–III, yet compounds present at lower abundance escaped detection [67].

Some of these limitations can be mitigated by coupling direct injection to ion mobility separation. DI–IM-MS has emerged as a particularly effective strategy for structural analysis of phenolic compounds, offering a structurally informative alternative to LC-based workflows. In anthocyanin-rich extracts of Brassica oleracea, DI–IM-MS resolved even isobaric species that notoriously co-elute in LC, while drift-time-aligned fragmentation provided precursor, MS/MS, and CCS identifiers within ~1.5 min, at low-picomole sensitivity [68]. Anthocyanins were organized into discrete drift-time bands according to their degree of glycosylation and acylation, allowing both rapid pattern recognition and confident identity confirmation via m/z + CCS without the need for chromatographic optimization. A related study demonstrated the technique’s utility for authenticating South African herbal teas: DI–IM-MS distinguished rooibos, honeybush, and their blends by detecting diagnostic phenolic markers such as aspalathin, orientin, and mangiferin, thus enabling rapid adulteration screening in the absence of lengthy separations [69]. Together, these examples illustrate how direct-injection MS, particularly when augmented by ion mobility, can provide high-throughput and structurally meaningful phenolic profiling while acknowledging the trade-offs in sensitivity and resolution.

3.2.3. Untargeted LC-MS/MS

In untargeted high-resolution MS/MS, the addition of a chromatographic dimension mitigates matrix effects, improves resolution of isomers, and expands metabolome coverage beyond direct injection. Accurate mass determination, combined with automated MS/MS acquisition and comparison against spectral libraries, enables annotation, although structurally related phenolic compounds often display highly similar fragmentation patterns that require orthogonal evidence such as retention behavior, predicted retention indices, or ion-mobility-derived collision cross section (CCS) values for confident identification. In this section, we review the techniques applied for detailed characterization of phenolic compounds in plant matrices. The most recent examples are summarized in Table 5.

As summarized in Table 5, recent LC–QTOF and LC–Orbitrap studies on plant phenolics increasingly combine comprehensive chemical profiling with functional bioassays, linking metabolite identity to biological activity rather than treating annotation as an isolated task [70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86]. Methodologically, however, the level of detail in compound identification remains heterogeneous. Several works, such as Tan et al. (2024) [70] and Boubertakh et al. (2025) [71], reported accurate mass (<5 ppm) and diagnostic MS² fragments verified by authentic standards, approaches consistent with MSI level 1 identification. Others relied on database dereplication (METLIN, HMDB, GNPS) with mass-error windows sometimes up to 100 ppm [72], a tolerance that substantially weakens confidence and often blurs structural isomers. Only a minority of papers specified spectral-matching scores or similarity thresholds, and very few provided retention-time or isotopic-pattern validation.

In contrast, Orbitrap-based workflows [73,74,75] showed higher spectral resolution and more explicit use of curated libraries (mzCloud, mzVault), achieving narrower mass tolerances (<3–5 ppm) and consistent annotation of multi-class phenols. Standardized reporting templates are necessary for untargeted phytochemical studies, including experimental and theoretical m/z with mass error (ppm), retention period, diagnostic MS² ions, similarity score, and MSI confidence level. Such harmonization would bring phenolic compound metabolomics in line with broader metabolite identification best practices, allowing for future meta-analyses that link chemical variety to bioactivity trends.

Table 5. Overview of liquid-chromatography-quadrupole-time-of-flight (LC-QTOF) based methods for phenolic-compound identification in plant matrices published within the last year (2025).

