Barley Wine in Focus: NMR Metabolomics Reveals Style and Barrel Aging Differences

Abstract

Barley wine is one of the most chemically complex and historically significant beer styles, yet its molecular composition remains largely unknown. This study aims to create the first detailed molecular framework for understanding the chemical diversity of barley wine and cereal wines. The chemical diversity of barley wines and related “cereal wines” made from wheat, oats, and rye, including barrel-aged varieties, is examined using ¹H nuclear magnetic resonance (NMR) metabolomics. Distinct cereal-dependent signatures were revealed by multivariate analyses. High levels of fusel alcohols and phenolic acids were present in barley wines. Elevated levels of pyruvate and aromatic amino acids were found in wheat wines, and high levels of maltodextrin, arabinose, and trigonelline were found in oat and rye wines. A comparison of sub-styles showed that English and American barley wines were different based on ester and complex sugar profiles. Barrel aging introduces changes dependent on the barrel’s origin. A reliable classification of barrel origin was allowed for by a decision tree with four diagnostic metabolites—5-hydroxymethylfurfural (HMF), acetaldehyde, mannose, and tryptophan. The way in which raw materials, fermentation conditions, and the reuse of barrels collectively influence their metabolomes is exemplified. Verifying the authenticity of beer, evaluating its quality, and generating new ideas for high gravity brewing are all cases in point for this approach.

1. Introduction

Barley wine is one of the strongest and most complex traditional beer styles. It is characterized by its high original gravity, elevated alcohol content, and rich sensory attributes. This style emerged in England during the late 18th and 19th centuries as an alternative to imported French wines. However, the history of strong grain-based fermented drinks is much older. Ancient Greek authors such as Xenophon (Anabasis) and Polybius (Histories) wrote about a drink called “krithinos oinos (κριθῖνος οἶνος)” or barley wine, indicating that wine-like drinks made from cereal grains have existed since antiquity. This drink was consumed in Thrace and Phrygia. Similar beverages were prepared by the Thracians and were sometimes referred to as bryton [1].

The designation “wine” reflects the strength and sipping character of these beers, which are often compared to fortified wines or cognac because of their intensity and aging potential. Barley wine gained prominence in England when trade conflicts and wars with France limited access to Bordeaux and Burgundy wines. Strong ales brewed with large amounts of grain and long boiling times were positioned as a domestic counterpart to refined wines. By the late 19th century, Bass marketed its No. 1 as “Barley Wine”, one of the first beers to use the label [2]. In 1976, Anchor Brewing introduced Old Foghorn, bringing the style to the United States and giving rise to the more hop-forward “American” variant [3].

Today, two major stylistic families are recognized: English barley wine, which exhibits malt-driven notes of caramel, toffee, dark fruit, and oxidative complexity, and American barley wine, which has higher hopping rates, bitterness, and resinous-citrus hop aromas [4]. Beyond these categories are international and hybrid interpretations, though systematic compositional studies remain scarce. An emerging area of diversification involves substituting barley with alternative cereals to produce “cereal wines”. First brewed commercially in the U.S. in the late 1980s, wheat wines are characterized by bready sweetness, a lighter body, and reduced bitterness compared to classic barley wines [3]. Rye wines incorporate high proportions of rye malt, which adds spiciness and phenolic notes while enhancing mouthfeel through arabinoxylans [5]. Though rare, oat wines are noted for their creamy texture and high β-glucan content, which influences viscosity and foam stability. While these experimental variations have attracted brewers’ interest, peer-reviewed chemical or metabolomic studies remain limited. This leaves the molecular composition of cereal-based analogs largely unexplored.

Another defining characteristic of barley wine and strong ales is barrel aging. In the United States, federal law requires bourbon to be aged exclusively in new, charred American oak barrels [6]. After a single use, these barrels enter a global secondary market and are often repurposed for aging beer. The barrels provide oak-derived compounds such as vanillin, lactones, and tannins, as well as residual bourbon congeners [7]. In Europe, brewers often use casks that previously held whisky, many of these casks were originally ex-bourbon or ex-sherry barrels [8]. These vessels promote oxidative, vinous, and phenolic characteristics to the beer. Fortified wine casks (sherry, port, and Madeira) impart nutty and dried fruit notes, while rum, brandy, and calvados barrels contribute estery and fruity complexity [9]. Repeated use of barrels alters extraction: first fills provide stronger oak and spirit influence, while subsequent fills emphasize oxidative and microbial development [10]. Despite their sensory importance, systematic studies on how barrel provenance and reuse shape beer metabolomes are limited.

