Skip to main content
NCBI home page
Food Chemistry: X logoLink to Food Chemistry: X
. 2026 Jul 13;38:104205. doi: 10.1016/j.fochx.2026.104205

Divergent effects of roasting on the volatile profiles and metabolomes of Pinus pumila and Pinus koraiensis nuts

Yong Yu a, Qingzhu Nie a, Xiaolong Chen a, Yu Di a, Aiyuan Wang a, Hongwei Yang a,⁎, Kaiming Ren b,⁎
PMCID: PMC13401034  PMID: 42502867

Abstract

To evaluate the effects of roasting and varietal differences on pine nut flavour, an electronic nose, HS-SPME-GC-Orbitrap MS, and non-targeted metabolomics were used to compare the volatile profiles and metabolomes of raw and roasted Pinus pumila (PP) and Pinus koraiensis (PK) nuts. Roasting markedly amplified the metabolic and flavour differences between the two species. In PK, roasting activated pantothenate and CoA biosynthesis, promoting lipid oxidation and the formation of lipid-derived aldehydes that contributed to a pronounced fatty aroma. In contrast, PP exhibited greater metabolic stability, with flavour formation primarily driven by the Maillard reaction, resulting in pyrazine accumulation and balanced nutty notes. 4,5-epoxydec-2(trans)-enal was identified as the principal aroma-active compound. Molecular docking indicated favourable binding of key odourants to the olfactory receptors OR51E1 and OR5K1. These findings elucidate species-specific flavour formation mechanisms and provide a theoretical basis for optimising roasting strategies for different pine nut varieties.

Keywords: Pinus pumila, Pinus koraiensis, Roasting, Flavour, Non-targeted metabolomics, Volatile compounds, Molecular docking

Highlights

  • •

    HS-SPME-GC-Orbitrap MS identified 993 volatiles (25 with OAV > 1).

  • •

    Metabolomics showed PP metabolic stability and PK pantothenate/CoA activation.

  • •

    PK aroma depends on lipid oxidation; PP uses Maillard with high stability.

  • •

    Roasting amplified interspecies metabolic and flavour divergence of two pine nuts.

1. Introduction

Pine nuts are the edible seeds of Pinus spp. (family Pinaceae) that have emerged as high-value commodities in the global market owing to their unique aromatic profiles and exceptional nutritional value (Bolling, Chen, McKay, & Blumberg, 2011). Global production is concentrated in Asia, particularly in China, North Korea, Siberia (Russia), Pakistan, and Afghanistan (Nayab, Valík, & Ačai, 2023). Accumulating evidence indicates that pine nuts possess diverse bioactive properties (e.g. antitumour, anti-inflammatory, and antibacterial activities) and exert preventive and therapeutic effects against chronic disorders, including cardiovascular diseases, diabetes, and obesity (Hao et al., 2025), highlighting their high edible quality and considerable industrial potential.

Pinus koraiensis (PK) is currently the most commercially representative pine nut species (Zhang, Zhang, Wang, & Zhao, 2019), whereas Pinus pumila (PP) is an important pine nut resource in Northeast China. However, inherent metabolic differences between PP and PK result in distinct processing characteristics (e.g. roasting stability, flavour balance, and lipid oxidation resistance), which have not been systematically compared. Existing studies lack a comprehensive analysis of the volatile and non-volatile metabolic changes that occur during roasting, hindering precise processing and high-value utilisation of PP resources.

Roasting is a pivotal thermal processing step in post-harvest nut processing and commercialisation. It directly determines the sensory properties of the final product (colour, aroma, and texture) and triggers a cascade of biochemical reactions, including the generation of volatile organic compounds (VOCs) and metabolic remodelling, making it a critical factor governing the final quality of nut products (Liu et al., 2026). During pine nut roasting, the Maillard reaction, lipid oxidation, and caramelisation collectively shape the sensory characteristics and substantially modify the VOC profile and global metabolome (Adelina, Wang, Zhang, & Zhao, 2021). Owing to differences in endogenous metabolic profiles, different nut varieties exhibit distinct responses to roasting, which largely determine their processing suitability and flavour characteristics—a phenomenon widely documented in peanuts, pistachios, almonds, and other nut varieties (Cialiè Rosso et al., 2021; Güler, Dursun, & Türkmen, 2022; Lakhlifi El Idrissi et al., 2024).

In recent years, multi-omics approaches integrating sensory evaluation, VOC analysis, and targeted and non-targeted metabolomics have emerged as powerful tools for elucidating the molecular mechanisms underlying food flavour formation (Yang et al., 2023). However, to date, no study has systematically compared the effects of roasting on the sensory quality, VOC profiles, and global metabolome of PP and PK nuts. Consequently, the quality evolution patterns, differential flavour compound accumulation, and metabolic response mechanisms of these two pine nut species during thermal treatment remain unclear.

Therefore, this study aimed to (1) compare the volatile profiles, sensory characteristics, and non-targeted metabolomes of raw and roasted PP and PK nuts; (2) reveal the species-specific metabolic pathways and flavour formation mechanisms under roasting; and (3) clarify the differences in processing suitability between the two varieties. This comprehensive analysis elucidated the variety-specific quality differences and the underlying biochemical mechanisms during roasting, providing theoretical support for optimising roasting processes, improving quality control for different pine nut varieties, and facilitating the diversified, high-value development of the pine nut industry.

2. Materials and methods

2.1. Reagents and chemicals

HPLC-grade methanol (CAS No. 67–56-1) was purchased from Fisher Scientific LLC. HPLC-grade acetonitrile (CAS No. 75–05-7) and HPLC-grade formic acid (CAS No. 64–18-7) were obtained from CNW Technologies Co., Ltd. Naphthalene‑d₈ (Cat. No. 013220191) was acquired from damas-beta, and n-pentadecane-d₃₂ (Cat. No. D874357–100 mg) was purchased from Macklin Biochemical Co., Ltd. Mixed n-alkane standards (C₁₀-C₂₅, Cat. No. 1069335HE) were supplied by LGC Standards Ltd. n-Hexacosane (Cat. No. BZRXD-KE) and n-heptacosane (Cat. No. DJQOK-KI) were purchased from Tokyo Chemical Industry Co., Ltd. (TCI). n-Octacosane (Cat. No. 74684–250 mg) and n-nonacosane (Cat. No. BCCF8352) were obtained from Sigma-Aldrich.

2.2. Sample preparation

PP and PK nut samples were collected from the Greater and Lesser Khingan Ranges (Heilongjiang Province, China) in September 2025. After the natural maturation and opening of the pinecones, the seeds were manually separated and stored in a dry environment prior to analysis. Roasting conditions were selected based on a previous study (Adelina et al., 2021). Samples were roasted in an electric oven (Midea, Guangzhou, China; dimensions: 441 × 400 × 366 mm) at 150 °C for 20 min. During roasting, the samples were evenly spread on stainless steel wire mesh trays and placed in the centre of the oven. The roasted PP and PK samples were designated PP_R and PK_R, respectively, whereas the unroasted samples were retained as raw controls.

2.3. Electronic nose analysis

Odour fingerprint profiling was performed using a portable electronic nose (E-nose) system (PEN3, Airsense, Germany) based on metal oxide semiconductor (MOS) sensor technology. The system characterises sample odour profiles by detecting conductivity changes induced by interactions between VOCs and sensor surfaces and is equipped with a 10-MOS sensor array (the target analytes for each sensor are listed in Table S1).

Prior to analysis, the sensors were zero-calibrated to ensure measurement accuracy and reproducibility. Clean air filtered through activated carbon was used as the reference gas for calibration. Sample analysis commenced only after the response signals of all sensors had stabilised at the baseline. For each analysis, 2.0 g of sample was accurately weighed into a dedicated headspace vial, sealed, and equilibrated at 60 °C for 30 min to establish headspace equilibrium of the volatile components. The instrumental parameters were as follows: sensor cleaning time, 120 s; sample preparation time, 5 s; injection flow rate, 400 mL/min; detection time, 120 s; and cleaning flow rate, 400 mL/min. Three biological replicates were performed for each group.