Matrix	Analyzed Compounds	Identification	Purpose	Ref.
LC-QTOF
Plukenetia volubilis leaves	16 compounds (Kaempferol trihexoside, Kaempferol-3-O- glucoside (astragalin), etc.)	Accurate mass measurements, fragmentation patterns, UV spectra, and comparison with reference standards and literature data	To characterize the phytochemical composition of Plukenetia volubilis leaves extract and evaluate its anti-Helicobacter pylori activity through in vitro assays and in silico docking, with focus on flavonoids such as astragalin.	[70]
Schima argentea	30 compounds (Quercetin 3-arabinoside, Nodularin, Ricinoleic acid methyl ester, etc.)	Compound identification was performed using Progenesis QI’s online METLIN database and a custom library from Biomarker Technologies Co., Ltd., with theoretical fragment ions considered. The mass error was controlled within 100 ppm	To investigate the antioxidant and photoprotective effects of 3,4-dihydroxybenzoic acid and (+)-catechin, identified from Schima argentea extract, in UVB-irradiated HaCaT keratinocytes.	[72]
Stem, Roots, and Leaves of Syzygium cumini	12 compounds (phenolic compounds and lignanas, for example: 3,4-O-Dimethylgallic acid, Scutellarein, etc.)	Compounds assigned by retention time (RT), isotopic pattern, and database/software matching (Agilent MassHunter)	To profile and compare phenolic constituents in stem, root, and leaf of Syzygium cumini and assess their antioxidant capacity across multiple assays, linking phenolic content to activity.	[76]
Symphorema polyandrum	11 compounds (Aciculatinone (O-Methylated flavonoids, 2″-p-Coumaroylvitexin (Flavonoid glycosides), etc.)	Scientific literature and mass-spectra databases such as the METLIN database, Chemspider, Pub-chem, NIST MS–MS database, and npatlas database	Characterization and in vitro assessment of its antioxidant, anticancer, and anti-inflammatory potential	[77]
Litsea monopetala bark	9 compounds (fraxetin, Kaempferol-3-O-alpha-L-rhamnoside, Kaempferol-3- neohesperidoside, etc.)	MassBank Europe Mass Spectral Database, the Human Metabolome Database, and relevant literature sources	Metabolic profiling and demonstration of hepatoprotective potential, supporting traditional use against jaundice and liver disorders	[78]
Gliricidia sepium leaves	22 bioactive compounds (flavonoids, phenolic acids, triterpenoid saponins, fatty acid derivatives, and coumarins)	Molecular formula prediction and peak identification via dereplication using ChemCalc online, Dictionary of Natural Products (DNP) Database, Global Natural Product Social Molecular Networking (GNPS) Database, Human Metabolome Database (HMDB), MassBank of North America (MoNA) Database, MassBank Europe	Metabolomic profiling with in vivo renoprotective assessment in diabetic hamsters, supporting use against diabetic nephropathy.	[79]
Achillea ligustica	45 phenolic compounds(caffeoylquinic and dicaffeoylquinic acid isomers, dihydroxybenzoic acid derivatives, coumarins, flavones, flavonols, lignin)	Accurate mass, MS/MS fragments and standards	Phytochemical profiling with evaluation of antioxidant activity (DPPH, ABTS, phenanthroline, reducing power), antimicrobial effects, acute toxicity, and analgesic activity, supporting traditional uses.	[71]
Strobilanthes sarcorrhiza root	55 compounds (terpenoids, phenylethanol glycosides, fatty acid derivatives, chain/other glycosides, flavonoid glycosides, sterols, alkaloids, nucleosides, esters, alkylene oxides, organic acids; 34 first reported in plant	Accurate mass, MS/MS fragments, RT; differences assessed by chemometrics	Chemical profiling and part-differentiation showing phenylethanol glycosides enriched in underground parts and terpenoids in aboveground parts, supporting rhizomes as an alternative medicinal part to improve resource utilization	[80]
Ziziphus budhensis Leaves	46 compounds (phenolic compounds, benzyl-isoquinolinic alkaloids, cyclopeptide alkaloids, triterpene aglycone and saponins)	Confirmation by diagnostic MS fragments and reference standards	Phytochemical profiling with evaluation of antioxidant, antibacterial, antifungal, cytotoxicity, and acute oral toxicity in mice to assess medicinal potential.	[81]
Lemon, lime, orange, and grapefruit juices	57 compounds ranging from polar phenolic acids over flavonoid glycosides to a polar coumarins, psoralens, and polymethoxyflavones in lemon, lime, orange, and grapefruit juices	Assignments confirmed using UV spectra, accurate mass, diagnostic MS fragments and authentic standards	Phytochemical profiling and quantitation of phenolic compounds/PMFs/coumarins/psoralens for authentication of Citrus juices and identification of species-specific chemical markers	[82]
Fagonia arabica	42 phenolic compounds (3 phenolic acids (cinnamic acid derivatives), 15 flavonols, 1 flavanol, 4 flavanones, 8 flavones, 2 isoflavones, 1 chalcone, 1 aurone O-glycosides, 1 stilbene and 6 anthocyanins)	ReSpect databases	Profiling and evaluation of cholinesterase inhibition potential through in vitro and in silico approaches	[83]
LC-Orbitrap
Shenhua Tablets	183 compounds (64 flavonoids, 52 terpenoids, 37 organic acids, 6 phenylpropanoids, 5 phenols, and 19 other phytochemicals)	mzCloud and mzVault library	Characterization	[73]
Lagopsis supina	114 compounds	opensource software, including GNPS web tools and MS-Dial, alongside public mass spectrometry databases (GNPS, HMDB, LipidMaps, KNApSAcK, and the American Mass Bank)	Chemical Composition and Antioxidant, Adipogenic, and Ani-Inflammatory activities	[84]
Rubi fructus	47 components (10 organic acids, 15 flavonoids, 12 phenols, 2 alkaloids, 4 terpenoids, 1 miscellaneous compound, 1 stilbene, 1 steroid and its derivatives, and 1 diterpenoid)	databases and relevant literature	Phytochemical profiling and evaluation of anti-diabetic mechanism through network pharmacology and experimental validation	[74]
Ribes nigrum leaf	24 compounds	Compound Discoverer (v. 2.1, Thermo, Waltham, MA, USA): accurate mass and mass fragmentation pattern spectra against customized database of different classes of phytochemicals created on the basis of literature data	Characterization and antioxidant and anti-inflammatory capabilities, concentrating on the influence of oxidative stress, gene expression, and enzymatic activity in microglial cells.	[75]
Thymbra spicata L.	31 compounds	calibration curves, RT and MS/MS fragmentation patterns (Quan Peak and confirming ions)	Phytochemical profiling and anti-inflammatory mechanism	[85]

Table 5. Overview of liquid-chromatography-quadrupole-time-of-flight (LC-QTOF) based methods for phenolic-compound identification in plant matrices published within the last year (2025).