Recent analytical research has expanded our understanding of how wood and aging affect beer chemistry. Gas chromatography-mass spectrometry (GC-MS) and liquid chromatography mass spectrometry (LC-MS) are commonly used to profile the volatile and phenolic compounds extracted from oak during maturation [11]. Approaches such as GC-olfactometry and aroma extract dilution analysis (SAFE-GC-O/AEDA) have identified vanillin, ethyl vanillate, and oak lactones as key aroma-active molecules in high-alcohol beers aged in wood [11]. Similarly, LC-based studies have demonstrated that the origin of the oak and the malt composition significantly impact the phenolic and bioactive profiles of wood-aged beers [10,12]. American oak increases the levels of vanillic and syringic acid levels compared to French oak [12]. Other studies have examined the dynamics of chemicals and microbes during beer maturation. These studies have revealed that barrel aging alters the concentrations of wood-derived compounds, such as vanillin, furfural, and whisky lactones. They have also shown that microbial activity within casks contributes to compositional evolution over time [13]. Research on other barrel-aged beverages, beyond beer, further demonstrates the analytical potential of linking molecular fingerprints to maturation history. For instance, Kew et al. used ¹H nuclear magnetic resonance (NMR) spectroscopy coupled with chemometrics to distinguish Scotch whiskies based on wood type and maturation parameters [14]. This demonstrates NMR’s ability to detect subtle compositional variations related to cask origin. Similar NMR-based metabolomic approaches in winemaking have revealed how barrel origin, utilization time, and volume affect the metabolite composition of red wines during oak maturation [15,16]. In brewing science, ¹H NMR metabolomics has successfully differentiated beers based on style, raw materials, and geographical origin [17,18,19,20]. Despite their historical significance, high extract content, and increasing popularity in the craft brewing sector, no comprehensive, NMR-based metabolomic study has focused on barley wine or its modern variants. The chemical complexity of these beers, which is shaped by various cereal compositions, prolonged boiling, and aging processes, has not been extensively studied at the molecular level.

To the best of our knowledge, this is the first study to systematically characterize barley wine and related cereal wines using NMR metabolomics. This study distinguishes between styles (English, American, and hybrid barley wines), compares cereal-derived variants, and provides preliminary evaluations of barrel-aged samples. Thus, it establishes a molecular framework for understanding the compositional diversity of high-gravity, wine-like beers. This approach lays the groundwork for future research. The research will focus on the authenticity of these beers. It will also focus on their aging mechanisms. Another focus will be how raw materials impact their chemical composition.

2. Materials and Methods

2.1. Samples

A total of 20 barley wine beers were acquired from several craft beer shops. The samples were categorized based on their raw material (barley, wheat, rye, or oats), sub-style (English, American, or other), and aging treatment (unaged, whisky, bourbon, fortified wine, or fruit spirit). Supplementary Table S1 provides detailed information on alcohol by volume (ABV), sub-style, country of origin, and brewery.

2.2. Sample Preparation

Beer samples were degassed in an ultrasonic bath (Elmasonic S30H, Elma Schmidbauer GmbH, Singen, Germany) for 15 min to remove the dissolved CO₂. Then, 500 µL of the degassed sample were combined with 50 µL of a 1.0 M deuterated phosphate-buffered solution (pD 4.4, prepared in D₂O) containing 1.27 g NaH₂PO₄ per 10 mL D₂O (99.9%, Deutero GmbH, Kastellaun, Germany), 0.1% TSP (sodium salt of 3-(trimethylsilyl)-2,2,3,3-tetradeuteropropionic acid, 98%, Merck KGaA, Darmstadt, Germany) as an internal chemical shift reference, 0.05% NaN₃ (Sigma-Aldrich, Darmstadt, Germany) as a preservative and 55 µL H₃PO₄ (Sigma-Aldrich, Darmstadt, Germany). All reagents were of analytical grade. The mixture was vortexed briefly to ensure homogeneity, then transferred into 5 mm NMR tubes (type 507-PP-7, Rototec-Spintec GmbH, Bad Wildbad, Germany) for the further spectroscopic acquisition.

2.3. NMR Spectra

Proton NMR spectra were acquired on a Bruker Avance NEO 400 MHz spectrometer (Biospin GmbH, Rheinstetten, Germany) equipped with a BBO probe at 300.0 ± 0.1 K using the zgcppr water-suppression sequence. The principal acquisition parameters were as follows: The acquisition included 42,104 data points, a spectral width of 13.2 ppm, an acquisition time of 4.0 s, and a relaxation delay of 4.0 s. Each spectrum was obtained by collecting 256 scans, preceded by 16 dummy scans. The free induction decay (FID) was zero-filled and apodised with a 0.3 Hz line-broadening function prior to Fourier transformation. Manual phasing and multipoint baseline correction were performed in MestReNova 14.2.3, to attenuate macromolecular resonances. All chemical shifts were referenced internally to the TSP signal at 0 ppm.

2.4. Metabolite Identification and Quantification

A comprehensive approach was used to assign and confirm metabolites. This approach integrated reference database matching (Human Metabolome Database and Biological Magnetic Resonance Data Bank), literature comparison and targeted experimental validation [17,21,22,23]. Two-dimensional experiments (heteronuclear single quantum coherence–HSQC and total correlation spectroscopy–TOCSY) were used to resolve overlapping signals and confirm key assignments. Spike-in experiments with authentic standards for representative sugars, amino acids and organic acids were performed to validate the selected metabolites. We quantified all 55 unambiguously identified metabolites using a standardized calculation formula recommended by the International Organisation of Vine and Wine (OIV) to ensure reproducible and comparable concentration values across samples [24]. Due to partial signal overlap in several spectral regions, deconvolution (line fitting) was applied in place of direct peak integration.

