Sensomics-based characterization of cinnamon-like aroma in Rougui Wuyi Rock Tea: core odorants, perceptual masking, and thermal modulation

Abstract

Rougui Wuyi Rock Tea (WRT) is characterized by a distinctive cinnamon-like aroma, however, its odorant basis and perceptual modulation remain unclear. Using cinnamon bark as a sensory reference, a sensomics strategy combining HS-SPME-GC–MS, GC-O, aroma recombination and addition tests was applied to Rougui tea infusions. Seven aroma-active candidates were selected, among which benzenepropanal (BP) and α-terpineol (AT) were made major contributions to the cinnamon-like profile. Linalool (Lin) reduced cinnamon-like intensity from 5.83 to 2.00 in the reconstructed BP + AT system and from 6.67 to 4.33 in CinRG infusion, indicating floral masking (p < 0.05). A simplified thermal model showed Lin depletion and limited AT formation, providing evidence from the model system for thermally related changes in terpene balance. Molecular modelling provided structural context for the predicted interaction environments of complexes containing BP. These findings provided a sensory-chemical basis for understanding how cinnamon-like aroma was expressed in Rougui WRT.

Graphical abstract

Highlights

• •Sensomics identified 7 key odorants of cinnamon-like aroma in Rougui WRT.
• •BP and AT constituted the cinnamon-like odorant core.
• •AT reinforced BP-driven cinnamon-like perception in tea infusion.
• •Lin masked cinnamon-like perception by enhancing the floral background.
• •Heating depleted Lin and shifted terpene-related aroma balance.

1. Introduction

Oolong tea, distinguished by its unique semi-fermentation process, is renowned globally for its rich sensory attributes and health benefits (Wu et al., 2024). Among its subcategories, Wuyi Rock Tea (WRT) from northern Fujian, China, is particularly noted for its characteristic “Yanyun” (Feng et al., 2024). Rougui, one of the traditional WRT cultivars, is known for a distinctive cinnamon-like aroma, which is also reflected in its cultivar name, meaning “cinnamon” in Chinese.

The cinnamon-like aroma of Rougui tea differs from that of cinnamon bark (Cinnamomum cassia), which is largely associated with (E)-cinnamaldehyde (Xing et al., 2025). In tea infusion, this sensory attribute is unlikely to be explained by a single cinnamon-bark marker compound. Instead, it is more appropriately considered a perceptual expression arising from the combined contribution and modulation of multiple aroma-active compounds in the tea matrix (Chen et al., 2025). Previous studies had characterized the volatile composition of Rougui and WRT from different perspectives. Qiu et al. (2018) reported that (E)-nerolidol, farnesene, and benzaldehyde were important discriminators between Rougui and Shuixian cultivars, indicating the relevance of volatile profiles to cultivar differentiation. Guo et al. (2018) focused on roasting related changes and showed that pyrazines, such as methylpyrazine and 2,5-dimethylpyrazine, contributed to the roasted and peanutty aroma of WRT. Our previous work further suggested a negative association between cinnamon-like aroma intensity and floral volatiles such as linalool in Rougui tea (Liang et al., 2024). However, the sensory-validated odorant basis of the cinnamon-like note, and the possible modulation by floral background compounds, remain insufficiently understood.

Roasting is a critical processing step that reshapes the aroma profile of WRT and has been associated with enhanced woody, roasted, and cinnamon-like attributes. Lightly roasted Rougui samples generally retain stronger floral characteristics (Peng et al., 2024), whereas increased roasting intensity was accompanied by the formation or accumulation of aldehydes, nitrogen-containing heterocyclic compounds, and furan derivatives related to roasted, woody, and cinnamon-like notes (Yang et al., 2021). Zhang, Dong, and Liu (2025), Zhang, Luo, and Liu (2025), and Zhang, Wu, and Ma (2025) proposed that β-myrcene, an odorant associated with cinnamon-like perception, might increase through the transformation of geraniol and linalool during roasting. These observations suggested that floral terpenoids might be involved not only as aroma contributors, but also as modulators of cinnamon-like aroma expression through changes in odorant balance. In this context, linalool is of particular interest because it is a major floral odorant in tea and can undergo acid- or heat-related rearrangement reactions to generate α-terpineol and other terpene derivatives (Riu-Aumatell et al., 2014; Takoi et al., 2010). In addition, molecular docking and molecular dynamics simulations have been increasingly used to provide structural information on odorant-receptor interactions and odorant mixture effects (Liu et al., 2023; Zhang, Dong, & Liu, 2025; Zhang, Luo, & Liu, 2025; Zhang, Wu, & Ma, 2025). Although such computational approaches cannot replace sensory validation, they can provide structural context for interpreting how coexisting odorants may differ in their interaction patterns.

This study aimed to clarify the sensory contribution of aroma-active compounds associated with cinnamon-like aroma expression in Rougui WRT infusions. A sensory-guided sensomics strategy integrating HS-SPME-GC–MS, GC-O, absolute quantitation, aroma recombination, omission, and addition tests was applied to characterize key odorants and their perceptual roles. In addition, linalool masking tests and a simplified thermal model were used to evaluate how floral background compounds and thermally related terpene changes may modulate cinnamon-like aroma expression. Molecular docking and molecular dynamics simulations based on OR1A1 and OR2T4 receptor models were further used to provide structural support for predicted odorant interaction patterns. This work provides a sensory-chemical framework for understanding and regulating cinnamon-like aroma expression in Rougui tea infusion.

2. Materials and methods

2.1. Materials

A total of 30 Rougui Wuyi Rock Tea (WRT) samples were collected from four representative WRT producers in Wuyishan City, Fujian Province, China (processed in Spring 2024 and sourced in November 2024). To identify samples with distinct cinnamon-like aroma profiles, all teas were evaluated by a sensory panel consisting of seven professional experts, following the GB/T 23776–2018. Based on this assessment, the samples with the highest and lowest perceived cinnamon-like aroma intensity were selected for further experiments and designated as strong cinnamon-like aroma (CinRG, RG10) and weak cinnamon-like aroma (RG, RG17), respectively (Table S1). Concurrently, shredded cinnamon bark (Cin) from Sichuan Province was obtained in 2024 to serve as a sensory reference for the cinnamon-like aroma. All tea samples were traditionally processed and stored at −20 °C prior to analysis.

A standard of n-alkanes (C7–C40) was purchased from O2SI (Shanghai, China). Ethyl decanoate (≥ 98%), linalool (≥ 98%), hexyl benzoate (≥ 98%), 2-ethylpyrazine (≥ 98%), phenylethyl acetaldehyde (≥ 98%), 1-ethyl-1H-pyrrole-2-carbaldehyde (≥ 98%), α-terpineol (≥ 96%), hexyl hexanoate (≥ 97%), (E)-cinnamaldehyde (≥ 97%), benzyl alcohol (≥ 99%), hexanal (≥ 95%) and β-damascone (≥ 90%) were obtained from Sigma-Aldrich (Shanghai, China). 1-Octen-3-ol (≥ 98%), 2-ethyl-5-methylpyrazine (≥ 98%), benzeneacetaldehyde (≥ 98%), acetophenone (≥ 98%), terpinen-4-ol (≥ 98%), β-myrcene (≥ 90%), geraniol (≥ 97%), (E,E)-2,4-heptadienal (≥ 90%), (E)-2-nonenal (≥ 85%), benzenepropanal (≥ 95%), γ-terpinene (≥ 95%), (E)-2-heptenal (≥ 95%), methyl phenylacetate (≥ 99%), citral (≥ 95%), furfural (≥ 99.5%) and 0.1 mol/L citrate-phosphate buffer (pH 5.5) were sourced from Shanghai YuanYe Biotechnology Co., Ltd. (Shanghai, China).