2.4. Headspace solid-phase microextraction coupled with gas chromatography–mass spectrometry analysis

Mixed n-alkane standards (C₁₀-C₂₉) were prepared in n-hexane at a stock concentration of 50 μg/mL and diluted to 10 μg/mL for retention index calibration. An accurately measured 2.0 g aliquot of pine nut sample was transferred into a 20 mL headspace vial. Subsequently, 2.5 μL of mixed internal standards (containing naphthalene-d8, 20 μg/mL, and n-pentadecane-d32, 50 μg/mL) and 4 mL of saturated sodium chloride solution were added, and the vial was immediately sealed. VOCs were extracted using a 120 μm DVB/Carbon WR/PDMS SPME Arrow fibre following incubation at 80 °C for 20 min (500 rpm) and headspace extraction at 80 °C for 10 min (200 rpm). The fibre was desorbed at 250 °C for 5 min and conditioned at 260 °C for 2 min before and after each injection to eliminate carry-over.

Gas chromatographic separation was performed using a Thermo Orbitrap Exploris gas chromatography (GC) system equipped with a VF-WAXms capillary column (25 m × 0.25 mm × 0.2 μm). The GC inlet temperature was maintained at 240 °C and operated in split mode (10,1), with high-purity helium as the carrier gas at a flow rate of 1.0 mL/min and a septum purge flow rate of 3 mL/min. The oven temperature programme started at 40 °C, increased to 120 °C at 8 °C/min, then to 230 °C at 20 °C/min, and was held at 230 °C for 4.5 min. Electron ionisation (EI, 70 eV) was employed, and full-scan acquisition (m/z 35–500, resolution 30,000) was performed at an ion source temperature of 250 °C. Quality control (QC) samples were prepared by pooling equal aliquots of all samples, and three QC samples were randomly inserted into the analytical sequence to assess instrumental stability and reproducibility. Three biological replicates were performed for each group.

2.5. Volatile metabolite identification

Raw GC-mass spectrometry (MS) data were processed using Thermo Compound Discoverer 3.3 SP3 software for peak extraction, alignment, and pre-treatment. Metabolites were annotated by matching the mass spectra against the NIST-2023 library, the GC-Orbitrap Flavour and Fragrances v1.0 library, and an in-house database. False-positive peaks (e.g. noise, column bleeding, and derivatisation reagent peaks) were removed, followed by redundancy elimination and peak merging. A three-dimensional data matrix containing metabolite names, retention times, molecular formulae, peak areas, and identification scores was generated for the subsequent statistical analyses.

2.6. Molecular docking studies

The three-dimensional (3D) structures of the human olfactory receptors OR51E1 (UniProt ID: Q8TCB6) and OR5K1 (UniProt ID: Q8NHB7) were downloaded from the UniProt database (https://www.uniprot.org/). The proteins were pre-processed using the “QuickPrep” module in Molecular Operating Environment (MOE). Dehydration, hydrogenation, and missing residue completion were performed, followed by constrained energy minimisation using the Amber10: EHT force field (RMS gradient <0.1 kcal/mol/Å2). 3D SDF files of the small-molecule compounds were obtained from the PubChem database, and ligand conformations were systematically generated in MOE using the conformation import method. The lowest-energy conformation was retained as the initial structure and subsequently subjected to energy minimisation using the MMFF94x force field.

The docking procedure followed the methods described by El-Demerdash et al. (El-Demerdash et al., 2021). Semi-flexible docking was performed using the “Dock” module in MOE. Binding sites were automatically identified using the “Site Finder” module, and the cavity with the largest volume and highest proportion of hydrophobic amino acids was selected as the active pocket. Docking was performed using a triangle-matcher algorithm to generate 50 initial conformations. These were ranked using the London dG scoring function, and the top 30 conformations were retained for rescoring using the GBVI/WSA dG scoring function. The conformation with the highest score and a reasonable binding mode was selected as the final docking pose. The RMSD_refine values for all conformations were < 2.0 Å, confirming the reliability of the docking results.

2.7. Non-targeted metabolomics analysis by ultra-high-performance liquid chromatography–mass spectrometry

A 100 ± 5 mg aliquot of pine nut sample was transferred into a 2 mL centrifuge tube containing a 6 mm grinding bead and extracted with 400 μL of ice-cold methanol/water (4,1, v/v) containing the internal standard (0.02 mg/mL L-2-chlorophenylalanine). Samples were ground at 50 Hz for 6 min (−10 °C), sonicated at 40 kHz for 30 min (5 °C), incubated at −20 °C for 30 min, and centrifuged at 13,000×g for 15 min (4 °C). The supernatant was collected for ultra-high-performance liquid chromatography–MS (UHPLC-MS) analysis. QC samples were prepared by pooling 20 μL of the supernatant from each sample.

Chromatographic separation was performed on a Waters ACQUITY UPLC BEH C₁₈ column (100 mm × 2.1 mm, 1.7 μm) at 40 °C. Mobile phase A consisted of 2% acetonitrile in water containing 0.1% formic acid, whereas mobile phase B consisted of acetonitrile containing 0.1% formic acid. The injection volume was 2 μL. Mass spectrometric analysis was performed using a Thermo Q Exactive HF-X system equipped with an electrospray ionisation (ESI) source operating in both positive and negative ion modes (m/z 70–1050). The operating parameters were as follows: sheath gas, 50 arb; auxiliary gas, 13 arb; heater temperature, 425 °C; capillary temperature, 325 °C; spray voltage, +3500 V/−3000 V; normalised collision energy, 20%/40%/60%; full MS resolution, 60,000; and MS2 resolution, 7500. QC samples were analysed every 5–15 study samples to monitor system stability and data reliability. Six biological replicates were performed for each group.

2.8. Non-targeted metabolite identification

Raw UHPLC-MS data were imported into Progenesis QI v3.0 software for baseline correction, peak detection, integration, retention time calibration, and peak alignment. Metabolites were identified by matching the MS/MS spectra against the in-house Majorbio plant-specific metabolite database (MJDBPM) using a mass accuracy tolerance of <10 ppm. Identification confidence was evaluated based on secondary spectral matching scores. A standardised data matrix (m/z, retention time, peak intensity, metabolite information, and HMDB/CAS IDs) was exported for bioinformatic analysis.

Data preprocessing was performed on the Majorbio Cloud platform (cloud.majorbio.com). Missing values were filtered using the 80% rule (retaining variables with non-zero values in ≥80% of samples) and imputed using the minimum value in the original matrix. Sum normalisation was applied to minimise variation arising from sample preparation and instrumental analysis, followed by the removal of variables with a QC relative standard deviation (RSD) >30% and log10 transformation to generate the final data matrix.

2.9. Data processing and statistical analysis

Principal component analysis (PCA) and orthogonal partial least squares-discriminant analysis (OPLS-DA) were performed to evaluate the overall metabolic differences. Differential metabolites were screened based on variable importance in projection (VIP) values derived from the OPLS-DA model and Student's t-test, with metabolites satisfying the criteria of VIP > 1 and P < 0.05 considered statistically significant. To control for the false positive rate inherent in high-dimensional metabolomics data, the Benjamini–Hochberg false discovery rate (FDR) correction was additionally applied to the P values from pairwise comparisons, and metabolites with an FDR < 0.05 were considered statistically robust differentiators. For comparisons where FDR correction yielded few or no significant metabolites owing to limited effect sizes (e.g. PP_R vs. PP), nominal P values were reported alongside VIP scores and fold-change values as exploratory indicators, and biological interpretations were based on converging multi-omics evidence rather than on individual statistical thresholds alone.