2.5. Statistical Analysis

The quantitative data of all identified compounds were subjected to multivariate statistical analysis to explore compositional differences among samples. The following methods were employed: hierarchical cluster analysis (HCA), orthogonal partial least squares discriminant analysis (OPLS-DA), orthogonal two-way partial least squares discriminant analysis (O2PLS-DA) and Chi-square automatic interaction detector (CHAID) decision tree.

We applied HCA using Ward’s linkage method and Euclidean distance to visualize natural groupings among samples based on metabolite composition. OPLS-DA was used to classify samples according to raw material (barley, wheat, or other, including rye and oat) and sub-style (English, American, or other), while O2PLS-DA was used to differentiate samples by barrel-aging type (whisky, bourbon, fortified wine, fruit spirits, or unaged beer). Unlike OPLS-DA, which models systematic variation in the predictor matrix (X) related and unrelated to class information, O2PLS-DA partitions shared and unique variation between X and Y. This improves interpretation in complex multifactorial datasets. CHAID decision trees generated hierarchical, rule-based classifications in which each split maximized between-group separation.

HCA, OPLS-DA, and O2PLS-DA analyses were performed using SIMCA 18.0 (Umetrics, Umeå, Sweden). We generated CHAID trees with SIPINA Research 3.12 (producer Ricco Rakotomalala). Variable importance in projection (VIP) scores identified the metabolites most relevant for class discrimination, and their relative contributions were visualized through contribution plots (Figures S1–S3). Model performance was evaluated using a receiver operating characteristic (ROC) curve analysis, including calculation of the area under the curve (AUC) to assess discriminatory power and predictive ability (Figure S4). We also examined misclassification matrices (Table S2). The CHAID decision tree was used to identify the key compounds that define each barrel-aging category.

3. Results and Discussion

3.1. NMR Profiling of Barley Wines

Fifty-five metabolites were identified and quantified using ¹H NMR spectroscopy across the cereal wine/barley wine samples (

Table 1. List of the 55 metabolites identified and quantified in the analyzed cereal wine samples, including their concentration ranges (minimum–maximum and mean, mg/L) and the chemical shift (δ in ppm) of the signal used for quantification.

Compound	Min–Max, Average, mg/L	1H δ (ppm), Multiplicities *
Alcohols
1-Propanol	27.1–135.5, 58.4	1.54, m
2-Phenylethanol	10.9–123.1, 44.5	7.31, m
2,3-Butanediol	170.4–851.1, 388.3	1.12, d
Ethanol	42,830.9–110,350.5, 75,876.0	1.17, t
Glycerol	1771.7–5645.1, 3399.4	3.54, dd
Isobutanol	27.3–132.3, 68.8	1.72, m
Isopentanol	55.6–151.7, 104.1	1.42, q
Methanol	2.9–93.1, 23.0	3.35, s
Saccharides
αα-Trehalose	12.1–572.9, 173.5	5.17, d
Arabinose	26.0–461.8, 125.7	4.50, d
Fructose	0.0–19,868.2, 1368.7	4.09, m
Glucose	4747.4–98,342.8, 22,545.4	5.21, d
Kojibiose	143.9–820.8, 385.2	5.07, d
Lactose	0.0–2693.9, 173.7	4.42, d
Maltodextrin	2385.1–78,436.4, 30,483.8	3.26, dd
Mannose	15.9–273.4, 88.5	5.16, d
Sucrose	25.7–3328.3, 289.0	4.20, d
Raffinose	51.8–395.6, 122.4	5.00, d
Xylose	0.0–1354.3, 319.6	5.16, d
Organic acids
Acetic acid	91.6–664.8, 320.0	2.02, s
Citric acid	40.1–656.2, 167.9	2.79, d
Formic acid	1.4–80.7, 20.5	8.40, s
Fumaric acid	0.6–21.9, 5.5	6.57, s
GABA	31.5–265.8, 92.5	3.02, t
Gallic acid	3.0–79.8, 35.9	7.03, s
Lactic acid	83.8–2429.2, 467.4	1.34, d
Maleic acid	3.7–85.7, 27.6	6.34, s
Malic acid	11.6–1588.7, 199.9	2.85, dd
Pyruvic acid	0.0–132.4, 54.2	2.35, s
Sorbic acid	0.0–120.7, 10.8	1.84, d
Succinic acid	136.6–479.1, 253.6	2.58, s
Tartaric acid	14.7–260.0, 78.9	4.54, s
Amino acids
Alanine	43.4–1199.5, 306.8	1.46, d
Histidine	0.2–8.2, 2.9	7.83, s
Isoleucine	5.8–63.5, 26.0	1.00, d
Leucine	27.6–203.3, 93.3	0.94, d
Phenylalanine	31.9–271.9, 121.6	7.40, m
Pyroglutamic acid	140.0–699.2, 269.0	2.48, m
Tryptophan	10.8–87.7, 25.4	7.26, t
Tyrosine	28.8–253.0, 114.2	6.86, d
Valine	4.1–313.2, 105.0	0.97, d
Nucleosides and Nucleobases
Adenosine	0.8–127.4, 47.5	8.36, s
Guanosine	8.1–56.4, 25.8	7.98, s
Inosine	0.4–27.8, 8.2	8.34, s
Thymidine	9.6–104.4, 37.2	7.65, s
Uracil	0.0–79.6, 22.0	7.51, d
Uridine	10.0–222.3, 134.9	7.87, d
Other organic compounds
Acetaldehyde	0.5–54.5, 6.6	9.66, q
Betaine	43.4–319.7, 169.9	3.24, s
Choline	57.6–270.9, 145.8	3.18, s
Furfural	0.0–11.1, 1.8	9.49, s
HMF	1.1–69.8, 9.2	9.44, s
Isoamyl acetate	0.0–131.5, 80.6	2.07, s
Trigonelline	5.8–22.5, 9.7	9.11, s
Vanillin	0.0–32.1, 2.2	9.69, s

