Decoding the structure-flavor interplay of Lanmaoa asiatica under different thermal processing methods: Insights from multimodal analysis and molecular docking

Abstract

This study aimed to elucidate the differential effects of various heating methods on the structure and flavor of Lanmaoa asiatica. A multi-scale structure-flavor correlation approach combined with molecular docking was employed. The results showed that frying formed a dense outer crust and immobilized water, thereby promoting lipid cleavage and Maillard reactions, which contributed fatty, roasted, and mushroom-like aromas. Steaming loosened tissue structure and enhanced the release of polar small molecules, thereby highlighting fruity-floral notes. Boiling exhibited a “dual-low” profile of both precursors and flavor products. Four key volatiles (nonanal, phenylacetaldehyde, 2,4-heptadienal, 3-methylbutanal) were highly correlated with sensory phenotypes, and molecular docking suggested favorable binding interactions with specific olfactory receptors, mainly through aromatic stacking and hydrophobic embedding. These findings advance the understanding of flavor generation and release by integrating structural characteristics and water distribution, providing a theoretical reference for flavor modulation and processing optimization of edible mushrooms.

Graphical abstract

Highlights

• •Multiscale structure-water coupling governs aroma generation and release.
• •GC–MS and GC-IMS complementarity identified four shared key volatiles.
• •E-nose, tongue-volatile-amino acid correlations reveal chemical-phenotype match.
• •Molecular docking elucidates specific recognition of key volatiles by receptors.
• •Cross-scale mechanism linking structure, volatiles, amino acid, sensory, and ORs.

1. Introduction

Lanmaoa asiatica (L. asiatica), commonly known as “Jianshouqing” because of its rapid blue discoloration after tissue damage (Wu et al., 2016), is a wild edible mushroom valued for its firm texture, umami taste, and nutritional components, including proteins, polysaccharides, and micronutrients (Tan et al., 2022; Zheng et al., 2024). However, its fresh fruiting bodies are highly perishable and contain heat labile toxins, making thorough thermal processing essential before consumption (Landi et al., 2021). Common cooking methods such as frying, steaming, and boiling differ in heat and mass transfer and can consequently affect microstructure, water distribution, texture, and flavor characteristics.

Different thermal processing methods alter food structure and flavor through differences in temperature, water activity, and heat and mass transfer. Heating media and transfer patterns affect cell walls, protein networks, pore structure, and water migration (Rahman et al., 2018; Yang et al., 2025). These structural and water changes influence texture (Liu et al., 2025) and can also affect molecular diffusion and key thermal reactions, including lipid oxidation, Strecker degradation, and the Maillard reaction (El Hosry et al., 2025; Vu et al., 2020), thereby shaping the formation and release of aroma-active compounds. High temperature and low water activity conditions generally favor more extensive thermal reactions and volatile accumulation (Luo et al., 2024), whereas moist heat conditions may facilitate precursor release (Bay Yılmaz et al., 2025), while liquid phase heating can promote the migration and loss of soluble flavor compounds (You et al., 2025). However, how different thermal processing methods affect the structure, water distribution, and aroma profile of L. asiatica remains insufficiently understood. Therefore, systematic comparison of these changes is needed to clarify their relationships with quality formation.

Multimodal analytical approaches can provide complementary information on aroma composition and sensory characteristics during thermal processing. Headspace solid-phase microextraction coupled with gas chromatography–mass spectrometry (HS-SPME-GC–MS) enables broad profiling of volatile compounds (Liu et al., 2025), while gas chromatography-ion mobility spectrometry (GC-IMS) provides complementary detection and visualization of volatile fingerprints (Parastar & Weller, 2024). Electronic nose (E-nose) and electronic tongue (E-tongue) can rapidly characterize overall odor and taste responses (Lu et al., 2022; Wei et al., 2023), whereas molecular docking provides in silico predictions of potential interactions between key aroma compounds and human olfactory receptors (Jin & Wei, 2024). However, studies on wild edible mushrooms have often focused on individual processing methods or analytical dimensions, leaving the relationships among structural changes, volatile profiles, and sensory responses insufficiently understood. Integrating these complementary approaches can therefore provide a more comprehensive assessment of thermal processing effects on edible mushroom quality.

This study evaluated the effects of thermal processing on the structure and flavor of L. asiatica using structural, water distribution, volatile, amino acid, sensory, and receptor interaction analyses. To this end, texture analysis, scanning electron microscopy, and low field nuclear magnetic resonance were used to characterize structure and water states; GC–MS and GC-IMS to profile volatile compounds; E-nose, E-tongue, and correlation analyses to assess sensory responses and their associations with volatile and amino acid profiles; and molecular docking to explore potential interactions between key volatiles and olfactory receptors. Overall, this integrated approach provides a better understanding of processing related quality and flavor differences in L. asiatica and offers a reference for optimizing thermal processing of edible mushrooms.