Differential metabolites were annotated using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database, and pathway enrichment analysis was performed using the Python scipy.stats package. Fisher's exact test was used to identify biological pathways significantly associated with the experimental treatments.

3. Results and discussion

3.1. E-nose analysis of raw and roasted pine nuts

The E-nose enables rapid, non-destructive characterisation of sample odour fingerprints through highly sensitive sensor arrays and has been widely applied to food flavour analysis (Zhou et al., 2025). In this study, a PEN3 E-nose equipped with 10 MOS sensors was used to profile the overall volatile fingerprints of four pine nut samples (raw PP, roasted PP [PP_R], raw PK, and roasted PK [PK_R]). These sensors exhibit selective sensitivity towards aromatic compounds, nitrogen oxides, ammonia derivatives, hydrogen, short-chain alkanes, methyl compounds, sulfides, alcohols/ethers/aldehydes/ketones, organic sulfides, and long-chain alkanes, enabling comprehensive profiling of volatile flavour characteristics.

All four samples exhibited the strongest responses for the W1W (sulfides), W5S (nitrogen oxides), and W1S (methyl-containing compounds) sensors (Fig. 1A), indicating that sulfur-containing compounds, nitrogen-containing heterocycles, and methyl-branched compounds were the major contributors to the volatile profiles of pine nuts. Roasting markedly increased the responses of these three sensors. For PP, W1W response increased from 7.13 to 12.44, W5S from 3.02 to 5.89, and W1S from 2.63 to 4.18. Similarly, for PK, the W1W response increased from 9.07 to 15.00, W5S from 3.23 to 6.07, and W1S from 3.13 to 5.16. In contrast, the W1C response decreased (PP: 0.59 to 0.30; PK: 0.47 to 0.21), consistent with the volatilisation and thermal degradation of thermally unstable terpenoids during roasting (Guo, Na Jom, & Ge, 2019). Notably, raw PK exhibited a higher W1W response (9.07) than that for raw PP (7.13), indicating a richer flavour precursor pool. This advantage was amplified after roasting: PK_R showed a 20.57% higher W1W response than that for PP_R (15.00 vs. 12.44), together with stronger W1S (5.16 vs. 4.18) and W2W responses (3.16 vs. 2.66), suggesting greater accumulation of sulfur- and nitrogen-containing heterocyclic aroma compounds in roasted PK.

Fig. 1.

Fig. 1

Open in a new tab

E-nose profiling of volatile characteristics in raw and roasted PP and PK nuts. (A) Radar plot of electronic nose sensor responses. (B) PCA score plot of electronic nose data. (C) LDA score plot of electronic nose data. (D) Confusion matrix of the LDA classification model.

PCA showed that the first two principal components explained 98.82% of the total variance (PC1 = 97.20%), with the four sample groups clearly separated and their 95% confidence ellipses showing no overlap (Fig. 1B). The loading plot indicated that W2W, W1C, and W1W contributed predominantly to PC1, reflecting roasting-induced changes in organic sulfides and aromatic compounds, whereas W5S, W6S, and W2S primarily contributed to PC2, distinguishing more subtle differences among the samples. Linear discriminant analysis (LDA) achieved 100% classification accuracy (Fig. 1C), demonstrating the strong discriminative capability of the E-nose. Feature contribution analysis identified W3C (20.6%), W1C (14.7%), and W5C (11.2%) as the principal sensors contributing to the discrimination of variety and roasting status. The cross-validation confusion matrix further confirmed the 100% correct classification of all 12 samples (Fig. 1D).

3.2. Overview of VOCs

VOCs are key determinants of food flavour and sensory perception and serve as important indicators of metabolic alterations and quality characteristics (Chen et al., 2025). Headspace solid-phase microextraction coupled with GC–MS (HS-SPME-GC–MS) is a widely used technique for volatile profiling, overcoming the limitations of conventional methods, such as complex sample preparation, organic solvent consumption, and limited sensitivity for trace compounds (Giannetti, Toscani, Rapa, Boccacci Mariani, & Marini, 2025).

A total of 993 VOCs belonging to 15 chemical classes were identified (Supplementary Material 2), with organic heterocyclic compounds (16.31%), esters (15.21%), and terpenes (12.99%) representing the dominant groups and together accounting for more than 44% of the identified VOCs (Fig. 2A). Relative abundance analysis (Fig. 2B) showed that terpenes, acids, hydrocarbons, organic heterocyclic compounds, aldehydes, and esters were the principal contributors to flavour. Roasting induced consistent compositional changes in both pine nut varieties, characterised by significant reductions in terpenes and hydrocarbons and marked increase in acids and organic heterocyclic compounds.

Fig. 2.

Fig. 2

Open in a new tab

Volatile metabolite profiling in raw and roasted PP and PK nuts. (A) Pie chart of volatile compound class distribution. (B) Stacked bar plot of relative abundances of volatile metabolite classes. (C) Correlation matrix heatmap of volatile metabolites. (D) PCA score plot of volatile metabolite profiles.

The reduction in terpenoids is likely attributable to their high thermal volatility (Adelina et al., 2021). Long-chain hydrocarbons undergo lipid oxidation during roasting, generating ketones and aldehydes as secondary oxidation products (Shahidi & Hossain, 2022). The substantial increase in organic heterocyclic compounds, predominantly pyrazines and furans, is consistent with enhanced Maillard reactions, the principal pathway responsible for generating roasted nutty aromas (Fan, 2005; Scalone, Cucu, De Kimpe, & De Meulenaer, 2015).

Notably, the magnitude of roasting-induced VOC changes differed considerably between the two varieties. PK exhibited greater depletion of terpenes and hydrocarbons together with increased accumulation of lipid oxidation-derived aldehydes, suggesting greater susceptibility to lipid oxidation during thermal processing. In contrast, PP exhibited greater metabolic stability, characterised by moderate increases in Maillard reaction products and comparatively limited losses of terpenoids. This genotype-dependent thermal response to roasting is consistent with previous observations in other nut species (Buthelezi, Tesfay, & Magwaza, 2021; Kulik & Waszkiewicz-Robak, 2025).

Intragroup correlation analysis demonstrated strong correlations within each sample group, whereas negative correlations were observed between different varieties and between raw and roasted samples (Fig. 2C), confirming the substantial influence of both variety and roasting on volatile composition. PCA clearly separated the four groups (Fig. 2D), with varietal differences (PC1 = 39.90%) representing the major source of VOC variation, while roasting induced consistent metabolic shifts along PC2 (19.60%).

3.3. Identification of differential VOCs

To further evaluate the effects of variety and roasting on volatile profiles, supervised OPLS-DA was performed. Model performance was assessed using R2X, R2Y, and Q2 values and validated using a 200-permutation test to minimise the risk of overfitting (Kalogiouri, Aalizadeh, Dasenaki, & Thomaidis, 2020) (Fig. S1). The OPLS-DA score plots showed clear separation among the four sample groups in all pairwise comparisons (Fig. 3A). Notably, the stronger performance of the PK_R vs. PK model (R2Y = 0.997, Q2 = 0.951) than of the PP_R vs. PP model (R2Y = 0.999, Q2 = 0.873) suggests that PK undergoes more extensive metabolic changes during roasting than PP. Furthermore, separation between the two species was most pronounced after roasting, with the PP_R vs. PK_R model exhibiting the highest predictive ability (Q2 = 0.981).

Fig. 3.

Fig. 3

Open in a new tab

OPLS-DA-based differential analysis of volatile metabolites in pine nut kernels. (A) OPLS-DA score plots for pairwise group comparisons. VIP plot and corresponding heatmap of differential volatile metabolites: (B) PP vs. PP_R; (C) PK vs. PK_R; (D) PP vs. PK; (E) PP_R vs. PK_R.