2.2. Sample preparation

Tea samples or shredded cinnamon bark (5.0 g each) were brewed in a standard lidded porcelain cup (Gaiwan) with 110 mL of boiling water for 2 min. The resulting infusion was filtered and a 30 mL aliquot was portioned into a lidded sensory evaluation cup, which was blinded with a three-digit random code. The prepared samples were then maintained at 50 °C in a water bath for all subsequent olfactory evaluations and volatile analysis.

2.3. Volatiles analysis by HS-SPME-GC–MS/O

Volatile compounds were extracted by HS-SPME using a PDMS/DVB fiber (65 μm, Supelco, Bellefonte, PA, USA). Specifically, a 5 mL aliquot of the infusion, prepared as described in Section 2.2, was transferred into a 20 mL headspace vial. The sample was then spiked with 1 μL of ethyl decanoate (5 mg/L, diluted with ethanol, IS). The vial was capped and equilibrated in a 60 °C metal heating block for 15 min, after which the SPME fiber was exposed to the headspace for 30 min for extraction. Finally, the fiber was withdrawn and inserted into the gas chromatograph inlet for thermal desorption at 250 °C for 5 min (Meng et al., 2024).

GC–MS and GC-O analysis was performed following Zheng et al. (2025), respectively using a Shimadzu GC-2030 coupled to a QP2020 NX MS and an ODE–2030 olfactometer (Shimadzu, Tokyo, Japan) with an Agilent DB-5MS column (30 m × 0.25 mm, 0.25 μm). Helium (≥ 99.99%) was used as the carrier gas at 48.40 cm/s. Samples were injected in splitless mode at 250 °C. The oven program was 35  °C (2  min), ramped at 5  °C/min to 210  °C (3 min), then 10  °C/min to 250 °C (1  min). MS was operated at 70 eV with a 230 °C ion source and transfer line; spectra were acquired over m/z 35–350 with a 4 min solvent delay. For GC-O analysis, the column effluent was split at a ratio of 1:1 between the MS detector and the sniffing port. Four trained assessors (2 males and 2 females, 22–26 years old) participated in GC-O evaluation. Each sample was sniffed twice by each assessor. Odor events were recorded with their retention time, odor description, and aroma intensity (AI). AI was rated on a four-point scale: 1 = weak, 2 = moderate, 3 = strong, and 4 = very strong. Odor-active regions with a detection frequency (DF) ≥ 2 were retained for further consideration, and regions with AI ≥2 were considered relevant odor-active regions for subsequent candidate selection (Wang et al., 2024).

Data processing and qualitative analysis of VOCs were performed using the Shimadzu GCMSsolution software (Version 4.41). To minimize misidentification arising from potential co-elution, compound identification relied on mass spectra (NIST20) and retention indices (RIs). Specifically, the experimental retention indices (RI_exp) were determined using a C7–C40 n-alkane series under identical chromatographic conditions and compared with reference retention indices (RI_ref) obtained from the NIST and Leibniz LSB@TUM odorant databases. Furthermore, 20 odor-active compounds were unequivocally verified by matching their mass spectra, retention indices, and odor qualities with those of authentic reference standards (Zheng et al., 2025). The quantification of VOCs was performed in two stages. First, the relative concentration of all detected volatiles was semi-quantified using ethyl decanoate as an internal standard (IS). Based on these estimations, absolute quantification was then performed for 8 key analytes (β-myrcene, (E,E)-2,4-heptadienal, 1-ethyl-1H-pyrrole-2-carbaldehyde, linalool, (E)-2-nonenal, benzenepropanal, α-terpineol, and β-damascone) using external standard method combined with the IS calibration method. The calibration curve for each compound was established in a range that bracketed the concentrations found in the samples (Wang et al., 2024). To ensure accuracy, standards and samples were analyzed under identical GC–MS conditions. These absolute concentrations formed the basis for the subsequent aroma recombination and omission tests.

2.4. Aroma recombination, omission,addition and masking tests

Aroma recombination was performed to validate the contribution of the selected aroma-active candidates to the cinnamon-like profile of CinRG. A seven-component recombination model (RM) was prepared by adding β-myrcene, (E,E)-2,4-heptadienal, 1-ethyl-1H-pyrrole-2-carbaldehyde, (E)-2-nonenal, benzenepropanal, α-terpineol, and β-damascone to 30 mL of RG infusion. The final concentrations of these compounds in the RM were adjusted to match their quantified concentrations in CinRG (Table S2). The RM and CinRG infusions were then evaluated by the trained sensory panel.

Omission tests were conducted to determine the contribution of each candidate odorant to the reconstructed cinnamon-like profile (Wang et al., 2024). A series of omission models were prepared by removing one compound at a time from the full RM. The sensory difference between each omission model and the full RM was evaluated using a triangle test according to ISO 4120:2004. A significant result (p < 0.05) indicated that the omitted compound made a perceptible contribution to the reconstructed aroma profile.

To evaluate the sensory contribution of benzenepropanal (BP) and α-terpineol (AT) in a tea infusion matrix, addition tests were performed using RG infusion as the base matrix. The addition levels were defined according to the quantified concentration differences of BP and AT between CinRG and RG, with 1× corresponding to Δ(CinRG − RG). BP and AT were added individually or in selected concentration-ratio combinations, including BP 1×, AT 1×, BP 1× + AT 1×, BP 1× + AT 0.5×, BP 1× + AT 2×, BP 0.5× + AT 1×, and BP 2× + AT 1×. Detailed concentrations were provided in Fig. 4a. The cinnamon-like intensity of each model was evaluated by the trained sensory panel using a 10-point intensity scale.

Fig. 4.

Sensory validation of BP and AT reinforcement and Lin masking in Rougui tea infusion. (a) Experimental flowchart of addition and masking tests. (b) Cinnamon-like intensity of RG, CinRG, and BP and AT addition models. (c) Effects of externally added Lin on cinnamon-like and floral intensities in the reconstructed BP + AT core system. (d) Effects of externally added Lin on cinnamon-like and floral intensities in the CinRG infusion. The lowercase letters in panel (b) show the significance of differences between the means as tested by ANOVA followed by Duncan's test (p < 0.05). In panels (c) and (d), solid lines connect the observed mean values, and dashed lines indicate fixed-effect predictions from linear mixed-effects models fitted using individual panelist scores. The slope and p value shown in each panel refer to the fixed effect of externally added Lin concentration. RG: Rougui with weak cinnamon-like aroma; CinRG: Rougui with strong cinnamon-like aroma, BP: Benzenepropanal, AT: α-Terpineol, Lin: Linalool.