2. Methods and materials

2.1. Sampling and cooking methods of L. asiatica

Fresh Lanmaoa asiatica samples were purchased from the Yunnan Mushuihua Wild Mushroom Market (Kunming, China). Uniform, fresh fruiting bodies without insect damage, mold, or decay were selected. After removing impurities, the mushrooms were rinsed with clean water to eliminate surface soil and debris, drained on a stainless-steel sieve, and longitudinally sliced along the stipe into slices of approximately 5 mm thickness. The prepared samples were then randomly divided into three groups for frying, steaming, and boiling treatments, as described below. The chosen conditions were based on common culinary practices for L. asiatica and supported by literature, ensuring both safety and flavor retention (Arora, 2008). Although there are differences in the heating conditions among the groups, this study ensured maximum consistency between the treatment groups by carefully controlling the thermal processing parameters, thereby effectively minimizing the influence of confounding variables. This provides a reliable experimental foundation for a comprehensive comparison of the effects of different thermal treatments on L. asiatica.

Frying treatment (FT): Refined rapeseed oil was added to a wok at a mushroom-to-oil ratio of 1:20 (g/mL, mass to volume). Fresh refined rapeseed oil from the same batch was used for each independent frying replicate and was not reused. When the oil temperature reached 160–170 °C, mushroom slices were fried for 20 min until the surface turned uniformly golden yellow without charring. Excess oil was removed using absorbent paper, and the samples were cooled to room temperature for further use.

Steaming treatment (ST): Water at ambient temperature was added to a steamer with a mushroom-to-water ratio of 1:20 (g/mL). Once the water began boiling, the mushroom slices were placed on a steamer tray and steamed for 30 min. After steaming, samples were cooled to room temperature for further use.

Boiling treatment (BT): Water at ambient temperature was added to a pot with a mushroom-to-water ratio of 1:20 (g/mL). After boiling, mushroom slices were added and further boiled for 30 min. The samples were then removed, drained, and cooled to room temperature for further analysis.

After thermal processing, all samples were cooled to room temperature, after which those for texture analysis were immediately tested, while the remaining samples were rapidly frozen at −80 °C for further analyses. Given that L. asiatica requires thorough cooking before consumption, the experimental design focused on comparing three commonly used thermal processing methods rather than tracing changes from the fresh state under a single processing treatment.

2.2. Texture profile analysis

The shear force and texture properties of L. asiatica samples were measured using a texture analyzer (TA. XT PlusC, Stable Micro Systems Ltd., UK). Shear force test: An HDP/BS shear blade probe was used with the following setting: pre-test speed of 10.0 mm/s, test speed of 2.0 mm/s, post-test speed of 10.0 mm/s, displacement 6 mm, and trigger force of 10 g. Texture profile analysis: A P/36R cylindrical probe was used under the following conditions: pre-test speed of 10.0 mm/s, test speed of 2.0 mm/s, post-test speed of 10.0 mm/s, strain set to 40%, time interval of 2 s, and trigger force of 10 g.

2.3. Scanning electron microscopy (SEM) analysis

After thermal treatment, samples from all groups were freeze dried prior to SEM analysis, approximately 1.0 mm-thick square tissue slices were cut from the same region of each treatment group. Samples were mounted on specimen stubs using conductive adhesive and coated with a thin layer of metal using via a sputter coater (Oxford Ultim Max 65, Thermo Fisher Scientific Brno Ltd., UK) under a current of 10 mA for 45 s. The surface microstructure of the samples was then observed using a scanning electron microscope (Hitachi Regulus 8100, Hitachi High-Technologies, Japan) at an accelerating voltage of 3 kV, and images were acquired with using an SE2 secondary electron detector. Multiple fields from different regions of each sample were examined to assess the consistency of the microstructural features. After confirming that the observed morphology was generally consistent across fields, a representative field was selected, and images at 50× and 200× magnifications were acquired from the same area.

2.4. Low-field nuclear magnetic resonance (LF-NMR) analysis

A low-field nuclear magnetic resonance analyzer (MesoMR23-060H-I, Niumag Analytical Instrument Corporation, Suzhou, China) was used to determine the transverse relaxation time (T2) of the samples. The measurement conditions were as follows: resonance frequency 23 MHz, magnetic field strength 0.5 T, coil diameter 25 mm, and magnet temperature maintained at 32 °C. The T2 was acquired using the Carr-Purcell-Meiboom-Gill pulse sequence with the following parameters: 90° and 180° radiofrequency pulse widths of 12.52 μs and 23.52 μs, respectively; waiting time (TW) 3000 ms; echo time (TE) 0.15 ms; number of echoes (NECH) 15,000; and number of signal accumulations (NS) 8.

2.5. HS-SPME-GC–MS analysis

Volatile aroma-active compounds were extracted using HS-SPME-GC–MS, according to a previously reported method with minor modifications (Santos et al., 2020). 2-Octanol was used as the internal standard for quantification. Briefly, 0.5 g of sample was directly transferred into a 20 mL headspace vial containing 10 μL of internal standard solution. After sealing, the vial was equilibrated at 60 °C for 10 min to promote the release of volatile compounds. A DVB/CAR/PDMS (50/30 μm) fiber was then inserted into the headspace vial and exposed to the sample headspace at the same temperature for 15 min. After extraction, the SPME fiber was immediately inserted into the GC–MS injection port and thermally desorbed at 250 °C for 5 min in splitless mode. To prevent cross-contamination, the fiber was reconditioned at 270 °C for 10 min prior to subsequent use.