Hierarchical cluster analysis of the top 30 differential volatile metabolites (VIP > 1) from each comparison was performed to visualise their accumulation patterns. In the comparison between raw PP and PK, the balanced distribution of up- and downregulated metabolites was primarily associated with terpenoids and aromatic compounds, and these differences became more pronounced following roasting (Fig. 3B).

Roasting of PP resulted in significant increases in characteristic Maillard reaction products, including pyrazines (e.g. 2,6-dimethylpyrazine and 2,5-dimethylpyrazine), furfural, and benzaldehyde, with almost no downregulated metabolites detected (Fig. 3C). This finding further supports the relatively high metabolic stability of PP during roasting. In contrast, PK exhibited a more complex metabolic response (Fig. 3D). Although pyrazines also increased, a substantially greater accumulation of lipid oxidation-derived aldehydes (e.g. octanal, nonanal, and hexanal), together with decreases in several terpenoids, indicated more extensive metabolic remodelling.

Comparison of the roasted samples (PP_R vs. PK_R) revealed distinct flavour formation pathways (Fig. 3E). The enrichment of pyrazines and furfural in PP_R suggests that flavour development was primarily driven by the Maillard reaction between amino acids and reducing sugars (Bi et al., 2022; Ouyang et al., 2024). Conversely, the enrichment of terpenoids and lipid oxidation products in PK_R suggests a greater contribution of lipid oxidation to flavour formation, resulting in a characteristic lipid-derived roasted aroma. The development of aroma in thermally processed nuts is generally attributed to the combined effects of lipid oxidation and the Maillard reaction, with aldehydes such as nonanal and hexanal representing characteristic products of lipid oxidation in polyunsaturated fatty acid-rich nuts.

Overall, the differential VOC analysis revealed two distinct flavour formation patterns: PP primarily followed a stable, Maillard reaction-driven pathway, whereas PK underwent more extensive metabolic remodelling characterised by lipid oxidation. The differentially accumulated metabolites, including pyrazines, aldehydes, terpenes, and furans, provide a robust foundation for the subsequent relative odour activity value (ROAV) and multi-omics correlation analyses.

3.4. Identification of odour-active compounds by odour activity value and ROAV

To quantitatively evaluate the contribution of individual volatile compounds to the overall aroma of pine nuts, odour activity values (OAVs) were calculated based on their concentrations and odour thresholds (Xie et al., 2025). ROAVs were subsequently calculated to compare the relative contributions of individual compounds across the different samples (Zhu, Chen, Chen, Chen, & Deng, 2020). By integrating the OAV and ROAV data with the VIP values obtained from the OPLS-DA models, key flavour markers with both high sensory contributions and statistical significance were identified (Supplementary Material 3).

Among the 25 volatile compounds with OAV >1, 4,5-epoxydec-2(trans)-enal consistently exhibited the highest OAV in all four sample groups (PP: 1515, PP_R: 2710, PK: 296, and PK_R: 4364), indicating that it is the principal contributor to the characteristic fatty aroma of pine nuts. Accordingly, its ROAV was set to 100 as the reference value. Although this compound exhibited VIP values >1 in some comparisons, its relatively small variation compared with its consistently high OAV suggests that it contributes to the stable aroma background rather than to flavour discrimination. Similarly, compounds such as valeric acid, 2-methylisoborneol, and limonene exhibited OAVs >1 but consistently low VIP values (<1), indicating that they contribute to the characteristic background aroma of pine nuts irrespective of variety or roasting treatment.

In contrast, the ROAVs of several key compounds changed markedly after roasting in a cultivar-dependent manner (Fig. 4A). Following roasting of PP, the ROAV of the Strecker aldehyde isovaleraldehyde increased from 0.22 to 0.58 (VIP = 1.77), together with that of caproic acid (5.87 to 6.51, VIP = 19.32), both of which are typical Maillard reaction products (Smit, Engels, & Smit, 2009). In PK, although the ROAV of 1-octen-3-one decreased (23.26 to 4.43), its absolute OAV increased from 64.07 to 192.11 (VIP = 1.75). Concurrently, the OAVs of lipid oxidation markers, including hexanal and nonanal, also increased substantially. Notably, the OAVs of 1-octen-3-one and 4,5-epoxydec-2(trans)-enal in PK_R were 3.7-fold and 1.6-fold higher, respectively, than those in PP_R, suggesting that arachidonic acid (AA) autoxidation was more pronounced during roasting of PK than of PP. Benzaldehyde emerged as a consistent roasting marker, exhibiting high VIP values across all roasting-related comparisons.

Fig. 4.

Fig. 4

Open in a new tab

Odour activity and sensory profile analysis of volatile compounds in pine nut kernels. (A) Scatter plot of ROAV distribution across sample groups. (B) Radar plot of odour profile characteristics.

Comparison of the raw samples showed that PK possessed a more pronounced mushroom-like green aroma than that of PP, primarily owing to the higher ROAV of 1-octen-3-one (23.26), together with subtle citrus and fatty notes contributed by compounds such as decanal (ROAV 1.47). After roasting, PP_R exhibited higher ROAVs for isovaleraldehyde and caproic acid, whereas PK_R retained substantially higher levels of 1-octen-3-one.

Overall, integration of the OAV, ROAV, and VIP analyses identified isovaleraldehyde, caproic acid, 1-octen-3-one, benzaldehyde, and hexanal as the principal discriminatory flavour markers among the four pine nut sample groups.

3.5. Odour profiles of raw and roasted PP and PK

To relate the chemical composition to sensory perception, an odour radar plot was constructed based on the sensory attributes of the identified volatile metabolites (Fig. 4B). Among the raw samples, PK exhibited stronger woody, camphor, and citrus notes, whereas PP was characterised by more pronounced sweet, fresh, and fruity notes, consistent with the E-nose responses. Roasting significantly enhanced buttery and spicy aroma notes in both species. PP_R exhibited a better balance between buttery and fresh notes, whereas PK_R displayed a more intense buttery aroma accompanied by stronger lingering woody notes. Sankey diagrams further illustrated the relationships between the key differential volatile compounds and their corresponding sensory attributes (Fig. S2). These diagrams highlighted the contributions of characteristic compounds, including 1-octen-3-one and hexanal in PK_R and pyrazines in PP_R, to the distinctive flavour characteristics of roasted pine nuts. They also provide an intuitive visualisation of how cultivar differences and roasting influence the relationships between volatile composition and odour perception, thereby further explaining the sensory divergence among the four pine nut samples.

3.6. Molecular docking of key odour-active compounds with olfactory receptors

Based on two selection criteria, the five core flavour markers identified above were selected for molecular docking studies: (1) functional relevance to well-characterised olfactory receptors and (2) structural complementarity to the two major flavour formation pathways in pine nuts. Although 4,5-epoxydec-2(trans)-enal was classified as a stable flavour background compound rather than a discriminatory marker in the VIP analysis, it was additionally included because it exhibited the strongest binding affinity towards both receptors in the preliminary docking analysis. The final ligand set comprised 4,5-epoxydec-2(trans)-enal (highest OAV across all groups; fatty aroma foundation), 1-octen-3-one (characteristic mushroom/metallic odourant of PK; OAV increased from 64 to 192), caproic acid (VIP = 19.32; the most sensitive indicator of roasted PP flavour), and hexanal (a lipid oxidation marker; VIP = 5.14 during PK roasting). Isovaleraldehyde and benzaldehyde were excluded because of their relatively weak binding affinities in the preliminary docking analysis.

Olfactory receptors (ORs) belong to the G protein-coupled receptor family and play essential roles in odour perception (Yang et al., 2025). OR51E1 preferentially recognises short-chain fatty acids associated with sour and acidic odours, whereas OR5K1 responds to a broader range of almond-like, floral, and nutty aroma compounds (Wang et al., 2025). Together, these receptors provide complementary recognition of odour molecules generated through both the Maillard reaction and lipid oxidation.