* Multiplicities: s—singlet, d—doublet, t—triplet, dd—doublet of doublets, q—quartet, m—multiplet.

). This dataset includes volatile compounds, higher alcohols, polyols, monosaccharides, oligosaccharides, organic acids, amino acids, and other nitrogenous compounds, as well as several minor markers associated with roasting, the Maillard reaction, and wood contact. The concentration ranges (minimum to maximum) and means from Table 1 are used below to describe the chemical space of the samples and to interpret the likely sources of the processes and ingredients.

A representative ¹H NMR spectrum of barley wine illustrating the main spectral regions and selected metabolite assignments is shown in Figure 1.

Ethanol dominated the metabolome (minimum 42.8 g/L, maximum 110.35 g/L, mean 75.9 g/L), confirming the high-gravity nature of the material. Higher alcohols, which are characteristic of amino acid catabolism, were present at high levels, including 2,3-butanediol (170–851 mg/L; mean 388 mg/L), 1-propanol (27–136 mg/L; mean 58 mg/L), isobutanol (27–132 mg/L; mean 69 mg/L), and isoamyl/isopentanol (55–152 mg/L; mean 104 mg/L). These levels are consistent with those observed in previous NMR and metabolomic studies of dark beers, in which elevated higher alcohol levels arise from enhanced flux in the Ehrlich pathway under nitrogen limitation and osmotic stress [17,18]. Glycerol levels were also high (1.77–5.65 g/L; mean 3.40 g/L), reflecting yeast osmoregulation during high-gravity fermentation. These levels contribute to the viscous mouthfeel characteristic of barley wines.

Carbohydrates comprised most of the non-volatile pool and showed significant variation among samples. Maltodextrin had a wide concentration range, from 2.4 to 78.4 g/L, with an average of 30.5 g/L. Interestingly, not all barley wines contained high levels of maltodextrin. Some samples, such as several fortified-wine-barrel and non-barrel-agedwines, showed relatively low dextrin levels (2–12 g/L). Others, notably several fruit-spirit and whisky-barrel samples, contained large pools of maltodextrin (>40 g/L). This distribution suggests variability in recipes and processes. High maltodextrin levels reflect high original gravity combined with mash regimes and starch conversion conditions that leave substantial non-fermentable dextrins. Such conditions include a high mash-out temperature, the use of specialty malts, or the deliberate addition of adjuncts to increase body. Conversely, low maltodextrin samples likely experience higher enzymatic conversion or more complete attenuation. The mean glucose concentration was approximately 22.5 g/L (range 4.7–98.3 g/L). The presence of sucrose and fructose (>1.5 g/L) in several samples indicates the use of residual or added fermentables in some recipes. Many samples contained lactose (up to 2.7 g/L), consistent with the addition of milk sugar to certain formulations. This carbohydrate patterns align with those observed in previous beer metabolomics studies using NMR. Strong, dark beers typically exhibit higher levels of dextrins and residual sugars than lighter lagers [19]. However, significant variability among samples reinforces the idea that barley wine is not chemically uniform. When interpreting metabolomic markers, one must consider recipe choices, such as adjuncts, kettle sugars, and mash profile, as well as yeast attenuation and fermentation completeness. Compared to similar dark beers described in the literature [17], elevated levels of pentoses and other minor sugars, such as arabinose, xylose, mannose, and raffinose, were often observed. The concentrations of these sugars were as follows: arabinose (minimum 26 mg/L, maximum 462 mg/L, and a mean of approximately 126 mg/L); xylose (minimum 0 mg/L, maximum 1354 mg/L, and a mean of approximately 320 mg/L).