Matrix	Analyzed Compounds	Identification	Purpose	Ref.
LC-QTOF
Plukenetia volubilis leaves	16 compounds (Kaempferol trihexoside, Kaempferol-3-O- glucoside (astragalin), etc.)	Accurate mass measurements, fragmentation patterns, UV spectra, and comparison with reference standards and literature data	To characterize the phytochemical composition of Plukenetia volubilis leaves extract and evaluate its anti-Helicobacter pylori activity through in vitro assays and in silico docking, with focus on flavonoids such as astragalin.	[70]
Schima argentea	30 compounds (Quercetin 3-arabinoside, Nodularin, Ricinoleic acid methyl ester, etc.)	Compound identification was performed using Progenesis QI’s online METLIN database and a custom library from Biomarker Technologies Co., Ltd., with theoretical fragment ions considered. The mass error was controlled within 100 ppm	To investigate the antioxidant and photoprotective effects of 3,4-dihydroxybenzoic acid and (+)-catechin, identified from Schima argentea extract, in UVB-irradiated HaCaT keratinocytes.	[72]
Stem, Roots, and Leaves of Syzygium cumini	12 compounds (phenolic compounds and lignanas, for example: 3,4-O-Dimethylgallic acid, Scutellarein, etc.)	Compounds assigned by retention time (RT), isotopic pattern, and database/software matching (Agilent MassHunter)	To profile and compare phenolic constituents in stem, root, and leaf of Syzygium cumini and assess their antioxidant capacity across multiple assays, linking phenolic content to activity.	[76]
Symphorema polyandrum	11 compounds (Aciculatinone (O-Methylated flavonoids, 2″-p-Coumaroylvitexin (Flavonoid glycosides), etc.)	Scientific literature and mass-spectra databases such as the METLIN database, Chemspider, Pub-chem, NIST MS–MS database, and npatlas database	Characterization and in vitro assessment of its antioxidant, anticancer, and anti-inflammatory potential	[77]
Litsea monopetala bark	9 compounds (fraxetin, Kaempferol-3-O-alpha-L-rhamnoside, Kaempferol-3- neohesperidoside, etc.)	MassBank Europe Mass Spectral Database, the Human Metabolome Database, and relevant literature sources	Metabolic profiling and demonstration of hepatoprotective potential, supporting traditional use against jaundice and liver disorders	[78]
Gliricidia sepium leaves	22 bioactive compounds (flavonoids, phenolic acids, triterpenoid saponins, fatty acid derivatives, and coumarins)	Molecular formula prediction and peak identification via dereplication using ChemCalc online, Dictionary of Natural Products (DNP) Database, Global Natural Product Social Molecular Networking (GNPS) Database, Human Metabolome Database (HMDB), MassBank of North America (MoNA) Database, MassBank Europe	Metabolomic profiling with in vivo renoprotective assessment in diabetic hamsters, supporting use against diabetic nephropathy.	[79]
Achillea ligustica	45 phenolic compounds(caffeoylquinic and dicaffeoylquinic acid isomers, dihydroxybenzoic acid derivatives, coumarins, flavones, flavonols, lignin)	Accurate mass, MS/MS fragments and standards	Phytochemical profiling with evaluation of antioxidant activity (DPPH, ABTS, phenanthroline, reducing power), antimicrobial effects, acute toxicity, and analgesic activity, supporting traditional uses.	[71]
Strobilanthes sarcorrhiza root	55 compounds (terpenoids, phenylethanol glycosides, fatty acid derivatives, chain/other glycosides, flavonoid glycosides, sterols, alkaloids, nucleosides, esters, alkylene oxides, organic acids; 34 first reported in plant	Accurate mass, MS/MS fragments, RT; differences assessed by chemometrics	Chemical profiling and part-differentiation showing phenylethanol glycosides enriched in underground parts and terpenoids in aboveground parts, supporting rhizomes as an alternative medicinal part to improve resource utilization	[80]
Ziziphus budhensis Leaves	46 compounds (phenolic compounds, benzyl-isoquinolinic alkaloids, cyclopeptide alkaloids, triterpene aglycone and saponins)	Confirmation by diagnostic MS fragments and reference standards	Phytochemical profiling with evaluation of antioxidant, antibacterial, antifungal, cytotoxicity, and acute oral toxicity in mice to assess medicinal potential.	[81]
Lemon, lime, orange, and grapefruit juices	57 compounds ranging from polar phenolic acids over flavonoid glycosides to a polar coumarins, psoralens, and polymethoxyflavones in lemon, lime, orange, and grapefruit juices	Assignments confirmed using UV spectra, accurate mass, diagnostic MS fragments and authentic standards	Phytochemical profiling and quantitation of phenolic compounds/PMFs/coumarins/psoralens for authentication of Citrus juices and identification of species-specific chemical markers	[82]
Fagonia arabica	42 phenolic compounds (3 phenolic acids (cinnamic acid derivatives), 15 flavonols, 1 flavanol, 4 flavanones, 8 flavones, 2 isoflavones, 1 chalcone, 1 aurone O-glycosides, 1 stilbene and 6 anthocyanins)	ReSpect databases	Profiling and evaluation of cholinesterase inhibition potential through in vitro and in silico approaches	[83]
LC-Orbitrap
Shenhua Tablets	183 compounds (64 flavonoids, 52 terpenoids, 37 organic acids, 6 phenylpropanoids, 5 phenols, and 19 other phytochemicals)	mzCloud and mzVault library	Characterization	[73]
Lagopsis supina	114 compounds	opensource software, including GNPS web tools and MS-Dial, alongside public mass spectrometry databases (GNPS, HMDB, LipidMaps, KNApSAcK, and the American Mass Bank)	Chemical Composition and Antioxidant, Adipogenic, and Ani-Inflammatory activities	[84]
Rubi fructus	47 components (10 organic acids, 15 flavonoids, 12 phenols, 2 alkaloids, 4 terpenoids, 1 miscellaneous compound, 1 stilbene, 1 steroid and its derivatives, and 1 diterpenoid)	databases and relevant literature	Phytochemical profiling and evaluation of anti-diabetic mechanism through network pharmacology and experimental validation	[74]
Ribes nigrum leaf	24 compounds	Compound Discoverer (v. 2.1, Thermo, Waltham, MA, USA): accurate mass and mass fragmentation pattern spectra against customized database of different classes of phytochemicals created on the basis of literature data	Characterization and antioxidant and anti-inflammatory capabilities, concentrating on the influence of oxidative stress, gene expression, and enzymatic activity in microglial cells.	[75]
Thymbra spicata L.	31 compounds	calibration curves, RT and MS/MS fragmentation patterns (Quan Peak and confirming ions)	Phytochemical profiling and anti-inflammatory mechanism	[85]