To examine whether a Lin-rich background modulated cinnamon-like perception, Lin masking tests were conducted in two sensory systems: the reconstructed BP + AT core system and the original CinRG infusion. In the reconstructed core system, Lin was added to the BP 1× + AT 1× model to obtain final added concentrations of 1, 5, 10, and 20 mg/L. In the CinRG infusion system, the same final added Lin concentrations were introduced into the original CinRG infusion. Samples without Lin addition served as controls. The original CinRG infusion contained endogenous Lin at 26.7 μg/L; therefore, the total Lin concentration in the CinRG matrix was calculated as the added Lin concentration plus 0.0267 mg/L endogenous Lin. All samples were coded with three-digit random numbers and evaluated for cinnamon-like and floral intensities using a 10-point intensity scale. The dose-dependent sensory changes induced by Lin addition were analyzed using the following linear mixed-effects model:

Y_ij = β0 + β₁Xij + bi + εij.

where Y_ij represents the sensory intensity score of panelist i at Lin addition level j, Xij represents the added Lin concentration, β0 is the fixed intercept, β₁ is the fixed effect of added Lin concentration, bi is the random intercept for panelist i, and εij is the residual error (Kuznetsova et al., 2015). Model estimates are reported as fixed-effect slopes with standard errors, 95% confidence intervals, and p values. For visualization, observed data are shown as mean ± SD, solid lines connect observed mean values, and dashed lines represent fixed-effect predictions from the mixed-effects models.

2.5. In vitro simulation of linalool transformation under thermal conditions

The in vitro thermal model was conducted in 20 mL headspace vials containing 5 mL of 0.1 mol/L citrate-phosphate buffer (pH 5.5), selected to match the pH of CinRG Maocha before roasting. Each vial was spiked with 500 μL of Lin standard solution (10 mg/L) and immediately sealed with a high-temperature resistant PTFE silicone septum. The vials were heated at 100 °C or 135 °C for 0, 2, 4, 8, and 16 h. After heating, the reactions were immediately cooled in an ice water bath.

The 0 h sample was analyzed under the same HS-SPME-GC–MS conditions and used as the baseline for evaluating thermal changes in Lin and its related products. The Lin concentration reported in the thermal model was the semi-quantified concentration determined by the internal standard method. Volatile analysis of the heated model systems followed the protocol described in Section 2.3. Sensory evaluation of the model systems was conducted according to the procedure described in Section 2.6.

2.6. Sensory analysis

2.6.1. Ethics statement

The ethical policies and procedures for all human sensory experiments were approved by the Ethics Committee of Fujian Agriculture and Forestry University (Approval No. PZCASFAFU25152). All participants provided written consent before their involvement.

2.6.2. Screening and training of sensory panelists

Panelists were recruited from the Fujian Agriculture and Forestry University (Fujian, China). Volunteers were non-smokers, non-drinkers, and possessed no taste or olfactory impairments. Following ISO 8586-2012, panelists were screened and trained using olfactory stimuli identification and intensity rating exercises.

2.6.3. Aroma profile evaluation (APE)

An aroma profile evaluation (APE) was conducted by a trained twenty-member panel (9 males, 11 females, aged 22–60; four with >10 years of experience). Following the method of Yang et al. (2022), the training was designed to ensure all panelists were proficient in identifying and rating six specific aroma descriptors: floral, fruity, woody, roasted, sweet, and cinnamon-like. To achieve this, panelists were trained using reference standards prepared in water at 100 times their odor thresholds (phenylethyl acetaldehyde, hexyl hexanoate, hexyl benzoate, 2-ethylpyrazine, benzyl alcohol, and (E)-cinnamaldehyde, respectively).

Panelists were also trained to apply a 10–point intensity scale (0–2: very weak; 2–4: weak; 4–6: moderate; 6–8: strong; 8–10: very strong). A 20 s rest was required between samples to minimize olfactory fatigue. All evaluations were conducted in triplicate by each panelist.

For Cin, RG, CinRG, the recombination model, omission models, and thermal-model samples, the six aroma descriptors were evaluated using the APE procedure. For BP + AT addition and Lin masking tests, cinnamon-like and floral intensities were evaluated as the primary sensory attributes.

2.7. Molecular docking and molecular dynamics simulation

Semi-flexible molecular docking was performed using AutoDock Vina 1.2.5 to explore the interactions between key aroma compounds and olfactory receptors (ORs). The 3D structures of BP (CID: 7707), AT (CID: 17100) and Lin (CID: 6549) were obtained from PubChem, while receptor models were retrieved from UniProt based on our previous study (Wu et al., 2025). Among 565 human ORs screened by preliminary docking, OR2T4 showed one of the lowest predicted binding energies and was therefore selected as a representative receptor model for subsequent structural visualization and molecular dynamics simulations (Table S3). OR1A1 was included as a broadly tuned olfactory receptor reference model supported by previous literature (Adipietro et al., 2012). Docking poses with RMSD <2.0 Å were accepted and visualized in PyMOL.

For each receptor model, docking was performed for BP alone and for ternary odorant systems containing BP + AT or BP + Lin. The lowest energy docking pose of each receptor-ligand system was selected for binding-mode analysis. Hydrogen-bond contacts, hydrophobic interactions, and π-stacking interactions were analyzed using Discovery Studio 2019 and visualized using PyMOL.

Molecular dynamics (MD) simulations were performed for the selected OR2T4 complexes, including OR2T4–BP, OR2T4–(BP + AT), and OR2T4–(BP + Lin). The simulations were conducted using GROMACS 2020.6 with the AMBER99SB force field and the SPC water model. Each complex was solvated in a cubic water box and subjected to energy minimization. The systems were then equilibrated under NVT and NPT ensembles for 100 ps each at 300 K, followed by a 100 ns production run with a 2 fs time step. Root-mean-square deviation (RMSD), radius of gyration (Rg), solvent-accessible surface area (SASA), hydrogen-bond number, and free energy landscape (FEL) were calculated from the trajectories. Binding free energies were calculated using MM/GBSA and MM/PBSA methods (Xu et al., 2025).

2.8. Statistical analysis

All data were expressed as the mean ± standard deviation (SD) of three replicates. For APE, each sample was evaluated in triplicate by each of the 20 trained panelists, and the mean panel score was used to represent the sensory intensity of each attribute. Statistical analyses were performed using SPSS 23.0 (SPSS Inc., IBM Corporation, Armonk, NY, USA). One-way ANOVA followed by Duncan's multiple range test was used for multiple comparisons among treatment groups at p < 0.05. For comparisons involving two sensory samples, Student's t-test was used. Mean absolute error, root mean square error, and Pearson correlation were calculated to evaluate the sensory similarity between RM and CinRG across the six aroma attributes. The significance of the sensory omission triangle tests was determined using a binomial test according to ISO 4120:2004. Linear mixed-effects models were fitted using R software (version 4.5.1; R Foundation for Statistical Computing, Vienna, Austria) (Chen et al., 2024; Zhang, Dong, & Liu, 2025; Zhang, Luo, & Liu, 2025; Zhang, Wu, & Ma, 2025). Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) was carried out in SIMCA-P (Version 13.0, MKS Umetrics AB, Umeå, Sweden). Molecular docking results were visualized with Discovery Studio 2019 (BIOVIA, Dassault Systèmes, San Diego, CA, USA) (Deng et al., 2025).