These pentose and hemicellulose sugars can originate from the husks or adjuncts of cereals, a phenomenon that is more prevalent with the inclusion of rye or oats—or from the partial hydrolysis of wood polysaccharides during barrel contact. αα-Trehalose levels were notably high, averaging several times higher than in some Weizenbock datasets, with minimum and maximum levels of 12 and 573 mg/L, respectively, and a mean of 173 mg/L [17]. These levels are most reasonably explained by a yeast stress response (αα-trehalose accumulation under osmotic or ethanol stress) and/or release during yeast autolysis in prolonged maturation. Together, elevated pentoses and αα-trehalose levels suggest the influence of raw materials and yeast physiology in high-gravity fermentation. The measured organic acids included lactic acid (with a mean concentration of approximately 467 mg/L and a range of 84 to 2429 mg/L), acetic acid (with a mean concentration of approximately 320 mg/L), and succinic acid (with a mean concentration of approximately 254 mg/L). Other acids present in smaller amounts included citric acid, malic acid, fumaric acid, and formic acid. High concentrations of lactic and succinic acids are characteristic of aged, full-bodied ales and may reflect yeast metabolism, as well as the limited activity of lactic acid bacteria associated with wooden surfaces. Moderate acetic acid concentrations suggest that although barrel porosity enables acetic production, overt acetification or spoilage was not widespread in this sample set. Similar increases in certain organic acids during storage and forced aging have been reported in comparative beer aging studies. Our data align with these mechanistic trends, while also showing pronounced sample heterogeneity, which is likely tied to differing barrel histories and microenvironmental conditions.

Free amino acids, such as alanine, leucine, valine, and phenylalanine, were quantifiable in the tens-to-hundreds-of-milligrams-per-liter range. These amino acids serve as precursors for fusel alcohols and esters, indicating high-gravity fermentation, wherein yeast uptake is limited by ethanol inhibition. Nucleosides (adenosine, inosine, and uridine) and osmolytes (betaine and choline) were also present at low yet measurable concentrations. These concentrations suggest yeast autolysis and cell turnover during extended maturation. These observations are comparable to those in other beer and wine metabolomics reports [19,25,26].

Several markers associated with Maillard chemistry and wood contact were detected: 5-hydroxymethylfurfural (HMF, mean concentration of approximately 9.75 mg/L; range of 0 to 69.8 mg/L), furfural (mean concentration of approximately 1.83 mg/L), and vanillin (mean concentration of approximately 2.20 mg/L). These furanic compounds are produced during malt kilning or roasting, as well as through sugar dehydration reactions, or can be released from charred wood during barrel aging. On the other hand, vanillin arises from lignin breakdown in oak and is a common byproduct of wood aging [10,14]. Importantly, vanillin was quantified only in two samples, both of which were labeled as vanilla-infused barley wines. The vanillin content of natural vanilla is known to be approximately 1–2% of the dry pod, according to the research by Ranadive et al. and Gu et al. The presence of other vanilla or wood marker ratios, such as vanillic acid or p-hydroxybenzaldehyde, is inconsistent in two of the measured beers (BW9 and BW10). Therefore, the vanillin in these samples is likely to be an additive, rather than a product of extraction from the barrel wood [27,28]. Conversely, several other beers declared to contain vanilla showed no detectable vanillin signal.

The compositional diversity observed across the dataset reflects the combined effects of fermentation dynamics, cereal composition, and maturation environment. This variability underscores the chemical heterogeneity of barley wine and provides an analytical foundation for classifying it based on raw material, style, and barrel origin.

3.2. Differentiation of Barley Wines by the Type of Cereal Used (Barley, Wheat, or Other)

All 55 compounds were put through the cluster analysis process, which was used to identify similarities among samples based on their chemical profiles. The resulting dendrogram, illustrated in Figure 2, reveals the presence of three distinct clusters.

The cluster analysis separated the samples into three distinct groups based on their main cereal types. The largest group, Cluster 1, contained all fifteen barley wine samples. This cluster was chemically the most distant from the others, highlighting the fundamental differences in the metabolomic profiles of traditional barley wines and alternative cereal wines. The remaining samples formed a connected supercluster. Cluster 2, which contained the three wheat wine samples and was directly linked to Cluster 3. Cluster 3 consisted of the highly distinct rye (OW1) and oat (OW2) wine samples. This connectivity suggests that, although wheat, rye, and oat wines are closely related, the barley-based matrix represents the dominant chemical divergence within the “cereal wine” category. We used OPLS-DA to maximize discrimination among these three clusters. Then, an OPLS-DA model was created to analyze differences in metabolites among barley wines (n = 15), wheat wines (n = 3), and other cereal wines (n = 2, including oat and rye). VIP analysis (VIP > 0.94) identified 19 metabolites that significantly contributed to group separation. These metabolites included core fermentation markers such as 1-propanol, ethanol, isobutanol, several organic acids including acetic acid, formic acid, fumaric acid, gallic acid, maleic acid, and pyruvic acid; amino acids such as isoleucine, phenylalanine, pyroglutamic acid, tryptophan, valine, thymidine. They also included saccharides (e.g., arabinose and maltodextrin), choline and trigonelline. The resulting optimized OPLS-DA model, built from this subset of discriminants, displayed robust explanatory power (R²X = 0.714, R²Y = 0.790) and acceptable predictive ability (Q² = 0.317). The model had two predictive and two orthogonal components. The score plot (Figure 3) clearly separated three groups of cereal-origin wines: barley wine (BW), wheat wine (WW), and other cereal wine (OW, from oats and rye). Thus, each group is defined by a coherent biochemical signature driven by the raw materials and fermentation dynamics.