Identification

Despite advances in high-resolution LC–QTOF and LC–Orbitrap platforms, compound identification remains the most critical bottleneck. Vendor-specific libraries (MassHunter, mzCloud) and open-access repositories (METLIN, HMDB, MassBank) provide valuable reference spectra, yet most databases remain heavily biased towards precursor masses rather than fully annotated MS/MS fragmentation patterns. This gap forces researchers into extensive manual curation, comparing experimental spectra one by one against published data, a task particularly demanding for glycosylated flavonoids and structurally diverse phenolic acids.

Accurate mass matching and spectral similarity scoring are key parameters that determine the reliability of compound identification in untargeted LC–HRMS. For high-resolution instruments such as Orbitrap and QTOF, mass accuracy thresholds of ±5 ppm are typically considered sufficient for reliable annotation, whereas values up to ±10 ppm may be acceptable when signal intensity is low or matrix effects are significant [66]. Reporting the precise mass deviation guarantees that each identification can be assessed independently, as even minor variations in m/z affect molecular formula determination and library matching results. In addition to accurate mass, most data-processing workflows assign a matching or similarity score, typically based on the cosine similarity, dot product, or entropy score, when comparing experimental MS² spectra to reference or in silico libraries. Higher scores (usually >0.7–0.9, depending on the algorithm) indicate better agreement between fragment ions, supporting higher confidence levels of identification [62,65]. Since scoring systems and tolerances differ among software tools, it is crucial that studies explicitly specify the ppm tolerance employed, the scoring algorithm used, and the acceptance threshold applied. A comprehensive identification table should therefore encompass: (i) the experimental and theoretical m/z values accompanied by the corresponding mass error (ppm), (ii) the RT, (iii) the fragment ions employed for matching, (iv) the matching score or similarity index, and (v) the assigned identification level (1–4). Such transparency ensures reproducibility, promotes comparison across laboratories, and permits future re-assessment of identifications as spectral libraries expand.

Untargeted identifications employ a stratified approach that incorporates orthogonal evidence [66,86]. Level 1 (confirmed) necessitates an authentic standard run conducted under identical LC–MS conditions, with matching accurate mass (sub-ppm), isotopic pattern, RT, and MS/MS spectra (

Table 6. Confidence levels for metabolite identification.