3. Results and discussion

3.1. Sensory profiling and screening of key cinnamon-like aroma-active compounds

3.1.1. Aroma profile evaluation (APE) of three samples

Prior to aroma profile evaluation (APE), 30 Rougui Wuyi Rock Tea (WRT) samples collected from four representative producers were preliminarily evaluated by seven professional tea experts according to GB/T 23776–2018. Based on this screening, two Rougui samples with contrasting cinnamon-like aroma intensities were selected for further study. CinRG (RG10) represented the sample with the strongest cinnamon-like perception, whereas RG (RG17) represented the weakest cinnamon-like perception among the screened samples (Table S1). Shredded cinnamon bark (Cin) was included as a sensory reference for cinnamon-like aroma.

The sensory profiles of Cin, CinRG, and RG were further quantified by a trained panel (n = 20, p < 0.05). As shown in Fig. 1a & 1b, Cin was characterized by pronounced cinnamon-like and woody notes. CinRG showed a clear cinnamon-like impression accompanied by a roasted background, consistent with the typical sensory style of roasted Rougui WRT (Liu et al., 2021; Yang et al., 2022). In contrast, RG showed a stronger floral profile but weaker cinnamon-like, woody, and roasted notes. In the correlation analysis, cinnamon-like intensity was negatively associated with floral and sweet attributes within the selected sensory set (r = −0.984, p < 0.0001; r = −0.820, p < 0.001). This result further supported the selection of RG and CinRG as contrasting Rougui tea infusions for subsequent chemical and sensory validation.

Fig. 1.

Sensory and volatile-profile differences among Cin, RG and CinRG. (a) Radar plot of the APE for Cin, RG, and CinRG; (b) Heatmap analysis among the aroma attributes; (c) Relative percentage category and quantity of VOCs for all samples; (d) Venn plot and correlation analysis among the shared volatile compounds and the key aroma attributes of Cin, RG, and CinRG. The lowercase letters indicated significant differences between means, as determined by ANOVA followed by Duncan's test (p < 0.05). The heatmaps show correlations between variables obtained by Pearson correlation analysis; ^⁎ indicates p < 0.05, and ^⁎⁎ indicates p < 0.01. APE: Aroma profile evaluation; Cin: shredded cinnamon bark; RG: Rougui with weak cinnamon-like aroma; CinRG: Rougui with strong cinnamon-like aroma.

3.1.2. Volatile profiling revealed different chemical bases of cinnamon bark and Rougui tea infusions

To obtain an overview of the volatile profiles of Cin, RG, and CinRG, HS-SPME-GC–MS analysis was performed. A total of 108 volatile features were annotated across the three samples, including 27 aldehydes, 20 ketones, 16 alcohols, 17 esters, 16 hydrocarbons, 11 heterocyclic compounds, and 1 other compound (Table S4).

The relative distribution of chemical classes differed markedly among the three samples (Fig. 1c). Although the total relative concentration of compounds was highest in the Cin sample, it contained the fewest compounds (n = 45), sharing only 12 compounds with the tea samples (Fig. 1d). This limited overlap indicated that the cinnamon-like perception of Rougui tea infusion was chemically distinct from cinnamon bark. In particular, (E)-cinnamaldehyde, the dominant odorant associated with cinnamon bark (Spence, 2024), was not detected in the Rougui tea infusions under the present analytical conditions.

Among the common volatiles, α-terpineol (AT) showed a positive association with cinnamon-like intensity (r = 0.834, p < 0.01), whereas linalool displayed an opposite distribution pattern between RG and CinRG. Specifically, linalool (Lin) was more abundant in RG, which was characterized by stronger floral perception, whereas CinRG showed lower linalool abundance and stronger cinnamon-like perception. This pattern was consistent with previous observations in roasted Rougui WRT, where enhanced woody, roasted, and cinnamon-like notes were accompanied by reduced floral perception and decreased Lin (Yang et al., 2022). These results suggested that AT and Lin might be involved in the sensory contrast between RG and CinRG, but their specific roles required validation by absolute quantitation and sensory experiments.

3.1.3. Chemometric and GC-O analyses selected aroma-active candidates for sensory validation

To further characterize the volatile compounds associated with cinnamon-like aroma, GC–MS/O analysis was performed. A supervised Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) model was then used as a chemometric tool to examine the sensory contrast among Cin, RG, and CinRG (Fig. 2). The model separated the three sample types (R²X = 0.994, R²Y = 0.991; Q²(cum) = 0.986), and the permutation test did not indicate apparent overfitting. Based on Variable Importance in Projection (VIP) score greater than 1 (Deng et al., 2024), 12 differential volatile candidates were selected (Table S5). These candidates provided a chemical basis for subsequent sensory-guided selection.

Fig. 2.

The OPLS-DA analysis for prioritizing differential volatile candidates among Cin, RG, and CinRG. The score scatter plot (a), loading scatter plot (b), and 200 permutations test (c) of OPLS-DA of the Cin, RG and CinRG. (d) The VIP scores and heatmap analysis between the volatile compounds with VIP ≥ 1 in OPLS-DA and the key aroma attributes. The heatmaps show correlations between variables obtained by Pearson correlation analysis, ^⁎ indicates p < 0.05, ^⁎⁎ indicates p < 0.01. Cin: shredded cinnamon bark; RG: Rougui with weak cinnamon-like aroma; CinRG: Rougui with strong cinnamon-like aroma.

To further identify the aroma-active compounds potentially contributing to cinnamon-like aroma in the tea infusions, it was analyzed using the Gas Chromatography-Olfactometry (GC-O) results (Table 1 & Fig. 3). A total of 54 odor-active compounds were perceived across all samples, including 15 in Cin, 26 in RG, and 36 in CinRG. Among these compounds, 46 were assigned to known odorants and 8 remained unidentified. The aroma intensity (AI) and specific attributes of these compounds are considered critical factors influencing the overall flavor quality of the tea infusion (Wei et al., 2026).

Table 1.

The aroma-active compounds were identified by GC–O analysis.