The contribution plots indicated that wheat wines (WW) are mainly defined by pyruvic acid and tryptophan. The tryptophan signal is consistent with NMR-based wheat beer profiling, which determined that the concentration of tryptophan in different wheat beer styles ranged from 5.7 to 82.2 mg/L, with an average of 36 mg/L for Weizenbock and 22 mg/L for light wheat beer styles [17]. Therefore, the wheat wines we measured occupy the same metabolite families as wheat beers. The elevated pyruvate suggests intensified glycolytic flux and altered pyruvate branching. These features can arise in high-gravity, nitrogen-rich wheat worts.

Barley wines exhibited a classic high-gravity profile, characterized by elevated levels of ethanol, 1-propanol, and isobutanol, as well as an extensive residual amino acid pool, including isoleucine, phenylalanine, pyroglutamic acid, and valine and increased levels of acetic and gallic acids. This pattern is mechanistically consistent with the hypothesis that amino acids from malted barley drive flux in the Ehrlich pathway and fusel alcohol formation. The results have been found to be in line with general NMR beer surveys, which have shown that higher levels of fusel alcohols and residual dextrins are present in craft and high-gravity beers compared to lighter industrial beers [18]. The elevated gallic acid in BW (41 mg/L) aligns with studies linking darker malts and longer wort concentration/aging to higher phenolic extraction [12]. The presence of thymidine and choline likely indicates nucleotide turnover or yeast autolysis during extended maturation.

The oat and rye OW samples are primarily distinguished by carbohydrate-derived markers and an alkaloid signature. Maltodextrin concentrations were extremely high in this group, with an average of 74.0 g/L in OW, vs. 27.6 g/L in BW and 15.8 g/L in WW. Similarly, the average arabinose concentration was 298 mg/L in the OW samples, increasing to 462 mg/L in the oat wine, which is approximately triple the average for barley wine (102 mg/L). These levels far exceed those typically found in wheat beers. For instance, Chorbadzhiev’s Weizenbock has an average of 47 mg/L of arabinose and 21.5 g/L of maltodextrin [17]. These levels are also consistent with the high levels of arabinoxylan, and β-glucan found in oats and rye [29,30,31]. Partial hydrolysis of these non-starch polysaccharides during mashing liberates pentose monomers (arabinose and xylose) and large, poorly fermentable oligosaccharides (maltodextrins). These compounds increase the residual extract, viscosity, and sweetness of oat- or rye-rich beers. This is why they have a significant impact on the OW profile. OW samples also show elevated levels of formic acid, fumaric acid, and trigonelline.

These results demonstrate that the molecular fingerprint of high-gravity beers is significantly impacted by the type of cereal used. This impact can be attributed to variations in carbohydrate composition, amino acid metabolism, and alkaloid content. The different ways that barley, wheat, and mixed-cereal wines change during production appear to stem from variations in grain chemistry and yeast response. These findings lay the groundwork for future sensory and process-level correlations, which will further our understanding of the subject.

3.3. Metabolomic Differentiation of Barley Wine Sub-Styles

We used OPLS-DA on the ¹H NMR profiles of 15 barley wine samples and achieved clear separation of the three sub-styles—English, American, or other—in the score plot (Figure 4). The optimized model consisted of three predictive and three orthogonal components. It demonstrated robust explanatory power (R²X = 0.877; R²Y = 0.858) and moderate predictive ability (Q² = 0.280). A total of 24 metabolites were identified as key discriminators from the full metabolomic dataset. These metabolites include higher alcohols, esters, sugars, organic acids, amino acids, and nitrogenous compounds, which together define the chemical signatures that are characteristic of each substyle.

English barley wines were chemically characterized by high concentrations of the higher alcohol isopentanol, and isoamyl acetate. These compounds contribute fruity, banana-like aromas and mild fusel warmth [11]. Their abundance indicates moderate attenuation and fermentation under controlled conditions, which are typical of traditional English ale yeast strains that favor ester synthesis over extensive sugar metabolism. Relatively low levels of complex carbohydrates further imply restrained adjunct use and a malt-forward wort composition. The resulting sensory profile emphasizes yeast-derived esters and balanced alcohol complexity rather than residual sweetness or heavy malt caramelization. This profile is consistent with the production practices of classical English barley wine [2].

American barley wines are distinguished by their high concentrations of complex sugars and nucleosides. Examples include kojibiose (424.9 mg/L), raffinose (172.1 mg/L), αα-trehalose (209.2 mg/L), inosine (11.3 mg/L), and betaine (187.4 mg/L). This accumulation reflects a metabolically rich wort, which is characteristic of the robust body and pronounced residual sweetness of the American style and is derived from extensive malt selection and the addition of adjunct sugars. Elevated levels of malic acid may indicate incomplete conversion during fermentation or the presence of adjunct fruits. The increased presence of nitrogenous compounds, such as inosine, uracil, and betaine, indicates a nutrient-rich environment that supports active yeast metabolism. This metabolism sustains high alcohol production and contributes to the full mouthfeel of these beers. These molecular markers align with the style’s signature balance of malt intensity and assertive hop bitterness [2,4].