Gas chromatography was performed on a LECO Pegasus® BT 4D system (LECO, USA) equipped with an Agilent 8890 A gas chromatograph and a time-of-flight mass spectrometer detector. A DB-Heavy Wax capillary column (30 m × 250 μm × 0.5 μm, Agilent, USA) was used with high-purity helium (≥99.999%) as the carrier gas at a flow rate of 1.0 mL/min. The oven temperature program was as follows: initial temperature 40 °C (held for 3 min), ramped at 6 °C/min to 200 °C, then increased at 10 °C/min to 250 °C, and held for 5 min.

Mass spectrometry was conducted in electron impact ionization mode at 70 eV, with both ion source and interface temperatures set to 250 °C. The mass scan range was m/z 35–550, the acquisition rate was 10 spectra/s, and the detector voltage was set to 2014 V. Data acquisition and processing were performed using LECO ChromaTOF software. Compound identification was based on comparison of mass spectra with the NIST mass spectral library in combination with retention indices (RIs). A minimum library similarity score of 700 was required, and the RI deviation was limited to within ±20 RI units. Compounds were assigned Level 2 identification confidence based on MS and RI matching. Volatile compounds were relatively quantified based on peak area ratios to the internal standard.

2.6. GC-IMS analysis

Volatile aroma-active compounds were analyzed using a FlavourSpec® GC-IMS (G.A.S., Dortmund, Germany). A 1.50 g sample was accurately weighed into a 20 mL headspace vial and sealed. Sample introduction was performed with an automatic headspace sampler (CTC Analytics AG, Switzerland). Prior to injection, the sample was incubated at 60 °C for 15 min with shaking at 500 r/min. Subsequently, 200 μL of headspace gas was drawn through a heated and sealed syringe needle (85 °C) and directly injected into the instrument in splitless mode.

GC conditions: An MXT-WAX capillary column (30 m × 0.53 mm, 1.0 μm film thickness, Restek, USA) was used, with the column temperature set at 60 °C. High-purity nitrogen (≥99.999%) was used as the carrier gas under a programmed flow mode: the initial flow rate was set at 2.0 mL/min and held for 2 min, then linearly increased to 10.0 mL/min over 8 min, followed by a further linear increase to 100.0 mL/min over 10 min, and finally maintained at 100.0 mL/min until 30 min. IMS conditions: The drift tube temperature was maintained at 45 °C, the electric field strength was 500 V/cm, and nitrogen was used as the drift gas at a flow rate of 150 mL/min. Tritium (³H) was applied as the ionization source. Compound identification: Normal ketones (C4-C9, purchased from Sinopharm Chemical Reagent Beijing Co., Ltd.) were used as external standards to calculate the retention index (RI) of each volatile compound according to a previously reported GC-IMS method (Jiang et al., 2023). Preliminary identification was based on RI values and comparison of drift times with those of reference compounds in the GC-IMS database. Data acquisition and analysis were carried out using LAV software (Laboratory Analytical Viewer, G.A.S.), with Reporter and Gallery Plot plug-ins applied for fingerprint and differential spectral analyses. Compound identification was performed by cross-referencing both the built-in NIST library and the IMS database for qualitative and relative quantitative analysis.

2.7. Calculation of relative odor activity value (ROAV)

The ROAV was utilized to assess the contribution of individual aroma-active compounds to the overall aroma profile. Odor threshold values for the identified compounds were obtained from relevant book (Van Gemert, 2003). The compound with the highest ROAV was assigned a reference value of 100 (Yang et al., 2024), and the ROAVs of the remaining compounds were calculated accordingly using the following equation.

$$\mathit{ROAVi}\approx \frac{\mathit{Ci}}{C\max }\times \frac{T\max }{\mathit{Ti}}\times 100$$

where Ci represents the relative concentration of each aroma-active compound, expressed as the peak area (or signal intensity) ratio of the compound to the internal standard obtained from GC–MS and GC-IMS analyses, and Cmax is the relative concentration of the most odor-active compound. Ti denotes the sensory threshold of each compound, while Tmax refers to the threshold concentration (mg/L) of the compound with the highest aroma impact. Odor activity value (OAV) was defined as OAVi = Ci/Ti. Accordingly, ROAV was calculated as a relative form of OAV, with the compound exhibiting the highest OAV assigned a reference value of 100.

2.8. Amino acid analysis

Preparation of standards and internal standard solutions: Stock solutions of amino acid standards were prepared in methanol or deionized water, mixed, and diluted with methanol-water (1:1, v/v) containing 10% formic acid to obtain working standards. Trp-d3 (tryptophan-d3) was used as the internal standard, prepared at 1000 ng/mL in the same solvent. All solutions were stored under refrigeration.

Sample preparation: Freeze-dried samples were extracted with 600 μL of methanol-water (1:1, v/v) containing 10% formic acid and vortexed for 60 s. Samples were frozen in liquid nitrogen for 5 min, thawed at room temperature, and subjected to a single freeze-thaw cycle. They were then ground at 55 Hz for 1 min and centrifuged at 12,000 rpm, at 4 °C for 10 min. The supernatant was diluted 50-fold, mixed with 100 μL of internal standard (20 ng/mL Trp-d3), vortexed, and filtered through a 0.22 μm membrane prior to analysis.