No.a	Aroma-active compoundsb	RIexpc	RIrefd	Odor descriptione	Aroma intensityf			Identificationg
					Cin	RG	CinRG
1	Hexanal	806	803	Green	/	1.0	/	MS,RI,O,S
2	Furfural	835	835	Caramel	/	/	1.5	MS,RI,O,S
3	(Z)-4-Heptenal	901	901	Fatty	/	/	1.8	MS,RI,O
4	1-(Furan-2-yl)ethanone	910	911	Roasted, Nutty	/	/	1.0	MS,RI,O
5	2,5-Dimethylpyrazine	914	912	Earthy, Nutty	/	1.9	/	MS,RI,O
6	Ethylpyrazine	916	913	Roasted	/	1.0	1.3	MS,RI,O
7	Unknown1	925	923	Green	/	/	1.2	O
8	(E)-2-Heptenal	957	957	Green, Herbal	/	1.7	/	MS,RI,O,S
9	1-Heptanol	971	972	Fatty	/	/	1.4	MS,RI,O,S
10	1-Octen-3-ol	981	983	Mushroom-like	/	1.6	/	MS,RI,O,S
11	6-Methyl-5-hepten-2-one	985	986	Green	/	1.0	1.0	MS,RI,O
12	β-Myrcene	991	991	Green, Spicy	/	1.5	3.1	MS,RI,O,S
13	2-Ethyl-5-methylpyrazine	1003	995	Roasted	/	/	1.0	MS,RI,O,S
14	(E,E)-2,4-Heptadienal	1011	1015	Fatty, Green	/	/	2.0	MS,RI,O,S
15	Unknown2	1025	1027	Spicy	1.0	/	/	O
16	Benzeneacetaldehyde	1043	1046	Spicy, Floral, Sweet	/	1.9	1.2	MS,RI,O,S
17	1-Ethyl-1H-pyrrole-2-carbaldehyde	1046	1046	Roasted, Smoky	/	2.3	3.0	MS,RI,O,S
18	Unknown3	1047		Floral	/	2.4	1.6	O
19	β-Ocimene	1048	1044	Green, Herbal	1.0	/	/	MS,RI,O
20	(E)-2-Octenal	1057	1059	Fatty	/	1.0	1.0	MS,RI,O
21	Acetophenone	1065	1062	Medicinal	/	/	1.2	MS,RI,O,S
22	(E,E)-3,5-Octadien-2-one	1068	1059	Herbal	/	/	1.4	MS,RI,O
23	3-Ethyl-2,5-dimethylpyrazine	1076	1079	Roasted, Nutty	/	1.0	1.4	MS,RI,O
24	Linalool	1102	1105	Floral	3.2	2.0	1.8	MS,RI,O,S
25	Hotrienol	1103	1106	Floral, Citrus	/	1.6	1.0	MS,RI,O
26	Nonanal	1104	1104	Citrus, Rose	2.4	/	/	MS,RI,O
27	β-Thujone	1107	1115	Herbal	1.2	/	/	MS,RI,O
28	Unknown4	1120	1134	Green	/	/	2.8	O
29	(E,Z)-3,6-Nonadien-1-ol	1155	1156	Cucumber-like	/	3.5	4.0	MS,RI,O
30	(E)-2-Nonenal	1158	1160	Fatty	/	/	2.0	MS,RI,O,S
31	Unknown5	1160		Woody	1.0	/	/	O
32	Benzenepropanal	1161	1163	Cinnamon-like, Woody, Floral	2.4	/	2.0	MS,RI,O,S
33	1-Phenyl-1-propanone	1163	1176	Herbal	/	1.6	2.4	MS,RI,O
34	Methyl phenylacetate	1173	1185	Honey-like	/	/	1.2	MS,RI,O,S
35	1-(Furan-2-ylmethyl)-1H-pyrrole	1177	1185	Earthy	/	1.2	/	MS,RI,O
36	Terpinen-4-ol	1181	1182	Earthy, Woody	2.0	/	/	MS,RI,O,S
37	Unknown6	1198	1198	Green	/	/	0.8	O
38	(E,E)-2,4-Nonadienal	1214	1216	Fatty	/	1.2	1.3	MS,RI,O
39	Hexyl 2-methylbutyrate	1234	1238	Fruity	/	1.6	1.4	MS,RI,O
40	Geraniol	1249	1256	Floral	/	2.0	1.2	MS,RI,O,S
41	Unknown7	1250		Minty-like	/	1.6	/	O
42	2-Phenylethyl acetate	1252	1256	Sweet, Rose-like	/	/	1.4	MS,RI,O
43	(E)-2-Decenal	1260	1265	Fatty, Green	1.0	/	/	MS,RI,O
44	Citral	1266	1276	Lemon-like	/	2.3	1.7	MS,RI,O,S
45	(E)-Cinnamaldehyde	1290	1282	Cinnamon-like	4.0	/	/	MS,RI,O,S
46	(E,E)-2,4-Decadienal	1316	1318	Fatty, Herbal	/	/	1.5	MS,RI,O
47	β-Damascone	1413	1415	Floral, Spice-like	/	1.0	1.1	MS,RI,O,S
48	α-Ionone	1421	1425	Woody, Floral	/	1.4	1.7	MS,RI,O
49	α-Curcumene	1482	1483	Herbal	2.0	/	/	MS,RI,O
50	(Z)-3-Hexenyl benzoate	1487	1488	Coconut-like, Fruity	/	1.1	1.3	MS,RI,O
51	α-Muurolene	1500	1497	Woody	1.6	/	/	MS,RI,O
52	Unknown8	1523		Floral	/	2.0	1.4	O
53	Caryophyllene oxide	1549	1575	Sweet	1.0	/	/	MS,RI,O
54	τ-Muurolol	1646	1640	Spice-like, Woody	1.0	/	/	MS,RI,O

^aOdorants were numbered consecutively according to their peak times on the DB-5MS capillary column.

^b“Unknown” denotes compounds that could not be identified by GC-O-MS.

^cRI_exp: Experimental retention indices calculated on a DB-5MS column using C7–C40 n-alkanes.

^dRI_ref: Literature retention indices from the NIST or Leibniz LSB@TUM odorant database for the DB-5MS columns.

^eOdor quality assessed at the sniffing port by panelists.

^fAroma intensity evaluated on a 4-point scale (1 = weak to 4 = very strong).

^gCompounds were identified using the following abbreviations: MS, mass spectra; RI, retention index; O, olfactometry; S, authentic reference compounds. A compound that was not able to be identified by an authentic reference compound was labeled as “tentatively identified”. “/” indicated the odor was not perceived at the sniffing port. Cin: shredded cinnamon bark; RG: Rougui with weak cinnamon-like aroma; CinRG: Rougui with strong cinnamon-like aroma.

Fig. 3.

Analysis of aroma-active compounds in three samples and recombination verification. (a) Aroma intensity (AI) of Cin, RG, and CinRG; (b) Comparison of aroma profiles between the CinRG and recombination model (RM). The significance of differences between the means as tested by t-test. “ns” means p > 0.05. Cin: shredded cinnamon bark; RG: Rougui with weak cinnamon-like aroma; CinRG: Rougui with strong cinnamon-like aroma.