As shown in Table S2, although the amino acid residues involved in ligand binding differed among the compounds, all binding energies were negative, indicating spontaneous ligand–receptor interactions. Binding energy reflects the strength of intermolecular interactions, with lower binding energies corresponding to stronger binding affinities and greater structural stability (Sun et al., 2024). Molecular docking analysis showed that OR5K1 and OR51E1 exhibited similar ligand selectivity. Among the tested compounds, 4,5-epoxydec-2(trans)-enal displayed the strongest binding affinity towards both receptors, with binding free energies of −5.09 kcal/mol for OR5K1 and − 4.76 kcal/mol for OR51E1. 1-Octen-3-one exhibited the second strongest binding affinity, with corresponding binding free energies of −4.67 kcal/mol and − 4.58 kcal/mol, respectively. These findings further support the key contribution of 4,5-epoxydec-2(trans)-enal to pine nut flavour perception.

Binding mode analysis showed that the interaction between 4,5-epoxydec-2(trans)-enal and OR5K1 was stabilised by a hydrogen bond with Ser243 together with extensive hydrophobic and van der Waals interactions involving residues such as Ile289 and Met59. Binding to OR51E1 involved a hydrogen bond between the aldehyde oxygen atom and the side chain of Arg265, with the ligand occupying a hydrophobic pocket formed by Leu183, Ile206, Leu203, Phe257, and Leu260 (Fig. 5A). 1-Octen-3-one interacted with OR51E1 through a hydrogen bond with His108 and surrounding hydrophobic residues, including Val205, Phe257, Met159, Ile206, Ile210, and Leu162. In OR5K1, its binding was mediated by an H–π interaction with Tyr259 together with hydrophobic contacts involving Leu264, Phe200, and Leu199 (Fig. 5B). Caproic acid formed two hydrogen bonds with OR5K1, through Arg122 and Glu296, while also interacting with the hydrophobic residues Met59 and Leu63. Although the binding free energy of caproic acid towards OR51E1 reached −4.20 kcal/mol, no hydrogen bonds were observed, and binding appeared to be primarily stabilised by hydrophobic interactions involving surrounding residues such as Phe88, Ile18, Ile87, and Phe28 (Fig. 5C). Hexanal interacted with OR51E1 through a hydrogen bond with Arg125, resulting in a binding free energy of −4.10 kcal/mol. In OR5K1, binding involved a hydrogen bond with Phe17 together with hydrophobic interactions involving Leu14, Phe85, and Pro21, with a binding free energy of −4.17 kcal/mol (Fig. 5D).

Fig. 5.

Fig. 5

Open in a new tab

Molecular docking analysis of key volatile compounds with olfactory receptors OR5K1 and OR51E1.

Collectively, these results suggest that the characteristic aroma compounds of pine nuts interact with olfactory receptors through hydrogen bonding and hydrophobic interactions, providing a plausible molecular basis for the diverse aroma profiles of pine nuts. However, molecular docking is based on static receptor conformations and does not account for receptor dynamics, multi-receptor cooperativity, or the physiological environment of the nasal cavity. Therefore, these findings should be further validated using in vitro functional assays and sensory evaluation.

3.7. Non-targeted metabolomic profiling of raw and roasted PP and PK

Plant-derived foods contain a diverse array of non-volatile flavour precursors, including amino acids, sugars, and fatty acids, whose composition and transformation during processing profoundly influence the final flavour (Xu, Zang, Sun, & Li, 2023). A non-targeted metabolomics approach using UHPLC coupled with Q Exactive HF-X MS (UHPLC-Q Exactive MS) was employed to comprehensively evaluate the metabolic changes induced by roasting.

A total of 2068 and 10,024 peaks were detected in the positive and negative ionisation modes, respectively. Following alignment with the reference databases and a series of preprocessing steps, including filtering, missing value imputation, normalisation, and log transformation, 405 compounds were putatively identified in the positive ion mode and 1486 in the negative ion mode (Supplementary Material 4). All annotated metabolites were subsequently classified into superclasses according to the HMDB database (Fig. 6A). Lipids and lipid-like molecules represented the predominant metabolite class, comprising 393 compounds (33.03% of all identified metabolites). The remaining major classes included phenylpropanoids and polyketides (14.62%), organic oxygen compounds (14.37%), organic acids and derivatives (12.86%), organoheterocyclic compounds (10.59%), and benzenoids (7.90%). All other metabolite classes, including alkaloids, hydrocarbons, and organosulfur compounds, each accounted for less than 3% of the annotated metabolites.

Fig. 6.

Fig. 6

Open in a new tab

Metabolite profiling and differential analysis of pine nuts. (A) Pie chart of metabolite superclass distribution. (B) PCA score plot of metabolite profiles. Heatmap and corresponding VIP plot of differential metabolites: (C) PP vs. PK. (D) PK vs. PK_R. (E) PP_R vs. PK_R. (F) Radar plot illustrating the relative abundances of 11 key differential metabolites between PP_R and PP groups.

To investigate the global metabolic differences among the four sample groups, unsupervised PCA was performed using the preprocessed metabolomics data (Fig. 6B). The four groups were clearly separated along PC1 (39.20% of the explained variance), with PP and PP_R clustering on the negative side and PK and PK_R clustering on the positive side, indicating that species is the primary driver of metabolome-wide variation. Additionally, separation along PC2 (7.22% of the explained variance) distinguished the raw and roasted samples within each species, suggesting that roasting exerted a consistent influence on the pine nut metabolome. This separation pattern was highly consistent with the volatilomics and E-nose analyses, further supporting the conclusion that both species and roasting are major drivers of metabolic and flavour differentiation.

3.8. Differential metabolite screening

OPLS-DA was performed to systematically evaluate the effects of variety and roasting on the non-volatile metabolome of pine nuts (Fig. S3). The top 30 metabolites with VIP > 1 were selected to generate hierarchical clustering heatmaps, thereby visualising their distribution patterns across the sample groups. In the comparison between raw PP and PK, the two groups were completely separated along the first predictive component, with model parameters of R2Y = 0.999 and Q2 = 0.963, indicating excellent model performance without evidence of overfitting. The differential metabolites were predominantly amino acid derivatives, flavonoids, and terpenoids (Fig. 6C).

For the PK_R vs. PK comparison, the samples were clearly separated along the first component, with non-overlapping 95% confidence ellipses. Permutation testing yielded R2Y = 0.994 and Q2 = 0.584, confirming satisfactory model fitness and predictive performance. The top 30 VIP-ranked metabolites were mainly classified as lipids, carbohydrates, and amino acid derivatives (Fig. 6D). Similarly, the comparison between PP_R and PK_R showed clear separation along the first predictive component, with R2Y = 1.000 and Q2 = 0.982 (P < 0.05), indicating excellent model performance. The differential metabolites were predominantly secondary glycosides, flavonoids, and terpenoids (Fig. 6E).

In contrast, roasting induced only limited metabolic changes in PP. In the PP_R vs. PP comparison, only 11 metabolites showed modest increases (FC range: 1.24–1.54, nominal P < 0.05), and none remained significant following FDR correction (Fig. 6F). This relatively small response contrasted markedly with the extensive metabolic remodelling observed in PK after roasting, in which hundreds of metabolites were differentially abundant. The PCA score plot (Fig. 6B) further corroborated this observation, showing substantial overlap between PP_R and PP, whereas PK_R and PK were clearly resolved along PC2. Although the corresponding OPLS-DA model yielded a negative Q2 value (−0.0384), indicating that the model is unsuitable for biomarker screening because of the extremely limited intergroup variation, the combined evidence from PCA, VOC profiling, and the enrichment of antioxidant pathways consistently suggests that PP exhibits greater metabolic stability during roasting. This enhanced metabolic stability may contribute to its balanced Maillard reaction-driven flavour profile and reduced susceptibility to off-flavour formation following roasting (Huang, Tang, Li, Chen, & Jiang, 2026).