Chromatographic and MS conditions: Separation was achieved on an ACQUITY UPLC® BEH C18 column (2.1 × 100 mm, 1.7 μm, Waters, USA) at 40 °C with 5 μL injections. The mobile phases were 50% methanol with 0.1% formic acid (A) and 10% methanol with 0.1% formic acid (B), with a programmed gradient from 90% to 0% B. The flow rate was 0.3 mL/min for 8 min, then increased to 0.4 mL/min. Mass spectrometry was performed in ESI-positive mode with multiple reaction monitoring (MRM). Key parameters were: ion source 500 °C, spray voltage 5500 V, collision gas 6 psi, curtain gas 30 psi, and nebulizer/auxiliary gases 50 psi each.

2.9. Sensory profiling through the E-nose and E-tongue

E-nose analysis: After thawing and equilibration to room temperature, the samples were cut into small pieces. A 10 g portion of each sample was placed in a 100 mL beaker, sealed with plastic film, and equilibrated at room temperature for 1 h before analysis. Olfactory characteristics of the samples were evaluated using a portable electronic nose (PEN3, Airsense Analytics GmbH, Schwerin, Germany) with a headspace sniffing method. The parameters were set as follows: sampling interval 1 s/group, sensor self-cleaning time 80 s, sensor zeroing time 5 s, sample preparation time 5 s, injection flow rate 400 mL/min, and analysis sampling duration 80 s. E-tongue analysis: For sample preparation, 30 g of each sample was mixed with 150 mL of purified water at 60 °C and homogenized for 1 min using a blender. The mixture was then centrifuged at 3000 rpm for 5 min and filtered through gauze, and the clear filtrate was collected for analysis. Taste attributes were assessed using an electronic tongue (TS-5000Z, INSENT, Japan) equipped with six sensors (AAE, umami; CT0, salty; CA0, sour; C00, bitter; AE1, astringent; GL1, sweet) and three standard electrodes. The procedure involved a 30 s measurement with a reference solution to obtain the initial taste values, followed by a 3 s rinse with the same reference solution and then another 30 s measurement for aftertaste evaluation. Each sample was analyzed in triplicate to ensure consistency and reliability of the results.

2.10. Homology modelling and molecular docking

To explore the potential binding interactions between key volatile aroma-active compounds and human olfactory receptors, four common volatiles identified by combined GC–MS and GC-IMS screening with ROAV >1 were selected as ligands: nonanal, phenylacetaldehyde, 2,4-heptadienal, and 3-methylbutanal. Six representative human olfactory receptor proteins-OR1A1, OR1G1, OR2J3, OR1D2, OR5K1, and OR52D1-were chosen as receptors. Due to the limited availability of experimentally resolved crystal structures for human olfactory receptors, their amino acid sequences were retrieved from the UniProt database, and three-dimensional structural models were constructed using the SWISS-MODEL online server based on homologous GPCR templates. The quality of the predicted models was evaluated using Ramachandran plot analysis to ensure reasonable stereochemical properties.

Ligand structures were obtained from the PubChem database and subjected to energy minimization. Molecular docking was performed using PyRx (version 0.8) with AutoDock Vina as the docking engine, and blind docking was adopted in the absence of predefined binding sites. A grid box covering the entire receptor structure was applied to allow full exploration of potential binding regions, and the top ranked docking poses were selected based on the lowest predicted binding energy and reasonable interaction geometry.

Visualization of receptor-ligand complexes was conducted using PyMOL, and the binding modes were further annotated with key interactions, including hydrophobic interactions, hydrogen bonds, and van der Waals forces. Molecular docking was employed solely as an in silico predictive and exploratory approach to assess potential ligand receptor interactions, rather than to demonstrate actual receptor binding, activation, or olfactory perception.

2.11. Statistical analysis

Each thermal treatment was independently performed three times using separate portions of the prepared mushroom samples, generating three independent processing replicates per treatment. Subsequent analyses were performed on these independent replicates (n = 3). One-way analysis of variance (ANOVA) was performed using IBM SPSS Statistics 29.0.1.0, and results were expressed as mean ± standard deviation (SD). Statistically significant differences were indicated by superscript letters (a-c). Data visualization was conducted using Origin 2024 and TBtools-II software. Significance levels were represented by asterisks: * denotes P ≤ 0.05, ** denotes P ≤ 0.01, and *** denotes P ≤ 0.001.