Level	Description	Minimum Evidence to Claim the Level	What You May Report
1—Confirmed identification	Confirmed structure	In-house authentic standard measured in the same method with matching retention time (RT) window, exact mass, isotope pattern, and MS/MS (key fragments and ratios).	Definitive identity and quantitative data.
2—putatively annotated compounds	Library/in silico match with orthogonal support	High-quality library or in silico MS/MS match. No in-house standard.	Probable identity (report as “putatively annotated”).
3—Putative compound class	Substructure/class only	Diagnostic fragments/neutral losses define a class, but isomeric structures are unresolved; MS/MS present but not unique to a single structure.	Class-level assignment only.
4—Unknown feature	Reproducible signal	Reproducible, alignable feature (m/z–RT; acceptable mass accuracy; clean peak shape); no reliable structural evidence.	Report as feature ID (m/z, RT) for statistics; do not name a compound.

). Level 2 (putatively annotated compounds) depends on a high-quality library or in silico MS/MS matches complemented by orthogonal agreement in RT and/or CCS (within instrument-specific tolerances). Level 3 (putative class/substructure) assigns compound class based on diagnostic fragments, neutral losses, and adduct/isotope logic when isomers cannot be definitively distinguished. Level 4 (unknown feature) retains aligned features for statistical analysis without making structural claims.

A central limitation is the incomplete reproducibility of fragmentation patterns across platforms and ionization modes. Positive and negative ion acquisition frequently yield distinct diagnostic ions, while adjustments in fragmentor voltage and CE strongly influence spectral outcomes. Importantly, fragmentation conditions are not standardized across studies, which complicates the direct comparison of metabolomic datasets. As summarized in

Table 7. Fragmentor voltages, collision energy settings, and ESI acquisition modes reported in untargeted metabolomics studies listed in Table 5.

Matrix	Mode	Fragmentor/V	Collision Energies/V	Ref.
Plukenetia volubilis leaves	negative	110	10 and 20	[70]
Schima argentea	positive and negative	NP *	NP	[72]
Stem, Roots, and Leaves of Syzygium cumini	positive and negative	NP *	NP	[76]
Symphorema polyandrum	positive and negative	NP	10–40 (ramp) (for positive mode) and 10–30 (ramp) (for negative mode)	[77]
Litsea monopetala bark	positive	NP	20	[78]
Achillea ligustica	NP	NP	NP	[71]
Ziziphus budhensis Leaves	positive	NP	30	[81]
Lemon, lime, orange, and grapefruit juices	positive and negative	NP	25–50 eV (stepping, negative and positive ion mode) and 40 eV (positive)	[82]
Fagonia arabica	positive and negative	NP	NP	[83]

* NP—not provided in the original manuscript.

, recent works employing untargeted metabolomics (listed in Table 5) applied different fragmentor voltages and CE regimes, as well as variable ESI modes, underscoring how instrumental settings critically shape spectral quality and compound annotation. Not all authors explicitly reported these instrumental parameters; while fragmentor voltage changes (e.g., from 100 to 150 V) do not drastically affect spectral profiles, variations in collision energy have a pronounced impact on MS/MS outcomes.

Consequently, while cheminformatics-driven dereplication has accelerated workflows, the field continues to rely on expert-guided spectrum validation. Current practice shows that phytochemical profiling is rarely pursued as an endpoint; rather, it establishes the molecular framework from which specific metabolites can be mechanistically linked to antioxidant, anti-inflammatory, hepatoprotective, or other biological effects. This methodological convergence underscores a broader trend: in untargeted phenolic research, chemical annotation is indispensable, but its ultimate purpose lies in bridging molecular structure with functional biology (Table 5).

3.2.4. Statistics of Untargeted MS Data

Untargeted metabolomics often utilizes multivariate statistical analyses to distinguish samples, identify clustering patterns, and evaluate group membership. Principal Component Analysis (PCA), Partial Least Squares (PLS), and Partial Least Squares Discriminant Analysis (PLS-DA) are the most commonly utilized methods. PCA is a statistical method employed to identify similarities and differences among samples by reducing data dimensionality and transforming it into a simplified coordinate system, thereby preserving the majority of variability [87]. Principal components (PCs) represent linearly uncorrelated directions that optimize data variance, with PC1 capturing the most significant portion and subsequent PCs accounting for decreasing amounts. This process results in a PCA plot that illustrates sample variance and grouping tendencies [88]. PLS complements PCA by aiding in classification and the identification of significant variables within a dataset [89]. This approach depends on extracting latent variables that significantly influence variation in the response, thereby reducing the risk of overfitting when the number of factors exceeds the number of observations. This regression-based method facilitates the concurrent modeling of various independent variables while reducing dimensionality without sacrificing model integrity. PLS-DA improves upon PLS by incorporating predefined group memberships, ensuring uniform weighting of all variables, and enabling more robust and accurate classification, while also identifying the variables that contribute to group differentiation [90]. Cross-validation is essential for evaluating predictive performance and determining the optimal number of components required to uphold statistical integrity. Additionally, Soft Independent Modeling of Class Analogy (SIMCA) is commonly employed in untargeted metabolomics. SIMCA develops separate PCA models for each predefined class and subsequently evaluates whether new samples are within the confidence intervals of these models, aiding in class membership identification and outlier detection [90].