In RG, floral odor-active odorants were prominent, with linalool and geraniol (both AI = 2.0) showing noticeable aroma intensities. In contrast, CinRG showed a shift toward spicy, woody, and roasted odor-active odorants, accompanied by weaker floral contributors. Benzenepropanal (BP, AI = 2.0) was perceived as a cinnamon-like odorant at the sniffing port, while β-myrcene (AI = 3.1) and β-damascone (AI = 1.5) contributed spicy notes. In addition, 1-ethyl-1H-pyrrole-2-carbaldehyde (AI = 3.0) was associated with the roasted background characteristic of WRT. These GC-O results were consistent with the sensory profiles and indicated that the cinnamon-like aroma of Rougui tea infusion was associated with a combination of cinnamon-like, spicy, woody, roasted, and floral odorants rather than a single cinnamon-bark marker.

Based on the integration of volatile profiling, OPLS-DA, GC–O aroma intensity, and odor attributes, seven sensory-selected candidates were selected for subsequent validation: AT, β-myrcene, (E,E)-2,4-heptadienal, 1-ethyl-1H-pyrrole-2-carbaldehyde, (E)-2-nonenal, BP, and β-damascone.

3.2. Aroma recombination and omission identify major contributors to cinnamon-like aroma

Aroma recombination was conducted using the seven sensory-selected candidates at their quantified concentrations in CinRG (Fig. 3b & Table S2). To further evaluate the fidelity of the recombination model, the sensory profile of RM was quantitatively compared with that of CinRG across the six aroma attributes. Based on the attribute intensity means, RM showed a mean absolute error of 0.79 points and a root mean square error of 0.97 points on the 10-point sensory intensity scale relative to CinRG. When normalized to the full scale range, these values corresponded to 7.9% and 9.7%, respectively. The two mean profiles were positively correlated across the six aroma attributes (r = 0.844, p < 0.05), supporting the overall similarity between RM and CinRG while still indicating deviations at the attribute level (Table S6). Together with the absence of significant differences in individual sensory attributes (p > 0.05), these results indicated that RM approximated the major odorant contribution to the cinnamon-like profile of CinRG, although some sensory nuances remained unreproduced.

The omission tests further clarified the relative contribution of individual candidates. Removing BP or AT caused a highly significant change in the reconstructed aroma profile (p < 0.001), indicating that both compounds were critical for maintaining the cinnamon-like perception. In contrast, omission of β-myrcene, (E)-2-nonenal or β-damascone produced no significant change (p > 0.05), whereas 1-ethyl-1H-pyrrole-2-carbaldehyde and (E,E)-2,4-heptadienal showed moderate effects (Table 2). These results identified BP and AT as core contributors to cinnamon-like aroma expression, while the remaining candidates mainly contributed to the spicy, roasted, woody, or background characteristics of the reconstructed profile.

Table 2.

The analysis of aroma omission test.

No.	Aroma compounds omitted from the spiked sample	CAS	Odor quality	n/20a	Significanceb
1	β-Myrcene	123-35-3	Green, Spicy	8	ns
2	(E,E)-2,4-Heptadienal	4313-03-5	Fatty, Green	11	*
3	1-ethyl-1H-pyrrole-2-carbaldehyde	2167-14-8	Roasted, Smoky	11	*
4	(E)-2-Nonenal	18829-56-6	Fatty	9	ns
5	Benzenepropanal	104-53-0	Cinnamon-like, Woody, Floral	17	***
6	α-Terpineol	98-55-5	Floral, Lilac-like	16	***
7	β-Damascone	23726-91-2	Floral, Spice-like	4	ns

^an/20 indicated the ratio of panelists who detected significant differences out of the total of 20 panel members.

^b*** for very high significant (p < 0.001). * for significant (p < 0.05), and “ns” for not significant (p > 0.05).

3.3. Sensory validation of core odorant reinforcement and floral masking in Rougui tea infusion

To evaluate the sensory roles of the two core odorants in a tea infusion matrix, BP and AT were added to RG infusion either individually or in selected concentration-ratio combination (Fig. 4). Compared with RG infusion, BP addition increased the cinnamon-like intensity from 2.33 to 4.33 (p < 0.05), whereas AT alone caused only a slight change from 2.33 to 2.67 (p > 0.05). When BP and AT were added together at the 1× level, the cinnamon-like intensity further increased to 5.83, approaching that of CinRG (p > 0.05).

The concentration-ratio addition models further supported the dominant sensory contribution of BP and the reinforcing contribution of AT. When BP was maintained at 1×, increasing AT from 0.5× to 2× increased cinnamon-like intensity from 4.67 to 6.33 (p < 0.05). Conversely, when AT was maintained at 1×, increasing BP from 0.5× to 2× produced a more pronounced increase in cinnamon-like intensity, from 3.33 to 7.87 (p < 0.05). These results indicated that BP acted as the major contributor to the cinnamon-like note, while AT further enhanced cinnamon-like perception when co-present in the Rougui tea infusion matrix. Therefore, cinnamon-like aroma expression was not determined solely by the presence of individual odorants, but was also influenced by their concentration balance.

Because RG was characterized by stronger floral perception and higher Lin abundance than CinRG, we further examined whether Lin-rich floral background could counteract the BP and AT associated cinnamon-like perception. Lin was therefore added to two sensory systems: the BP and AT reconstructed core system and the original CinRG infusion. In the reconstructed core system, increasing Lin concentrations from 0 to 20 mg/L reduced cinnamon-like intensity from 5.83 to 2.00 (p < 0.05), while floral intensity increased from 3.83 to 8.67 (p < 0.05). A similar pattern was observed in the CinRG infusion, where Lin addition decreased cinnamon-like intensity from 6.67 to 4.33 (p < 0.05)and enhanced floral perception from 2.33 to 7.33 (p < 0.05).

Linear mixed-effects models based on individual panelist scores further supported the dose-dependent masking effect of Lin. In the reconstructed BP + AT core system, added Lin concentration was negatively associated with cinnamon-like intensity (p < 0.001) and positively associated with floral intensity (p < 0.001). Similar trends were observed in the original CinRG infusion, where added Lin concentration was negatively associated with cinnamon-like intensity (p < 0.001) and positively associated with floral intensity (p < 0.001). These results indicated that increasing Lin progressively weakened cinnamon-like perception while strengthening floral perception over the tested concentration range.

Together, the addition and masking experiments indicated that cinnamon-like aroma expression in Rougui tea infusion was shaped by both core odorant reinforcement and perceptual masking. BP acted as the major cinnamon-like contributor, AT further enhanced BP-associated cinnamon-like perception under the tested concentration-ratio conditions, and Lin attenuated cinnamon-like perception by increasing the floral background in a dose-dependent manner. In parallel, Chigo-Hernandez and Tomasino (2023) reported that a combination of AT and Lin could produce a distinct ginger-like aroma in wine. These results indicated that Lin addition induced a dose-dependent shift in sensory balance, enhancing floral perception while weakening cinnamon-like expression.

3.4. Thermal response of linalool and associated sensory shifts

Previous studies have shown that roasting can enhance woody, roasted, and cinnamon-like notes in WRT, accompanied by a decrease in floral volatiles such as Lin (Yang et al., 2022; Zhang, Dong, & Liu, 2025; Zhang, Luo, & Liu, 2025; Zhang, Wu, & Ma, 2025). In addition, Lin can undergo acid- or heat-related rearrangement reactions to form AT and other terpene derivatives (Riu-Aumatell et al., 2014; Takoi et al., 2010). Based on the masking effect of Lin and the reinforcing role of AT observed in the tea-infusion experiments, a simplified thermal model was established to examine whether heating could shift the odorant balance between Lin and terpene-related products.