3. Results and discussion

3.1. Texture, microstructure, and water distribution analysis

The texture of food determines the initial oral sensation, and changes in cell walls, protein networks, and pore structures during different thermal processes shape a unique mouthfeel (Li et al., 2021; Paudel et al., 2016). L. asiatica exhibited significant textural differences under different cooking methods, as reflected by multiple texture parameters (Fig. 1A-F). The frying treatment (FT) group showed the firmest texture, with shear force, hardness, gumminess, and chewiness significantly higher than those of the steaming treatment (ST) and boiling treatment (BT) groups (P ≤ 0.05). The fried samples were characterized by a “tough” bite and moderate resilience. The steaming group displayed intermediate hardness and shear force, slightly higher resilience, and a flexible texture. The boiling group had the lowest hardness and shear force, with gumminess and chewiness significantly reduced, whereas springiness and resilience were the highest, resulting in a soft yet springy texture. Textural differences often reflect microstructural variations, therefore, SEM was employed to reveal the structural features of L. asiatica under different cooking processes and to elucidate the morphological basis of textural changes. As shown in Fig. 1G, the fried samples exhibited smooth, dense surfaces without cracks or pores, which are a typical result of protein instant coagulation and cell wall shrinkage as well as cross-linking induced by high-temperature frying (Khalid et al., 2023). Steamed samples showed cracks and reticular pores, resulting in looser structures; boiled samples exhibited the most severe ruptures, with large cracks and fractured networks covering the surface, indicating that prolonged heating in water caused pronounced swelling and structural destruction of cell tissues (Narra et al., 2024). These features were consistent with the textural properties: intact and dense structures provided stronger resistance to deformation during chewing, whereas porous and fractured tissues were more easily fragmented. Similar texture structure relationships have been reported in shiitake mushrooms, where frying increased hardness and chewiness with moisture loss, while steaming and boiling caused marked decreases in hardness accompanied by cooking induced microstructural changes (Chen et al., 2025; Yao et al., 2023).

Fig. 1.

Texture, microstructure, and water distribution of Lanmaoa asiatica under thermal processing. (A) Hardness; (B) Shear force; (C) Springiness; (D) Gumminess; (E) Chewiness; (F) Resilience; (G) SEM; (H) Relaxation time T2 spectrum; (I) Percentage of water distribution.

Because microstructure governs water retention and redistribution, LF-NMR further evaluated water states of L. asiatica under different cooking methods. Based on relaxation times, water was classified as bound water (T2b, 0.01–10 ms), immobilized water (T21, 10–100 ms), and free water (T22, 100–1000 ms) (Wang et al., 2023). As shown in Fig. 1H, the FT group displayed a three-peak distribution, with a small fraction of T2b in addition to T21 and T22. By calculating the peak areas, the proportions of the three water types in each group were obtained (Fig. 1I): In the FT group, immobilized water accounted for the highest proportion (68.81%), free water was lower (31.06%), and bound water was minimal (0.13%). In the ST group, free water increased to 55.15% while immobilized water decreased to 44.85%, whereas the bound water component accounted for only approximately 0.0027% of the total peak area. In the BT group, free water reached the highest level (75.63%) and immobilized water decreased to 24.37%, while the bound water component accounted for approximately 0.0015%. Because the bound water proportions in the ST and BT groups were both below 0.01%, they are displayed as <0.01% in Fig. 1I. A similar water distribution pattern was reported in boiled Boletus species, with one sample showing 76.43% free water and 23.56% immobilized water, consistent with the water distribution observed in the BT group of the present study (Zhang et al., 2025).

In summary, different thermal processes reshaped the structural characteristics of L. asiatica, showing consistent patterns in texture, microstructure, and water distribution: Frying formed a compact outer layer that restricted water migration and retained a higher proportion of immobilized water, contributing to greater hardness and chewiness. In contrast, steaming and boiling resulted in sponge-like reticular structures under moist heat, with the BT group in particular exhibiting a loose, porous, and highly permeable structure, a higher proportion of free water, and a limited capacity to retain immobilized and bound water, leading to a softer texture. Notably, these structural alterations were associated with differences in aroma profiles. Therefore, the following analysis focuses on the flavor attributes of L. asiatica to further explore their potential relationships with structural characteristics.

3.2. HS-SPME-GC–MS qualitative analysis of volatile compounds

3.2.1. Composition and differential analysis of volatile compounds

A total of 60 volatile compounds were identified (Fig. 2A), including aldehydes (19), aromatic hydrocarbons (10), alcohols (9), alkanes (8), heterocyclic compounds (5), olefins (4), ketones (3), esters (1), and others (1). The detailed information for all detected compounds is provided in Table S1. All three thermal processing methods significantly altered the volatile profiles (Fig. 2B, C). Alcohols were dominant in all groups, with the BT group showing the highest proportion (95.22%), likely due to increased porosity and elevated free water content, which facilitated the exchange and enrichment of water-soluble alcohols (Shen et al., 2016). In contrast, the proportions of alcohols decreased in the FT and ST groups, as part of the alcohols underwent oxidation or participated in esterification and Maillard reactions (Hwang et al., 2020). Alkanes accounted for nearly half of the volatiles in FT (45.69%) and ST (51.75%), which were significantly higher than in BT (3.98%). This enrichment originated from lipid thermal degradation and deoxygenation reactions (Grebenteuch et al., 2021), while boiling, due to lower temperatures and leaching of volatiles into the broth, produced and retained fewer alkanes. Aldehydes were most abundant in FT (8.49%) and least in BT (0.18%), primarily arising from lipid oxidation and amino acid chain cleavage (Hidalgo & Zamora, 2016). High-temperature frying was associated with enhanced lipid oxidation and Maillard reactions (Liu et al., 2024), which may have contributed to the higher abundance of short-chain aldehydes, whereas during boiling, aldehydes tended to dissolve into the cooking liquid. Notably, because refined rapeseed oil was used as the frying medium, some lipid derived volatiles detected in the FT group may have originated from both endogenous lipids of L. asiatica and thermal oxidation of the rapeseed oil. Accordingly, these compounds are interpreted as frying associated lipid oxidation products. Heterocyclic compounds and aromatic hydrocarbons were only detected in FT and ST at certain proportions, but were almost absent in BT, as they mainly originated from advanced reactions such as high-temperature Maillard reaction and protein degradation (Kathuria et al., 2023; Wei et al., 2025). A similar processing dependent shift was reported in Boletus edulis, where high temperature cooking, particularly frying, favored the formation of aldehydes and pyrazines, whereas steaming and boiling produced distinct volatile profiles (Wu et al., 2026). Overall, these findings were consistent with structural observations: frying formed a dense crust and favored the retention of thermal reaction products; steaming partially retained the structure while promoting lipid cleavage; boiling, with its loose structure and abundant free water, facilitated flavor leaching and limited the extent of advanced reactions.