The other barley wine sub-style had a distinctly different and more complex metabolomic profile. It was enriched with 2,3-butanediol, arabinose, mannose, lactose, tyrosine, gamma-aminobutyric acid (GABA), gallic acid, maleic acid, pyruvic acid, succinic acid, tartaric acid, guanosine, uridine, and choline. Average levels of these compounds were 625.8 mg/L for lactose, 168.3 mg/L for tyrosine, 135.5 mg/L for GABA, 52.3 mg/L for gallic acid, and 173.5 mg/L for choline. These elevated levels suggest the use of unfermentable adjuncts and non-standard ingredients, which contribute to body, residual sweetness, and distinctive flavor complexity. The presence of amino acids and organic acids, along with osmolytes such as betaine and choline, indicates yeast stress and incomplete precursor assimilation, which often occurs under high-gravity or mixed-fermentation conditions. These characteristics are consistent with experimental, barrel-aged, or “pastry” barley wines, in which extended fermentation and maturation promote the accumulation of secondary metabolites [2]. The result is a chemically diverse matrix that produces a pronounced sensory richness and complexity that goes beyond what is observed in traditional sub-styles.

The molecular outcomes of high-gravity brewing depend on yeast behavior, malt and adjunct composition, and fermentation strategy. The distinct metabolomic patterns observed among English, American, and experimental barley wines demonstrate this. These results clearly demonstrate the biochemical basis of stylistic choices. These choices lead to compositional diversity. They also pave the way for investigating the effects of barrel aging.

3.4. Discrimination According to Barrel-Aging Type (Bourbon, Whisky, Fruit Spirits, Fortified Wines, and Non-Barrel Aged

We used an O2PLS-DA analysis to study how barrel reuse type influences the metabolite composition of 15 barley wine samples. While the model achieved acceptable explanatory power (R²X = 0.876, R²Y = 0.658), its predictive ability was modest (Q² = 0.197), indicating that while the main chemical differences between groups are well captured, some within-group variation remains. Nineteen discriminant metabolites were identified using VIP > 0.92, including alcohols (1-propanol, isopentanol, isoamyl acetate, and acetaldehyde), saccharides (αα-trehalose, arabinose, mannose, and xylose), organic acids (citric and succinic acids), amino acids and derivatives (isoleucine, leucine, pyroglutamic acid, and tryptophan), nucleosides (uridine and thymidine), and furanic compounds (furfural and HMF). The corresponding score plot (Figure 5) shows clear clustering according to the previously used barrel type: fortified wine, whisky, bourbon, fruit spirits, none.

Barley wines aged in fortified-wine barrels had higher levels of tryptophan, uridine, and isoamyl acetate, with average concentrations of 24.9, 154.3, and 94.6 mg/L, respectively. These increased levels of tryptophan and uridine may result from the extraction of nitrogenous compounds and nucleoside derivatives from the wine-saturated oak barrel. Wine barrels often contain yeast lees or residual wine that has undergone autolysis, a process that releases these compounds into the wood’s pores. This makes them available for extraction by the high-alcohol barley wine [32]. Tryptophan metabolism during secondary aging can also lead to the formation of indolic and floral aroma compounds that contribute to complex fruity notes [33]. The relatively high isoamyl acetate content aligns with residual yeast esterification activity under micro-oxygenated conditions in reused barrels. Such esters are associated with the sweet, vinous character typical of wood-aged strong ales.

Bourbon-barrel-aged barley wines were characterized by elevated levels of αα-trehalose (233.6 mg/L), mannose (117.9 mg/L), xylose (557.0 mg/L), leucine (125.5 mg/L), and acetaldehyde (13.0 mg/L). The high saccharide levels likely originate from the partial hydrolysis of the hemicellulosic components of the oak staves, a process that is enhanced by prior bourbon contact and charring [10,34]. These sugars and nitrogen-containing compounds can undergo Maillard reactions during storage, producing color- and flavor-active intermediates such as furfural and HMF [35]. The increased acetaldehyde, an intermediate in ethanol oxidation, further supports the occurrence of oxidative transformations during maturation. The higher leucine concentration may be linked to the release of amino acids from yeast autolysis and subsequent ester formation pathways, which provide mild malty or nutty background notes [36].

Barley wines aged in whisky barrels had significantly higher concentrations of glycerol (4.8 g/L), isopentanol (133.2 mg/L), succinic acid (302.9 mg/L), isoleucine (5.0 mg/L), pyroglutamic acid (412.2 mg/L), thymidine (56.5 mg/L), and acetaldehyde (19.7 mg/L). These results suggest a notable impact from yeast metabolism and oxidative aging. The increased presence of glycerol, isopentanol, and succinic acid indicates changes in yeast metabolism and the accumulation of primary and secondary fermentation byproducts in a spirit-rich, nutrient-poor environment. This phenomenon has been previously observed in studies of whisky matrices [37]. High acetaldehyde levels are best explained by partial ethanol oxidation and redox changes that commonly occur during barrel aging, contributing to aldehydic/sherry-like notes [14]. The enrichment of pyroglutamic acid is consistent with amino acid cycling and the oxidative transformation of glutamate and glutathione pools during aging. The increase in thymidine likely reflects nucleic acid breakdown or release from microbial or barrel residues. Taken together, these markers suggest that whisky barrels promote the extraction and retention of yeast-related metabolites, as well as the oxidative transformation of ethanol and amino acid pools. These changes produce a fuller body (glycerol), increased organic acidity (succinic acid), and a more pronounced aldehydic character (acetaldehyde).