4. Extraction of Phenols from Complex Natural Matrices

Extracting phenolic acids and flavonoids from complex natural matrices requires balancing selectivity, throughput, and compatibility with downstream MS analysis. In the following sections, we provide a concise, practice-oriented overview of solvent selection and three widely used extraction modalities: ultrasound-assisted extraction (UAE), microwave-assisted extraction (MAE), and solid-phase extraction (SPE) (

Table 8. Comparison of ultrasound-assisted extraction (UAE), microwave-assisted extraction (MAE), and solid-phase extraction (SPE) for phenolic acids and flavonoids from complex natural matrices.

Method	Extraction Principle and Typical Conditions	Strengths (Target Classes/Matrices)	Limitations and Risks	LC–MS(/MS) Implications for the Resulting Extract
UAE	Cavitation → rapid desorption/diffusion 50–80% aqueous alcohol; mild acid ≤40–50 °C; 10–30 min; controlled amplitude/duty	Fast; low thermal load Strong for cinnamates, flavanones/flavones Good in sugar-rich matrices when parameters constrained	Hot-spot oxidation if aggressive Matrix-specific tuning (pH/T/amplitude) needed	Align solvent with initial LC phase; favor negative ESI ([M–H]−) Often fewer sugar-borne co-extracts than MAE → lower ion suppression; light cleanup if waxes/lipids present (propolis)
MAE	Dielectric/volumetric heating; minutes-scale 60–100% MeOH/EtOH; mild acid Tight control of power/hold/temperature	Very rapid; high throughput Efficient for hydroxybenzoic and hydroxycinnamic acids Effective on robust plant residues	Over-power/time → glycoside cleavage, oxidation In sugar-rich matrices aglycones can degrade Tends to co-extract interferents without cleanup	Plan SPE to remove sugars/organic acids and stabilize adducts Tune polarity/pH to preserve glycosides when needed Useful for flavonols/benzoates with constrained dielectric input
SPE	Adsorptive cleanup/enrichment Acidified load; LC-compatible MeOH/ACN elution Sorbent selection: C18, polymeric, mixed-mode	Removes sugars/salts/lipids → ↑S/N, reproducibility Stabilizes electrospray; matrix-agnostic cleanup	Polar acids can break through C18 without strong acidification Glycosides show variable recovery; risk of over-retention	Essential to curb ion suppression and adduct variability Standardize loading/elution pH Consider polymeric/mixed-mode phases when C18 under-retains (honey/propolis/juices)

).

4.1. Solvent

The selection of a solvent is significantly influenced by the analytical objective. For the purpose of quantifying phenol concentrations, matrix-optimized extraction protocols are frequently employed. In characterization workflows, extracts are prepared in LC–MS–compatible solvents to maintain chromatographic integrity and limit ion suppression [72,78,79]. The composition of the solvent/acid dictates co-elution into the electrospray, thus influencing the ionization efficiency of phenolic acids and flavonoids.

Mid-polarity aqueous organics are the most reliably effective first line in botanicals, propolis, and honey. Across 37 vegetables, the literature showed that 70% acetone is the most efficient solvent for recovering polyphenolic antioxidants, delivering the highest Total Phenolic Content (TPC) overall [91]. For berries, acetone/water 70/30 maximized total phenolic compounds, while methanol/water/acetic acid 70/29.5/0.5 gave the best anthocyanins, showing how solvent and mild acidification steer class-selective recovery [92]. In lignocellulosic agro-byproducts such as brewer’s spent grain, 60% acetone–water extracted antioxidant phenolic compounds more efficiently than aqueous alcohols [93]. However, acetone-rich extracts frequently contain chlorophylls, waxes, and other nonpolar compounds that exacerbate ion suppression in ESI, while water-dominant alcohols from sugar-laden matrices, particularly honey, co-extract saccharides and organic acids that compete for charge and diminish signal; both phenomena are consistently reported in phenolic workflows and honey case studies [81]. Extracts reconstituted to align with the initial mobile phase and analyzed via negative ESI (preferred for phenolic acids and numerous flavonoids) produce more stable [M–H]⁻ signals and fewer matrix-induced artifacts and most papers listed in Table 4 and Table 5 obtain that practice.