To systematically evaluate the contribution of flavour precursors to species differences and roasting responses, differential metabolites with VIP > 1 from the three robust pairwise comparisons (PK_R vs. PK, PP vs. PK, and PP_R vs. PK_R) were classified into three principal flavour precursor groups: amino acids and their derivatives, lipids, and carbohydrates and their derivatives (Supplementary Material 5).

3.8.1. Amino acids and derivatives

This metabolite class showed the greatest enrichment in the comparison between PP and PK. Among the top 30 VIP-ranked metabolites, peptides and amino acid derivatives, including Tyr-Ser-Arg (VIP = 3.00) and Arg-Gly-Asp (VIP = 2.30), were consistently more abundant in PP. These results suggest that raw PP contains a larger pool of free amino acid-related metabolites, potentially providing a richer precursor supply for Maillard reactions during roasting (Luo et al., 2020).

In the PK_R vs. PK comparison, amino acid derivatives such as N-acetyl-L-glutamate 5-semialdehyde (VIP = 3.12) were markedly upregulated, suggesting enhanced amino acid metabolism during roasting. In contrast, only one amino acid derivative exhibited a marginal change in the PP_R vs. PP comparison (oxfenicine; FC = 1.36), and this difference was non-significant after FDR correction, further supporting the relatively high stability of amino acid metabolism in PP during roasting.

3.8.2. Lipids

Lipid metabolites exhibited the greatest diversity and the most pronounced differential abundance in the PK_R vs. PK comparison. Among the top 30 VIP-ranked metabolites, (Z)-10-hydroxy-8-decene-4,6-diynoic acid (VIP = 3.99), mannosylglucosylglyceric acid (VIP = 3.92), and several oxidised lipids were significantly upregulated, reflecting extensive lipid oxidation and degradation during PK roasting (Pedron, Jaouhari, & Bordiga, 2025). In the PP_R vs. PK_R comparison, glycolipids, including a 6-feruloylglucose derivative (VIP = 2.39), were more abundant in PP_R, indicating fundamental differences in lipid metabolism between PP and PK following roasting. In contrast, differences between the raw PP and PK samples were relatively modest, with only a few glycerides, such as 1,2-dibutyrin (VIP = 2.55), showing higher abundance in PP.

3.8.3. Carbohydrates and derivatives

Carbohydrates and their derivatives accounted for a substantial proportion of the differential metabolites identified across the three comparisons. In the comparison between PP and PK, glycosides such as swertiamarin (VIP = 2.83) and fortuneanoside J (VIP = 2.38) were more abundant in PP, representing important contributors to the inherent metabolomic differences between the two species. In the PK_R vs. PK comparison, changes in the abundances of furaneol glycoside (VIP = 3.26), acetylmaltose (VIP = 2.85), and d-ribose (VIP = 2.75; downregulated) suggest that reducing sugars were consumed as substrates during the Maillard reaction (Fan et al., 2023), accompanied by the hydrolysis or transformation of certain glycosides. In the PP_R vs. PK_R comparison, differences in carbohydrate derivatives were particularly pronounced. Most of these metabolites, including swertiamarin and caffeic acid 3-glucoside, were more abundant in PP_R. These findings suggest that PP_R has a greater capacity to retain or generate secondary glycosidic metabolites after roasting, potentially owing to its higher oxidative stability and a comparatively milder Maillard reaction environment (Oh et al., 2026).

3.9. KEGG pathway enrichment analysis

To elucidate the pathway-level mechanisms underlying species-specific metabolic responses to roasting, differentially expressed metabolites from the three comparison groups (PP vs. PK, PK_R vs. PK, and PP_R vs. PK_R) were mapped to the KEGG database for pathway enrichment analysis. The enrichment results were visualised using topological bubble plots, in which the x-axis represents the pathway Impact Value; the y-axis represents -log₁₀(P-value); bubble colour corresponds to P value; and bubble size reflects pathway impact. These results were further validated using multidimensional enrichment concentric circle plots, in which concentric layers represent, from the innermost to the outermost, the direction of metabolite regulation, pathway significance, number of enriched metabolites, and KEGG pathway identifiers. Together, these visualisations provide complementary information on pathway significance, biological impact, and metabolite abundance.

In the comparison between raw samples, differential metabolites were significantly enriched in pathways related to phenylpropanoid biosynthesis, flavonoid biosynthesis, arginine biosynthesis, and branched-chain amino acid (valine/leucine/isoleucine) biosynthesis (Fig. 7A). Among these, phenylpropanoid biosynthesis exhibited the highest -log₁₀(P-value) and pathway impact score, indicating that it is the principal pathway distinguishing the metabolic profiles of the two species. Multidimensional enrichment analysis further showed increased abundance of flavonoid and phenolic glycoside antioxidant metabolites associated with this pathway in PP (Fig. 7D), suggesting constitutively higher activity of the phenylpropanoid and flavonoid pathways. These pathways are well recognised as antioxidant defence systems that scavenge free radicals and mitigate oxidative stress in plants (Grace & Logan, 2000; Pineda-Hidalgo et al., 2025).

Fig. 7.

Fig. 7

Open in a new tab

KEGG pathway enrichment analysis of differential metabolites in pine nuts. KEGG topology analysis plot: (A) PP vs. PK. (B) PK vs. PK_R. (C) PP_R vs. PK_R. Circular enrichment diagram of differential metabolite pathways: (D) PP vs. PK. (E) PK vs. PK_R. (F) PP_R vs. PK_R.

In the PK_R vs. PK comparison, differential metabolites were primarily enriched in pantothenate and CoA biosynthesis, nucleotide metabolism, and phenylalanine metabolism (Fig. 7B). Pantothenate and CoA biosynthesis exhibited the highest pathway impact, indicating a central role in the metabolic response to roasting. Pantothenate is an essential component of coenzyme A and plays a key role in acyl transfer during fatty acid oxidation, fatty acid synthesis, and the integration of amino acid carbon skeletons into energy metabolism (Naquet, Kerr, Vickers, & Leonardi, 2020). Enrichment of this pathway therefore suggests coordinated activation of lipid and energy metabolism during roasting. The enrichment circle plot further showed that metabolites associated with pantothenate and CoA biosynthesis, together with other lipid metabolism-related metabolites, were predominantly upregulated in PK_R, consistent with the elevated levels of lipid oxidation-derived aldehydes (e.g., hexanal, nonanal, and 1-octen-3-one) identified by volatile metabolite analysis (Fig. 7E). Similar enrichment of lipid metabolism pathways has also been reported during peanut drying (Deng et al., 2025).

Following roasting, phenylpropanoid biosynthesis remained the principal enriched pathway differentiating the two species, while arginine biosynthesis and stilbene secondary metabolism were also significantly enriched (Fig. 7C). Enrichment network analysis further showed that antioxidant-related phenylpropanoid and flavonoid pathways remained more active in PP_R, whereas PK_R was characterised by enhanced lipid metabolism (Fig. 7F). These findings indicate that roasting did not eliminate the inherent metabolic differences between the two species but instead accentuated their metabolic and flavour divergence through the preferential activation of species-specific pathways. Specifically, PP maintained greater activity of phenylpropanoid- and Maillard-related pathways, whereas PK exhibited stronger activation of lipid oxidation pathways. This interpretation is supported by the clear separation of the roasted samples in the E-nose analysis together with the enrichment of pyrazines in PP_R and aldehydes in PK_R identified by volatilomics.