Fig. 2.

Analysis of HS-SPME-GC–MS. (A) Categories of volatile compounds; (B) Proportion comparison of different classes of volatile compounds; (C)The chord diagram of volatile compounds; (D) PCA score scatter plot; (E) OPLS-DA score scatter plot; (F) Cross validation plot by 200 permutation tests; (G) Cluster heatmap of volatile compounds.

Principal component analysis (PCA) was performed to systematically evaluate how different thermal processes affected volatile composition (Fig. 2D). PC1 and PC2 explained 26.3% and 21.2% of the total variance, respectively, with clear separation among groups and good intragroup clustering. Orthogonal partial least squares-discriminant analysis (OPLS-DA) was further applied to evaluate the separation of volatile profiles among the three processing groups (Fig. 2E). The model showed R² = 0.99 and Q² = 0.22, and permutation testing showed generally lower R² and Q² values than those of the original model (Fig. 2F), supporting its use for exploratory discrimination among the processing groups without evident overfitting. Hierarchical clustering based on volatile abundances was further performed to assess the overall similarities and differences in volatile profiles among the three processing groups (Fig. 2G). The FT group clustered separately, whereas ST and BT were grouped together, indicating that frying samples were chemically distinct from moist-heat-treated samples. Although ST and BT both underwent moist-heat processing, differences still existed in the abundance of certain compounds. Considering the observed differences in volatile composition, the distinct clustering of the FT group was consistent with its higher aldehyde abundance and the presence of heterocyclic and aromatic compounds, together with the compact structure observed by SEM. In contrast, ST and BT were more closely grouped and both exhibited looser structures and higher proportions of free water, although differences in the abundance of individual volatiles remained. These findings indicate that the three thermal processing methods produced distinct volatile profiles accompanied by corresponding differences in structural and water characteristics.

3.2.2. Identification of key aroma-active compounds

ROAV was calculated based on odor thresholds, and compounds with ROAV >1 were considered key contributors to the overall aroma. (Xie et al., 2023). As shown in Table 1, a total of 19 key volatile compounds were identified in this study, including aldehydes, terpenes, furans, and a few ketones and olefins, which collectively shaped the characteristic flavor of L. asiatica. Among the three groups, the FT group contained the largest variety, whereas the BT group had the fewest. In the FT group, the key compounds could be broadly classified into three categories based on their predominant odor characteristics and reported formation pathways: (1) Lipid thermal degradation products, such as hexanal, 2-heptenal, (E)-2-octenal, 2,4-heptadienal, and 2,4-decadienal, are well known to impart fatty and nutty notes (Franklin & Mitchell, 2019; Xu et al., 2023) and were considered major contributors to the characteristic background aroma of fried L. asiatica; (2) Compounds associated with the interaction between Maillard reactions and lipid oxidation, including 2-amylfuran and 4-oxohex-2-enal, contributed caramel-like and roasted notes (Wu et al., 2025), which are commonly linked to enhanced aroma complexity; (3) l-limonene introduced fresh citrus-like nuances (Ibáñez et al., 2020), which may contribute to balancing the heavier notes generated during frying and improving the overall sensory harmony. High frying temperatures were associated with enhanced lipid oxidation and Maillard reactions, while the formation of a dense surface crust may have facilitated flavor retention, thereby contributing to a multilayered and complex aroma profile.

Table 1.

ROAV of volatile compounds in L. asiatica under different thermal processing methods (GC–MS).

Flavor compound	Flavor description	Threshold values (mg/kg)	ROAV
			FT	ST	BT
n-decane	paraffinic, waxy	3.6	0.45	1.25	0.91
phenylacetaldehyde	honey-like, floral, green	0.003	4.40	18.29	3.42
octylaldehyde	fatty, waxy, citrus-like	0.0008	27.97	8.03	9.51
3-methylbutanal	nutty, roasted, malty	0.0011	20.26	33.84	10.85
nonanal	citrus peel, floral, green grass, oily	0.0011	35.94	100	100
(d)-limonene	citrus, sweet orange, piney	0.006	8.82	95.42	64.56
acrolein	pungent, irritating, spicy, burnt	0.11	0.56	1.20	0.63
(E)-2-Octenal	fatty, green, mushroom-like	0.0025	5.58	0	0
2-nonenal	fatty, green, waxy	0.0001	95.63	0	0
methylheptenone	fruity, floral, green	0.015	1.24	0.28	0.38
2-undecenal	fatty, waxy, citrus-like	0.001	1.85	0	0
l-limonene	piney, citrus-like	0.02	4.97	0	0
2-heptenal, (z)-	fatty, mushroom, green	0.0002	91.70	0	0
2,4-heptadienal	fatty, vegetable, green, cinnamon	0.0006	27.26	0	0
hexanal	green, grassy, fatty	0.005	10.17	9.41	0
2-amylfuran	beany, green, nutty	0.005	13.78	9.36	1.19
2,4-decadienal	fried, fatty, nutty	0.000077	27.86	0	0
trans-2-decenal	fatty, waxy, green	0.0003	100	0	0
4-oxohex-2-enal	fatty, toasty, herb	0.0002	2.82	0	0