Acidification

Reducing solvent pH enhances the extraction of phenolic acids and flavonoids by converting weak-acid phenolic compounds (such as hydroxybenzoates and hydroxycinnamates) into their neutral, less-ionized states. This diminishes electrostatic interactions with cell-wall polysaccharides and promotes solubility in aqueous alcohols; this fundamental mechanism explains why acidified hydroethanolic systems are more effective than neutral media in various plant matrices [94]. The type and strength of the acid influence both extraction kinetics and the distribution of chemical forms. In purple maize and related matrices, formic acid at low pH increased anthocyanin recovery and antioxidant measurements relative to other acids, indicating that organic acids can preserve the protonated anthocyanin cation without extensive hydrolysis [95]. In contrast, Nuutila et al. showed that robust mineral acid conditions (1.2 M HCl in 50% MeOH, 80 °C, 2 h) quantitatively hydrolyze flavonoid glycosides to aglycones [96]. It is useful because it converts a chemically dense, glycoside-rich matrix into a manageable analyte collection. By converting diverse flavonoid glycosides to a small number of aglycones (for which calibrated standards exist), the Nuutila et al. protocol enabled reliable, comparable HPLC quantification without exhaustive glycoside libraries, while explicitly balancing complete hydrolysis against degradation of labile phenolic compounds. This is analytically advantageous, yet it alters the native profile if employed during extraction instead of as a distinct hydrolysis step.

4.2. Ultrasound (UAE) and Microwave Assisted Extractions (MAE)

UAE accelerates mass transfer via cavitation-driven cell disruption, which shortens contact time at modest bulk temperatures [97]. For phenolic acids and flavonoids, this improves recovery while reducing thermal and oxidative exposure, provided amplitude, duty cycle, and temperature are constrained to avoid hotspot-induced oxidation of o-dihydroxylated species or acid-catalyzed high phenolic yields at 50–80% aqueous alcohol can be achieved through careful temperature control (≤40–50 °C) and pulsed sonication, with stability gains due to rapid desorption and enzyme deactivation. This is demonstrated in citrus peel and other lignocellulosic byproducts, where UAE increased phenolic acid and flavonoid recovery compared to conventional maceration [97]. The UAE time was the primary factor in propolis (50:50 methanol–water), with an optimal extraction time of approximately 15 min [98]. Class selectivity was influenced by pH, with neutral pH favoring flavonoids (luteolin, quercitrin, hesperetin) and acidic pH favoring phenolic acids (benzoates, chlorogenic acid). It is important to note that the same compounds extracted in matrix degraded significantly more than standards sonicated in solvent, which emphasizes the necessity of optimizing in-matrix. When compared to 95 °C reflux (15 min, pH 2), UAE preserved both classes better [98].

MAE combines volumetric dielectric heating with steep, short-lived temperature/pressure gradients that speed up the release and diffusion of flavonoids and phenolic acids from complex matrices [99]. This process typically reduces extraction times, thereby limiting hydrolysis and oxidation in comparison to conventional reflux, provided that power, hold time, and solvent polarity are optimized to prevent glycoside cleavage or oxidation. MAE demonstrated significant advantages in both food and botanical matrices. Response-surface-optimized MAE (~185 °C, 1000 W, 20 min; 100% MeOH; 10:1 solvent:solid) facilitated the rapid recovery of hydroxycinnamic and hydroxybenzoic acids in rice grains, including ferulic, p-coumaric, sinapic, vanillic, and protocatechuic acids [100]. In onions, MAE tuned by Box–Behnken design maximized quercetin glycosides (quercetin-3-O-glucoside) with ≈94% MeOH at pH 2 and 50 °C for flavonols, and a slightly more aqueous solvent at higher temperature for peak DPPH activity [101]. This resulted in rapid, precise, and high-yield extracts.

From a stability standpoint, both UAE and MAE work because they compress the time–temperature integral of extraction while enhancing solvent access. However, the comparison of UAE and MAE in honey reveals a distinct stability split: in neat standards, ultrasound maintained phenolic compounds in a largely intact state (approximately 90% recovery), whereas microwaves reduced benzoic-acid derivatives and flavonoid aglycones to approximately 70–80% [102]. In honey, phenolic acids and glycosides (quercitrin, rutin, hesperidin) remained stable under both methods, but aglycones collapsed to below 10%. In a sugar-only mimic, sequential microwave and then ultrasound nearly eliminated flavonols. These findings suggest that matrix and energy input determine what survives to MS [102]. In monofloral honeys, UAE and MAE show complementary selectivity: with 70% ethanol, ultrasound yields more total flavanones, flavones, hydroxycinnamic acids, and total phenols, whereas microwaves enrich total flavonols and hydroxybenzoic acids, consistent with cavitation-driven desorption versus dielectric heating and class-specific stability [103]. Antioxidant readouts follow suit, with higher ORAC and DPPH in microwave extracts and higher ABTS in ultrasound extracts; method choice should match the target classes and assay end points, favoring ultrasound for cinnamates and flavanones or flavones and microwaves for flavonols and benzoates [103]. Overall, method choice should be aligned to target classes and matrix: UAE for cinnamates and flavanones or flavones in sugar-rich systems, MAE for flavonols and benzoates, or for time efficient processing of robust plant wastes, with solvent and pH tuning and minimal thermal load to safeguard structural integrity.