Barley wines aged in fruit-spirit barrels had the highest levels of 1-propanol (85.6 mg/L), as well as roughly twice the amount of total saccharides (e.g., mannose and xylose), citric acid (448.9 mg/L), furfural (5.5 mg/L), HMF (29.3 mg/L), and isoamyl acetate (108.4 mg/L). The concurrent increase in furanic compounds (furfural and HMF) and monosaccharides can be explained by the extraction of hemicellulose-derived sugars from heated or toasted oak, followed by their acid-catalyzed degradation via the Maillard and dehydration pathways. These processes are enhanced by prior spirit contact and residual fruit acids in the barrel matrix. Fruit-derived residues and lower pH microenvironments can also favor mild microbial transformations and ester retention/reformation. This is consistent with the high levels of isoamyl acetate and citric acid observed. Together, these produce a sweeter, fruitier, and caramel-like aromatic profile. As with other barrel types, these effects are a result of a combination of pure extractives from the wood, the chemical transformation of released sugars, and residual contributions from spirits previously stored in the barrel.

In contrast, non-barrel-aged barley wines lacked the extractive and oxidative signatures observed in the barrel-treated samples and exhibited only a slight increase in uridine (142.1 mg/L). The absence of wood-derived saccharides, furans, and aldehydes indicates that the distinctive chemical fingerprints of the other groups primarily originate from barrel contact and prior barrel filling history, as well as the oxidative and extractive transformations that occur during maturation.

We created a decision tree model to make our findings as useful and understandable as possible (Figure 6). This method simplifies the complex multivariate separation process, making it easier to understand and use. It transforms the process into a simple, step-by-step diagnostic tool for quickly identifying problems. The tool uses a small group of four key markers: HMF, acetaldehyde, mannose, and tryptophan. The clear demonstration of the driving differences behind the O2PLS-DA separation in the decision tree can be condensed into a few diagnostic chemical features. This provides a practical and reliable method for authenticating barrel reuse history.

Because the beers analyzed in this study were produced by different craft breweries, variations in yeast strain, fermentation temperature, wort composition, and pitching rate were expected, and these variations may have contributed to metabolite variability. Yeast-dependent traits, such as fusel alcohol formation via the Ehrlich pathway, ester production, and organic acid assimilation, influence how samples are positioned in multivariate space. While this process-level heterogeneity introduces background noise, supervised models such as OPLS-DA and O2PLS-DA can effectively separate the systematic variation associated with cereal type, substyle, and barrel aging from brewery- or fermentation-specific variation. Similarly, the CHAID decision tree partitions samples based on the metabolites that discriminate the strongest, thus mitigating the influence of such variability. The consistency of the key discriminant metabolites across groups indicates that these markers remain robust despite differences in craft brewing practices.

This study provides novel molecular insights. However, several limitations should be noted. First, the sample size was modest and included beers from various commercial sources, introducing natural variability in recipes, fermentation, and aging conditions. Therefore, the reported metabolomic trends should be considered illustrative rather than comprehensive. Future studies using standardized brewing trials and expanded datasets will allow for stronger statistical validation and clearer links between chemistry and sensory expression. This will advance our understanding of the relationship between beer’s chemical composition and its sensory characteristics.

4. Conclusions

This study demonstrates the ability of ¹H NMR metabolomics to capture the compositional complexity of barley and other cereal wines. Fifty-five metabolites were analyzed, revealing distinct molecular fingerprints. These fingerprints were associated with three factors: the type of cereal used, stylistic differentiation, and barrel-aging history. Barley samples were characterized by fusel alcohols and phenolic acids, wheat wines contained pyruvate and aromatic amino acids, and oat and rye variants contained maltodextrin, arabinose, and trigonelline. Barley wine sub-styles varied in their ester-driven profiles. English versions were notable for their accumulation of complex sugars and nucleosides. American versions featured adjunct-derived compounds, such as lactose and unassimilated organic acids, which contributed to their unique characteristics.

Barrel aging introduces additional layers of molecular diversity, with each cask type imparting its own unique combination of saccharides, furanic compounds, and nitrogenous metabolites. The decision tree model is based on four diagnostic markers: HMF, acetaldehyde, mannose, and tryptophan. The model demonstrates the potential of using targeted metabolomic fingerprints. These fingerprints can verify barrel provenance and production history.

This work provides compositional insights and establishes a methodological and conceptual basis for applying NMR metabolomics to research on high-gravity brewing. This approach enables robust chemical authentication of beverages and can be used to study sensory-chemical correlations, aging kinetics, and production optimization in complex fermented beverages. By linking raw materials, fermentation strategies, and maturation processes, the study enhances our molecular understanding of how brewing choices impact the quality and characteristics of strong beers. While the dataset clearly shows molecular trends, further analysis using larger, more controlled samples is necessary to confirm and expand these findings.