4. Conclusion

This study systematically elucidated the distinct effects of roasting on the volatile profiles and metabolomes of PP and PK pine nuts. Among the identified volatile compounds, 4,5-epoxydec-2(trans)-enal was the principal aroma-active compound in both varieties. Roasting not only preserved but also amplified the metabolic and flavour differences between the two species. In PK, flavour formation was characterised predominantly by lipid oxidation associated with pantothenate and CoA biosynthesis, resulting in pronounced fatty aroma characteristics. In contrast, PP exhibited greater metabolic stability, which may be associated with constitutively activated phenylpropanoid and flavonoid antioxidant pathways that potentially limit lipid oxidation while favouring a balanced, Maillard reaction-driven flavour profile. Molecular docking further indicated that the key odour-active compounds can bind spontaneously to the olfactory receptors OR51E1 and OR5K1 through hydrogen bonding and hydrophobic interactions, providing a molecular basis for their contribution to the characteristic aroma of pine nuts. Overall, these findings highlight the importance of interactions between variety and roasting in shaping pine nut flavour quality and provide a theoretical basis for developing roasting strategies tailored to the characteristics of PP and PK.

CRediT authorship contribution statement

Yong Yu: Writing – original draft, Data curation. Qingzhu Nie: Conceptualization. Xiaolong Chen: Methodology. Yu Di: Resources. Aiyuan Wang: Investigation. Hongwei Yang: Writing – review & editing, Supervision. Kaiming Ren: Writing – review & editing, Supervision, Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgement

This work was supported by the Joint Program of Science and Technology Program of Liaoning Province (2024JH2/102600324).

Footnotes

Appendix A

Supplementary data to this article can be found online at https://doi.org/10.1016/j.fochx.2026.104205.

Contributor Information

Hongwei Yang, Email: yanghw@sj-hospital.org.

Kaiming Ren, Email: renkm@sj-hospital.org.

Appendix A. Supplementary data

Supplementary material 1: Supplementary Figures S1–S3 and Tables S1–S2
mmc1.docx (346.8KB, docx)
Supplementary material 2: Summary of Identified volatile organic compounds
mmc2.xlsx (173.7KB, xlsx)
Supplementary material 3: Summary of odor-active compounds
mmc3.xlsx (13.2KB, xlsx)
Supplementary material 4: Summary of Identified non-volatile organic compounds
mmc4.xlsx (312KB, xlsx)
Supplementary material 5: Significantly differential metabolites from pairwise comparisons
mmc5.xlsx (150.9KB, xlsx)

Data availability

Data will be made available on request.