In comparison, the ST group exhibited fewer key compounds, mainly including phenylacetaldehyde, which imparted an elegant honey- and floral-like note (Piornos et al., 2020), enhancing aroma delicacy; 3-methylbutanal, which provided roasted and nutty notes (Guan et al., 2022); and hexanal, contributing fatty and grassy nuances (Yang et al., 2025), complementing other volatiles. In addition, 2-amylfuran supplied a light roasted note, while nonanal and (d)-limonene delivered citrus-like freshness (Al-Khalili et al., 2025). Since the steaming temperature was lower than that of frying, pyrazines and high concentrations of unsaturated aldehydes-products of advanced reactions were limited (Xia et al., 2022). However, prolonged heating still induced partial lipid cleavage and amino acid degradation, resulting in a balanced profile characterized by nutty, roasted, and fruity-floral notes. The BT group showed the most simplified profile, with only a few compounds such as 3-methylbutanal, nonanal, and (d)-limonene, and their ROAV values were generally lower than those in the ST group. As a result, only mild nutty, roasted, and citrus-like notes were present, and the overall aroma intensity was weaker. This result corresponded with structural and water distribution observations: boiling increased porosity and free water content, facilitating the migration of volatiles into the broth and restricting advanced reactions, ultimately leading to a thin and less layered aroma profile. Similar findings were reported for cooked Boletus edulis, in which phenylacetaldehyde and 2-amylfuran were identified as characteristic volatile compounds, while boiling resulted in weaker flavor characteristics, consistent with the trends observed in the present study (Yang et al., 2026).

In summary, different thermal processes regulated lipid oxidation, Maillard reactions, and the migration and retention of volatile compounds, thereby generating distinct flavor characteristics: Frying enriched fatty, nutty, and roasted notes, accompanied by fresh citrus nuances, giving L. asiatica the most complex and full-bodied aroma profile. Steaming emphasized floral and nutty notes, yielding a milder yet layered flavor profile. Boiling, however, resulted in a simplified and weakened aroma, with limited complexity and intensity.

3.3. GC-IMS qualitative analysis of volatile compounds

3.3.1. Composition and differential analysis of volatile compounds

Using GC-IMS, 56 volatile compounds were detected across all samples (Fig. 3A), and their detailed information is summarized in Table S2, including aldehydes (21), ketones (11), alcohols (11), esters (5), heterocyclic compounds (3), acids (2), furans (2), and aromatic hydrocarbons (1). The proportions of volatile constituents in L. asiatica were significantly affected by the three thermal processing methods (Fig. 3B). Aldehydes were the most abundant volatiles in all groups, accounting for 55.26% in FT, which was higher than in ST (50.03%) and BT (47.23%). This was likely associated with enhanced lipid oxidation within the frying system and Maillard reactions under high temperature conditions, which favored the formation and accumulation of short chain aldehydes and contributed roasted, nutty, and fruity notes. In contrast, aldehyde generation and accumulation in BT were restricted due to lower temperatures and substantial water migration under open conditions. Ketones were more abundant in ST (23.66%) and BT (22.10%) than in FT (18.52%), as moist-heat treatment tends to promote amino acid decarboxylation and ketonization reactions, providing favorable conditions for ketone formation (Park et al., 2023). Acids were highest in BT (13.17%), suggesting that moist-heat conditions may enhance hydrolytic reactions. Comparable results were reported in cooked Boletus edulis, where GC-IMS revealed clear processing dependent differences in volatile profiles, with high temperature frying favoring aldehyde rich profiles compared with steaming and boiling (Wu et al., 2026).

Fig. 3.

Analysis of GC-IMS. (A) Categories of volatile compounds; (B) The chord diagram of volatile compounds; (C) PCA score scatter plot; (D) Euclidean distance plot; (E) 3D spectrum of volatile compounds; (F) 2D spectrum of volatile compounds; (G) Difference spectrum of volatile compounds; (H) The fingerprints of volatile compounds; (I) Cluster heatmap of volatile compounds.

PCA visualized the overall impact of different thermal processes on volatile composition (Fig. 3C). PC1 and PC2 together explained 87.0% of the total variance. The groups were clearly separated with compact intragroup clustering, indicating pronounced differences in flavor composition among the three processing methods. Furthermore, fingerprint similarity analysis (Fig. 3D) confirmed significant differences in the Euclidean distances of volatile compounds across treatments, reflecting distinct divergences in flavor-related reaction pathways and resulting in unique flavor profiles. This trend was not only evident in the variations of compound categories and abundances but also in the differential activity of associated metabolic pathways, demonstrating that processing temperature and medium type jointly shaped the material basis of the final flavor profile.