4.3. Solid Phase Extraction (SPE)

Complex matrices saturate electrospray with salts, sugars, and lipids, resulting in ion suppression/enhancement and unstable adduct formation. A well-designed SPE step enriches analytes (yield), provides consistent recovery, and prepares the extract for effective ionization. Recent metabolomics and analytical reviews have concluded that ESI performance is inextricably linked to sample cleanup and solvent composition [104].

SPE improves signal-to-noise and reduces ion suppression compared to direct injection or liquid–liquid extraction. However, small, polar phenolic acids may show weak retention and breakthrough on silica C18 unless strongly acidified or captured on polymeric or mixed-mode phases. Glycosylated flavonoids may also exhibit variable recoveries depending on hydrogen-bonding and π-π interactions, and over-strong retention can prevent quantitative desorption. These trade-offs are widely reported for honey phenolic compounds and in larger SPE reviews of phenols, which encourages matrix- and analyte-aware cartridge selection [105].

Recent improvements in SPE sorbent-coating materials address wash-off, which π-π stacking alone cannot prevent, by introducing hydrophobic interactions: Ma et al. created a new sorbent by coating hydrophilic NH₂-MIL-101(Fe) with a hydrophobic TAPB-FPBA COF shell. The sorbent’s combined hydrophobic, π-π, and hydrogen-bonding domains stabilized retention and, with only 5 mg per cartridge, outperformed C18 (R² ≥ 0.9967; LODs 0.02–0.08 ng mL⁻¹; ≈85–105% recoveries in water, grape juice, and honey) [106].

4.4. Data-Driven Optimization of Phenolic Extraction

Reports employing response surface methodology show that solvent composition, duration, and thermal input can be systematically optimized to maximize yield. For example, in rambutan peel, an agro-industrial by-product, accelerated solvent extraction optimized with a Box–Behnken design identified an optimal extraction process. In detail, 54 percent aqueous ethanol at 60 °C for 34 min demonstrated to be the optimal because water swells the cell wall and accelerates diffusion while ethanol solubilizes less polar phenolic acids and flavonoid aglycones and limits sugar co-extraction, thereby maximizing extraction yield, TPC, total anthocyanin content, and ABTS activity [107]. In addition, Brzezińska et al. applied response surface methodology to optimize extraction from spent coffee grounds, identifying 65% ethanol–water, a solvent-to-matrix ratio of 51 mL g/mL, and ultrasound for 30 min at 60 °C as optimal [108]. As another illustration, Dahmen-Ben Moussa et al. optimized ultrasound-assisted extraction of phenolic compounds from Scenedesmus and Chlorella using a central composite (RSM) design that varied solvent (methanol, ethanol, ethyl acetate), temperature (4–80 °C), and time (0.1–4 h) [109]. Methanol yielded the highest TPC (4 °C/4 h for Scenedesmus; 80 °C/4 h for Chlorella). In general, the literature indicates that response-surface methods effectively adjust solvent systems and time/temperature to the matrix and analytical objective, thereby promoting mobile-phase-compatible, frequently hydroalcoholic mixtures, to optimize phenolic recovery and downstream LC–MS performance.

Recently, researchers have begun optimizing extraction processes with machine-learning models. In particular, machine-learning-guided MAE of pomegranate peel accurately mapped power, time, temperature, and solid-to-solvent ratio to total phenolic compounds, tannins, and antioxidant activity, revealing microwave power as the key factor affecting phenolic acid and flavonoid stability [110]. Applied to complex plant, propolis, and honey matrices, the same workflow trains supervised models on a small, carefully designed experiment set and uses their insights (feature importance/SHAP with Bayesian or active-learning optimization) to locate operating points that raise yields while constraining degradation [110]. Looking ahead, coupling such models to real-time readouts could fine-tune MAE or UAE during extraction, delivering sample-specific, near-optimal protocols that balance recovery with preservation of molecular integrity. This is particularly important as UAE and MAE can co-extract matrix components that afterwards can influence electrospray response (for example: waxy/lipidic fractions from propolis or high sugar load in honey), thereby increasing ion suppression and adduct formation in MS/MS analysis.

5. Conclusions

This review relates the fundamental chemistry of phenolic acids and flavonoids to what is seen in MS and on the column. With today’s high-resolution technologies and established libraries, structural identification and sample fingerprinting of phenolic compounds and their derivatives are generally reliable. The challenge is quantification, particularly across heterogeneous natural matrices: propolis, resinous exudates, and green plant tissues all present co-extractives that alter recovery, ionization, and calibration behavior. Shared libraries, explicit reporting of identification confidence, and, most importantly, consistent quantitative methodology (matrix-matched calibration, stable-isotope internal standards, and inter-laboratory benchmarks) are required to ensure that results are truly comparable across studies and matrices.