References

  1. Adelina N.M., Wang H., Zhang L., Zhao Y. Comparative analysis of volatile profiles in two grafted pine nuts by headspace-SPME/GC–MS and electronic nose as responses to different roasting conditions. Food Research International. 2021;140 doi: 10.1016/j.foodres.2020.110026. [DOI] [PubMed] [Google Scholar]
  2. Bi S., Niu X., Yang F., Xu Y., Dai Y., Liu Y., Zhou Q. Roasting pretreatment of walnut (Juglans regia L.) kernels: Improvement of the oil flavor profile and correlation with the chemical composition. Food & Function. 2022;13(21):10956–10969. doi: 10.1039/d2fo01990f. [DOI] [PubMed] [Google Scholar]
  3. Bolling B.W., Chen C.Y.O., McKay D.L., Blumberg J.B. Tree nut phytochemicals: Composition, antioxidant capacity, bioactivity, impact factors. A systematic review of almonds, Brazils, cashews, hazelnuts, macadamias, pecans, pine nuts, pistachios and walnuts. Nutrition Research Reviews. 2011;24(2):244–275. doi: 10.1017/S095442241100014X. [DOI] [PubMed] [Google Scholar]
  4. Buthelezi N.M.D., Tesfay S.Z., Magwaza L.S. Influence of roasting on antioxidants, fatty acids, sensory properties and oxidative stability of macadamia nuts. Scientia Horticulturae. 2021;278 doi: 10.1016/j.scienta.2020.109850. [DOI] [Google Scholar]
  5. Chen Y., An H., Huang Y., Jiang Y., Ying J., Wang S.…Huang J. Characterization of key aroma components of jasmine tea with diverse aroma types: Focus on jasmine tea scented with multipetal jasmine. Food Chemistry: X. 2025;29 doi: 10.1016/j.fochx.2025.102734. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Cialiè Rosso M., Stilo F., Bicchi C., Charron M., Rosso G., Menta R.…Cordero C. Combined untargeted and targeted fingerprinting by comprehensive two-dimensional gas chromatography to track compositional changes on hazelnut primary metabolome during roasting. 2021;11(2) https://www.mdpi.com/2076-3417/11/2/525 [Google Scholar]
  7. Deng J., Hou M., Cui S., Liu Y., Li X., Liu L. Integrative analysis of transcriptome and metabolome reveals molecular mechanisms of dynamic change of storage substances during dehydration and drying process in peanuts (Arachis hypogaea L.) 2025;16 doi: 10.3389/fpls.2025.1567059. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. El-Demerdash A., Al-Karmalawy A.A., Abdel-Aziz T.M., Elhady S.S., Darwish K.M., Hassan A.H.E. Investigating the structure-activity relationship of marine natural polyketides as promising SARS-CoV-2 main protease inhibitors. RSC Advances. 2021;11(50):31339–31363. doi: 10.1039/d1ra05817g. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Fan X. Formation of furan from carbohydrates and ascorbic acid following exposure to ionizing radiation and thermal processing. Journal of Agricultural and Food Chemistry. 2005;53(20):7826–7831. doi: 10.1021/jf051135x. [DOI] [PubMed] [Google Scholar]
  10. Fan Y., Yu M., Li D., Zhao G., Zhang M., Wang Z.…Zhou D. Effects of non-enzymatic browning and lipid oxidation on color of ready-to-eat abalone during accelerated storage and its control. 2023;12(7) doi: 10.3390/foods12071514. https://www.mdpi.com/2304-8158/12/7/1514 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Giannetti V., Toscani G., Rapa M., Boccacci Mariani M., Marini F. Volatile profile analysis of pasta using HS-SPME/GC–MS and chemometric modelling for process and quality marker identification. Food Research International. 2025;222 doi: 10.1016/j.foodres.2025.117692. [DOI] [PubMed] [Google Scholar]
  12. Grace S.C., Logan B.A. Energy dissipation and radical scavenging by the plant phenylpropanoid pathway. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences. 2000;355(1402):1499–1510. doi: 10.1098/rstb.2000.0710. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Güler Z., Dursun A., Türkmen D. Comparison of pistachio (Pistacia vera L.) kernel volatile compounds of “Siirt”, “Uzun” and “Kirmizi” grown in Turkey: Effect of traditional and industrial in-shell roasting. Journal of Food Process Engineering. 2022;45(2) doi: 10.1111/jfpe.13957. [DOI] [Google Scholar]
  14. Guo S., Na Jom K., Ge Y. Influence of roasting condition on flavor profile of sunflower seeds: A flavoromics approach. Scientific Reports. 2019;9(1):11295. doi: 10.1038/s41598-019-47811-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Hao J., Hou D., Yu W., Zhang H., Guo Q., Zhang H.…Li Y. Metabolomic and transcriptomic analysis of the synthesis process of unsaturated fatty acids in Korean pine seed kernels. Food Chemistry. 2025;481 doi: 10.1016/j.foodchem.2025.143895. [DOI] [PubMed] [Google Scholar]
  16. Huang J.-B., Tang Y.-H., Li J., Chen J.-R., Jiang P.-F. LC-MS-based non-targeted metabolomics analysis of sea urchin during frozen storage under different packaging methods. Food Chemistry. 2026;505 doi: 10.1016/j.foodchem.2026.148132. [DOI] [PubMed] [Google Scholar]
  17. Kalogiouri N.P., Aalizadeh R., Dasenaki M.E., Thomaidis N.S. Application of high resolution mass spectrometric methods coupled with chemometric techniques in olive oil authenticity studies - a review. Analytica Chimica Acta. 2020;1134:150–173. doi: 10.1016/j.aca.2020.07.029. [DOI] [PubMed] [Google Scholar]
  18. Kulik K., Waszkiewicz-Robak B. The effect of roasting on the health-promoting components of nuts determined on the basis of fatty acids, polyphenol compounds, and antioxidant capacity. Molecules. 2025;30(23) doi: 10.3390/molecules30234594. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Lakhlifi El Idrissi Z., Oubihi A., Ibourki M., El Youssfi M., Gharby S., El Guezzane C.…Tabyaoui M. Peanut (Arachis hypogaea L.) flour and oilcake flour: Exploring the influence of roasting and varietal differences on proximal composition, elemental profiling, antimicrobial and antioxidant properties. Food Chemistry: X. 2024;24 doi: 10.1016/j.fochx.2024.101791. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Liu Y., Huang Y., Hu H., Chen Y., Li C., Yu Q., Wang Y. Thermal generation of acrylamide, 5-HMF, and AGEs in roasted cashews: Correlation with aroma-active volatiles and quality evaluation. Food Chemistry. 2026;510 doi: 10.1016/j.foodchem.2026.148693. [DOI] [PubMed] [Google Scholar]
  21. Luo W., Xue H., Xiong C., Li J., Tu Y., Zhao Y. Effects of temperature on quality of preserved eggs during storage. Poultry Science. 2020;99(6):3144–3157. doi: 10.1016/j.psj.2020.01.020. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Naquet P., Kerr E.W., Vickers S.D., Leonardi R. Regulation of coenzyme a levels by degradation: The “ins and outs”. Progress in Lipid Research. 2020;78 doi: 10.1016/j.plipres.2020.101028. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Nayab A.K., Valík Ľ., Ačai P. Sorption isotherm modelling of chilgoza pine nuts (Pinus gerardiana) Lwt. 2023;174 doi: 10.1016/j.lwt.2023.114437. [DOI] [Google Scholar]
  24. Oh S.C., Lee S.-J., Ding K., Shen J., Huang C., Kang S.N.…Shim Y.Y. Kimchi fermentation-driven detoxification of flaxseed: Impact on physicochemical quality and antioxidant potential. 2026;15(4) doi: 10.3390/foods15040632. https://www.mdpi.com/2304-8158/15/4/632 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Ouyang H., Xu W., Peng W., Zhang R., Yu X., Jin L.…Li S. The potential formation mechanism of key volatile compounds in pumpkin seeds performed with different heat processing methods based on Maillard reaction and lipid oxidation. Journal of Food Measurement and Characterization. 2024;18(6):4764–4778. doi: 10.1007/s11694-024-02532-8. [DOI] [Google Scholar]
  26. Pedron G., Jaouhari Y., Bordiga M. Conventional and innovative drying/roasting technologies: Effect on bioactive and sensorial profiles in nuts and nut-based products. 2025;15(3) https://www.mdpi.com/2076-3417/15/3/1258 [Google Scholar]
  27. Pineda-Hidalgo K.V., Flores-Paredes G., Garzón-Tiznado J.A., Salazar-Salas N.Y., Chávez-Ontiveros J., López-Angulo G.…Lopez-Valenzuela J.A. The expression of genes involved in phenylpropanoid biosynthesis correlates positively with phenolic content and antioxidant capacity in developing chickpea (Cicer arietinum L.) seeds. Plants. 2025;14 doi: 10.3390/plants14162489. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Scalone G.L., Cucu T., De Kimpe N., De Meulenaer B. Influence of free amino acids, oligopeptides, and polypeptides on the formation of pyrazines in Maillard model systems. Journal of Agricultural and Food Chemistry. 2015;63(22):5364–5372. doi: 10.1021/acs.jafc.5b01129. [DOI] [PubMed] [Google Scholar]
  29. Shahidi F., Hossain A. Role of lipids in food flavor generation. Molecules. 2022;27 doi: 10.3390/molecules27155014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Smit B.A., Engels W.J.M., Smit G. Branched chain aldehydes: Production and breakdown pathways and relevance for flavour in foods. Applied Microbiology and Biotechnology. 2009;81(6):987–999. doi: 10.1007/s00253-008-1758-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Sun Z., Lin Y., Yang H., Zhao R., Zhu J., Wang F. Characterization of honey-like characteristic aroma compounds in Zunyi black tea and their molecular mechanisms of interaction with olfactory receptors using molecular docking. Lwt. 2024;191 doi: 10.1016/j.lwt.2023.115640. [DOI] [Google Scholar]
  32. Wang J., Zhang C., Qian J., Wang S., Fan W., Shi Q.…Chai G. Structure-threshold relationship in food aroma molecules: Insights from S-curve method, molecular docking, and dynamics simulations. Current Research in Food Science. 2025;10 doi: 10.1016/j.crfs.2025.101073. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Xie X., Wang Y., Wen B., Tian J., Cheng Z., Tang S.…Li B. Characterization and metabolism pathway of volatile compounds in blueberries of different varieties and origins analyzed via HS-GC-IMS and HS-SPME-GC–MS. Food Chemistry. 2025;480 doi: 10.1016/j.foodchem.2025.143813. [DOI] [PubMed] [Google Scholar]
  34. Xu L., Zang E., Sun S., Li M. Main flavor compounds and molecular regulation mechanisms in fruits and vegetables. Critical Reviews in Food Science and Nutrition. 2023;63(33):11859–11879. doi: 10.1080/10408398.2022.2097195. [DOI] [PubMed] [Google Scholar]
  35. Yang P., Luo T., Yang S., Zhang A., Tang Y., Chen L.…Lu L. Identification of olfactory receptors responding to androstenone and the key structure determinant in domestic pig. 2025;47(1) doi: 10.3390/cimb47010013. https://www.mdpi.com/1467-3045/47/1/13 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Yang W., Liu J., Liu H., Zhang Q., Lv Z., Jiao Z. Characterization of strawberry purees fermented by Lactobacillus spp. based on nutrition and flavor profiles using LC-TOF/MS, HS-SPME-GC/MS and E-nose. Lwt. 2023;189 doi: 10.1016/j.lwt.2023.115457. [DOI] [Google Scholar]
  37. Zhang S., Zhang L., Wang L., Zhao Y. Total phenols, flavonoids, and procyanidins levels and total antioxidant activity of different Korean pine (Pinus koraiensis) varieties. Journal of Forestry Research. 2019;30(5):1743–1754. doi: 10.1007/s11676-018-0744-0. [DOI] [Google Scholar]
  38. Zhou F., Wang A., Wang X., Wang C., Zhao J., Cao W.…Xu L. Effects of pollination cultivars on the characteristic volatile organic compounds in pears using E-nose, HS-SPME-GC–MS, and HS-GC-IMS. Food Chemistry: X. 2025;32 doi: 10.1016/j.fochx.2025.103266. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Zhu Y., Chen J., Chen X., Chen D., Deng S. Use of relative odor activity value (ROAV) to link aroma profiles to volatile compounds: Application to fresh and dried eel (Muraenesox cinereus) International Journal of Food Properties. 2020;23(1):2257–2270. doi: 10.1080/10942912.2020.1856133. [DOI] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary material 1: Supplementary Figures S1–S3 and Tables S1–S2
mmc1.docx (346.8KB, docx)
Supplementary material 2: Summary of Identified volatile organic compounds
mmc2.xlsx (173.7KB, xlsx)
Supplementary material 3: Summary of odor-active compounds
mmc3.xlsx (13.2KB, xlsx)
Supplementary material 4: Summary of Identified non-volatile organic compounds
mmc4.xlsx (312KB, xlsx)
Supplementary material 5: Significantly differential metabolites from pairwise comparisons
mmc5.xlsx (150.9KB, xlsx)

Data Availability Statement

Data will be made available on request.

Articles from Food Chemistry: X are provided here courtesy of Elsevier

RESOURCES