Lin was incubated in citrate-phosphate buffer at pH 5.5, with the pH selected according to the pH of the CinRG maocha stage prior to roasting. Two heating temperatures, 100 °C and 135 °C, were applied for 0–16 h (Fig. 5). At 135 °C, Lin decreased markedly from 66.0 μg/L at 0 h to 1.9 μg/L after 16 h. In parallel, AT was formed during heating and reached a maximum concentration of 2.8 μg/L at 8 h before declining. To exclude potential GC injection-related artifact formation, the 0 h sample was analyzed under the same instrumental conditions. AT was not detected in this control, despite the high initial linalool concentration, indicating that AT formation was not attributable to GC analysis (Reinhardt & Steinhaus, 2025). In contrast, Lin showed a slower decline at 100 °C and AT formation remained negligible, with only 0.02 μg/L detected after 16 h. The amount of AT formed was much lower than the amount of Lin lost, indicating that AT was only one of the transformation products rather than the sole endpoint of Lin degradation. Other terpene-related products, such as limonene, terpinolene, and β-myrcene, were also detected during heating (Table S7), consistent with the possible occurrence of competing cyclization, isomerization, and dehydration reactions (Chang et al., 2021).

Fig. 5.

Thermal transformation of linalool and sensory shifts in a simplified acidic model system. (a) Schematic overview of the thermal model. (b) Semi-quantitative changes in Lin during heating at 100 °C and 135 °C. (c) Semi-quantitative changes in AT during heating at 100 °C and 135 °C. (d) Sensory shifts of the heated model systems. The letters show the significance of differences between the means as tested by ANOVA followed by Duncan's test (p < 0.05). BP: Benzenepropanal, AT: α-Terpineol, Lin: Linalool.

Consistent with the compositional changes, sensory evaluation showed that the samples treated at 135 °C exhibited lower floral and fruity intensities and higher woody-like perception than the samples treated at 100 °C. This sensory shift was consistent with the marked depletion of Lin and the transient formation of AT under intensified heating. Together with the addition and masking experiments, these results suggested that thermal treatment might modulate cinnamon-like aroma expression by altering the balance between floral masking components and terpene-related contributors.

It should be noted that this experiment was conducted in a simplified acidic aqueous model system and was not intended to fully reproduce industrial roasting conditions. Therefore, the observed changes in Lin, AT, and other terpene-related products should be interpreted as model-system evidence under the tested conditions, rather than direct proof of transformation pathways during Rougui roasting. Further studies using stable isotope labeled precursors, together with tea samples collected from defined processing stages, are needed to clarify the transformation routes under authentic roasting conditions.

3.5. Molecular modelling of core and masking odorants in olfactory receptor models

3.5.1. Molecular docking in OR1A1 and OR2T4 models

Molecular docking is commonly used to predict favorable binding conformations of ligands within receptor binding pockets (Liu et al., 2023). In this study, molecular docking was used to compare the predicted binding environments of BP complexes in selected olfactory receptor models, providing structural context for subsequent MD simulation. Based on the sensory validation results, BP was selected as the primary odorant. A binary receptor-ligand complex containing BP alone was first constructed, and AT or Lin was then introduced to generate ternary complexes containing BP + AT or BP + Lin. This design was used to evaluate whether coexisting odorants altered the binding conformation of complexes containing BP and the interacting residues around the binding site. Two olfactory receptor models were selected for comparative modelling. OR1A1 was included as a literature supported reference receptor model because it has been reported as a broadly tuned human olfactory receptor responsive to diverse odorants (Adipietro et al., 2012). OR2T4 was retained as a candidate receptor model from the virtual screening of 565 human olfactory receptors, showing the lowest docking energy with BP (−6.8 kcal/mol, Table S3). For each receptor-ligand system, the lowest-energy conformation was selected for binding-mode analysis.

In the OR2T4 model (Fig. 6A & Table S8), BP was surrounded mainly by hydrophobic residues such as VAL-137, VAL-213, and MET-231, together with aromatic residues such as PHE-144, PHE-191, and TYR-291. PHE-191 was involved in π-stacking, while VAL-137 participated in hydrophobic interactions and appeared near a predicted hydrogen-bond contact in the docking model. After AT was introduced, the BP + AT complex showed a lower docking energy of −10.3 kcal/mol and involved a broader set of contact residues, including VAL-137, ALA-140, PHE-191, HIS-207, VAL-213, ALA-215, ASN-218, LEU-219, TYR-291, and PRO-300. Additional hydrogen-bond contacts involving VAL-137 and ASN-218 and π-stacking with HIS-207 were observed, suggesting that AT changed the local interaction environment of the complexes containing BP (Tam et al., 2022; Zhang et al., 2024). For the BP + Lin complex, the docking energy was −9.8 kcal/mol. This system involved different hydrogen-bond contacts, mainly ASP-187 and MET-292. Hydrophobic interactions were associated with VAL-137, PHE-144, VAL-213, MET-231, TYR-291, PRO-214, GLU-228, and TYR-232, while π-stacking with PHE-191 was retained. These differences suggested that AT and Lin might differentially alter the predicted residue-level interaction patterns and interaction types of complexes containing BP within the selected receptor models.

Fig. 6.

Molecular modelling of benzenepropanal-containing odorant complexes in OR1A1 and OR2T4 receptor models. (a) Representative docking conformations of BP, BP + AT, and BP + Lin in OR1A1 and OR2T4 receptor models. Interacting residues were shown in yellow, ligands were shown in cyan, and dashed lines indicated predicted hydrogen-bond contacts. Molecular dynamics analysis of selected OR2T4 complexes during 100 ns MD simulation: (b) Root-mean-square deviation (RMSD), radius of gyration (Rg), solvent-accessible surface area (SASA), number of hydrogen bonds; and (c) Free energy landscape (FEL). BP: Benzenepropanal, AT: α-Terpineol, Lin: Linalool. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

A similar distinction was observed in the OR1A1 model. BP alone showed a docking energy of −6.6 kcal/mol, with predicted contacts involving TYR-258, TYR-276, VAL-203, and PHE-206. The BP + AT complex showed the lowest docking energy among the OR1A1 systems (−12.2 kcal/mol), with additional contacts involving ASN-109, ASN-155, ASN-176, TYR-178, and TYR-265. In contrast, the BP + Lin complex showed a docking energy of −9.1 kcal/mol and mainly involved TYR-113, TYR-285, TYR-276, TYR-120, LEU-144, VAL-203, PHE-206, and VAL-254. Thus, in both receptor models, AT and Lin altered the predicted residue contact profiles of BP containing complexes. These results provided a structural basis for subsequent MD simulation and a hypothesis for possible odorant interaction patterns, rather than direct evidence of receptor activation or sensory interaction (Kurahashi et al., 1994; Yuan et al., 2025).

3.5.2. Molecular dynamics simulation of OR2T4-odorant complexes

To further evaluate the dynamic stability of the docked conformations, 100 ns molecular dynamics (MD) simulations were performed for the selected OR2T4–odorant complexes (Fig. 6B & C). The root mean square deviation (RMSD) profiles of OR2T4–BP, OR2T4–(BP + AT), and OR2T4–(BP + Lin) gradually reached equilibrium after approximately 20 ns, indicating that the complexes attained relatively stable conformations during the simulation. During the equilibrium stage of 20–100 ns, the average RMSD values were 0.93 ± 0.06, 1.25 ± 0.12, and 1.02 ± 0.01 nm, respectively. Consistent trends in the radius of gyration (Rg) and solvent-accessible surface area (SASA) suggested the formation of relatively compact and dynamically stable conformational states (Mahuri et al., 2023). The number of hydrogen bonds also remained within a limited fluctuation range during the equilibrium stage, indicating persistent polar contacts between the ligand complexes and the OR2T4 binding pocket. In addition, the free energy landscape (FEL) plots showed that each complex occupied a defined low-energy basin after equilibration, further supporting the conformational stability of the simulated systems.

Binding free energies were then estimated using MM/GBSA and MM/PBSA methods (Table 3). The MM/GBSA binding free energies of OR2T4–BP, OR2T4–(BP + AT), and OR2T4–(BP + Lin) were − 33.32 ± 1.44, −62.34 ± 0.66, and − 62.56 ± 2.96 kcal/mol, respectively. The corresponding MM/PBSA values were − 27.82 ± 1.62, −48.82 ± 2.48, and − 48.72 ± 2.97 kcal/mol, respectively. Energy decomposition showed that van der Waals interactions made the major contribution to complex stabilization, consistent with the hydrophobic and aromatic residues identified in the docking analysis (Guo et al., 2024; Guo et al., 2025).

Table 3.

The binding energies of ligand-receptor complexes using MM/GBSA and MM/PBSA methods (kcal/mol).

Methods	Parameters	OR2T4–BP	OR2T4–(BP + AT)	OR2T4–(BP + Lin)
MM/GBSA	Delta van der Waals Energy	−33.67 ± 0.97	−62.78 ± 0.54	−63.96 ± 2.06
	Delta Electrostatic Energy	−2.36 ± 0.95	0.00	−3.77 ± 1.22
	Delta Generalized Born Energy	6.32 ± 0.47	7.16 ± 0.38	12.40 ± 0.03
	Delta Solvent Accessible Surface Area Energy	−3.61 ± 0.01	−6.72 ± 0.07	−7.22 ± 0.20
	Total Binding Energy	−33.32 ± 1.44	−62.34 ± 0.66	−62.56 ± 2.96
MM/PBSA	Delta van der Waals Energy	−33.67 ± 0.97	−62.78 ± 0.54	−63.96 ± 2.06
	Delta Electrostatic Energy	−2.36 ± 0.95	0.00	−3.77 ± 1.22
	Delta Polar Solvation Energy	10.00 ± 0.88	14.93 ± 2.43	23.06 ± 0.28
	Delta Non-Polar Solvation Energy	−1.79 ± 0.09	−0.96 ± 0.02	−4.05 ± 0.02
	Total Binding Energy	−27.82 ± 1.62	−48.82 ± 2.48	−48.72 ± 2.97

Overall, the MD results showed that the OR2T4–odorant complexes maintained stable conformations during the simulation and that their stabilization was mainly associated with van der Waals-driven interactions. Together with the docking analysis, these results provide a structural perspective on how AT and Lin may influence the predicted interaction environment of BP containing complexes in selected olfactory receptor models. However, these computational results should be interpreted as structural context and require future validation using functional odorant receptor assays.

4. Conclusions

This study characterized the sensory-chemical basis and perceptual modulation of cinnamon-like aroma in selected Rougui Wuyi Rock Tea (WRT) infusions using a sensory-guided sensomics strategy. Based on volatile profiling, chemometric analysis, GC-O, and absolute quantitation, seven aroma-active candidates were selected for sensory validation: α-terpineol (AT), β-myrcene, (E,E)-2,4-heptadienal, 1-ethyl-1H-pyrrole-2-carbaldehyde, (E)-2-nonenal, benzenepropanal (BP), and β-damascone. Aroma recombination and omission tests indicated that BP and AT were major contributors to the reconstructed cinnamon-like profile of CinRG.

Further addition experiments in the tea infusion matrix showed that BP provided the major cinnamon-like contribution, while AT reinforced BP-associated cinnamon-like perception when the two compounds coexisted. In contrast, linalool (Lin) addition reduced cinnamon-like intensity and increased floral perception in both the reconstructed core system and the original CinRG infusion, supporting a masking effect of the Lin-rich floral background. These results indicated that cinnamon-like aroma expression in Rougui tea infusion was shaped not only by core odorants, but also by the balance between enhancing and masking odorants. A simplified acidic thermal model further showed that Lin decreased markedly under intensified heating, accompanied by limited formation of AT and other terpene-related products. The corresponding sensory shift from floral/fruity toward woody perception suggested that thermal treatment might influence cinnamon-like aroma expression by reshaping the balance between floral masking components and terpene-related contributors. However, these results chemically support the possibility that heating alters the balance between linalool and related terpene products. Molecular docking and molecular dynamics simulations using OR1A1 and OR2T4 receptor models provided structural context for the predicted binding modes and contact patterns with the receptor for complexes containing BP. The BP + AT and BP + Lin complexes showed different interacting residues and interaction types in the receptor models, while MD simulations indicated that the selected OR2T4 complexes maintained relatively stable conformations. However, these computational results require future validation using functional odorant receptor assays.

Overall, the cinnamon-like aroma of Rougui WRT infusion was not attributable to a single cinnamon-bark marker compound. Instead, it represented a matrix-dependent perceptual expression primarily supported by BP and AT and modulated by Lin-mediated floral masking and thermally induced changes in terpene-related odorant balance. These findings provided a sensory-chemical framework for understanding the distinctive cinnamon-like aroma of Rougui tea and offered a basis for targeted aroma regulation during WRT processing.

CRediT authorship contribution statement

Sihan Deng: Writing – original draft, Methodology, Data curation. Huanqin Luo: Writing – review & editing, Conceptualization. Wei Tao: Validation, Formal analysis. Xinran Sun: Methodology, Investigation. Chenglong Li: Investigation, Data curation. Weiwei Wu: Software. Fuming Lin: Visualization, Investigation. Lili Wang: Resources. Yan Huang: Supervision. Yong-Quan Xu: Writing – review & editing, Conceptualization. Weijiang Sun: Writing – review & editing, Project administration, 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.

Appendix A. Supplementary data

Supplementary material

Sample information and sensory evaluation data; aroma recombination and addition test designs; volatile compound quantification and VIP analysis; external standard calibration data; in vitro simulation procedures and results; molecular docking parameters and binding energy data (PDF).

Data availability

Data will be made available on request.