3.3.2. GC-IMS spectrum analysis

GC-IMS spectral analysis revealed how different thermal processing methods affected the distribution of volatile compounds in L. asiatica. As shown in Fig. 3E, three-dimensional ion mobility spectra exhibited distinct differences among the groups in terms of volatile compound distribution and peak intensity. For easier comparison, 2D top-view plots were constructed (Fig. 3F), where the background was blue, the vertical red line at 1.0 represented the reaction ion peak (RIP), and colored spots indicated different volatiles, with color intensity reflecting signal strength. The major signals of volatile compounds were concentrated within 1.0–1.6 ms in all groups, but peak areas and signal densities varied. In particular, within the red-framed region, the FT and ST groups displayed denser and stronger signals, whereas the BT group showed markedly weakened responses. Differential spectra were further generated using FT as the reference (Fig. 3G). In region a, both ST and BT differed from FT, while in region b, numerous blue signals in BT indicated significantly lower compound abundances compared with FT. Overall, flavor compound retention in ST was more similar to FT, suggesting that frying and steaming were superior to boiling in retaining volatile compounds, as the latter caused extensive migration and loss.

The GC-IMS fingerprint plots (Fig. 3H) allowed a more intuitive comparison of volatile distribution across treatments. Each processing method produced a unique “chemical fingerprint,” where red indicated higher concentrations and lighter colors represented lower levels. Overall, the FT group showed the densest distribution and strongest signals, followed by ST, while BT exhibited the weakest signals. Further examination revealed that: region a was a characteristic enrichment zone of FT, containing 1-octen-3-one, p-xylene, octanal, 2,4-heptadienal, (E)-2-hexenal, 1-penten-3-ol, pentanal, and propanal, associated with mushroom-like, nutty, fatty, and herbal aromas; region b was the enrichment zone of ST, mainly including 3-methyl-1-butanol, 2-hexenal, 2-butanone, and 6-methyl-5-hepten-2-one, imparting mild roasted and fruity notes; region c was the only relatively prominent area in BT, containing benzaldehyde and cyclohexanone, but both the types and intensities were markedly lower than in the other two groups, indicating that boiling led to substantial flavor loss.

The advantage of heatmap analysis lies in its ability to visualize the distribution and clustering of samples across compound variables, compensating for the inability of fingerprint plots to reflect overall hierarchical relationships. Therefore, cluster analysis was further performed (Fig. 3I). The results showed that FT and ST were clustered together, both exhibiting higher expression of multiple volatiles, whereas BT formed a separate cluster with markedly fewer volatile compounds. Similar differentiation of volatile fingerprints among processing treatments has been reported in shiitake mushrooms, where GC-IMS fingerprint and clustering analyses clearly distinguished samples subjected to different drying conditions (Chen et al., 2021). In summary, the GC-IMS spectrum analysis demonstrated the reshaping effects of thermal processing on the volatile profiles of L. asiatica, with distinct flavor differences observed among the treatment groups.

3.3.3. Identification of key aroma-active compounds

11 key volatile compounds with ROAV >1 were identified (Table 2). The FT group contained the most diverse set of key aroma-active compounds, with generally higher ROAV values. Among them, 1-octen-3-one, the most representative compound, contributed the typical mushroom-like aroma (Yu et al., 2023). This compound is generally associated with linoleic acid oxidation and lipid degradation, and its formation can be promoted under high temperature conditions. Ethyl 3-methylbutanoate and isoamyl acetate imparted sweet fruity notes, likely derived from esterification reactions of Maillard-derived intermediates (Charnock et al., 2022; Holt et al., 2018). Other compounds, such as 3-methylbutanal, methional, 2,4-heptadienal, and butanal, were largely generated through branched-chain amino acid degradation or Maillard reactions, contributing roasted, onion-like, and fatty notes (Chen et al., 2022; Li et al., 2021).

Table 2.

ROAV of volatile compounds in L. asiatica under different thermal processing methods (GC-IMS).

Flavor compound	Flavor description	Threshold values (mg/kg)	ROAV
			FT	ST	BT
phenylacetaldehyde	honey-like, floral, green	0.003	1.88	2.19	1.46
2,4-heptadienal	fatty, vegetable, green, cinnamon	0.0006	7.35	2.09	2.21
1-octen-3-ol	mushroom, lavender, rose, hay	0.0015	0.55	4.46	0.5
methional	onion, meat, fruity	0.0002	10.81	7.66	5.65
nonanal	citrus peel, floral, green grass, oily	0.0011	2.97	2.86	2
1-octen-3-one	mushroom, vegetable, fishy, chicken	0.00003	100	20.62	22.65
octanal	waxy, citrus, orange, fruity, fatty	0.0005	2.22	1.53	1.42
3-methylbutanal	nutty, roasted, malty	0.0011	83.22	72.09	61.36
butanal	pungent, fruity, green leaf	0.002	4.44	3.37	1.61
isoamyl acetate	sweet, banana, fruity	0.0006	0.9	1.89	0.4
ethyl 3-methylbutanoate	apple, banana, sour and sweet	0.00001	93.14	100	100

In the ST group, although the overall ROAV values were slightly lower than those in FT, a considerable number of key aroma-active compounds were still retained, forming a distinctive flavor profile. Phenylacetaldehyde and 1-octen-3-ol were representative contributors: the former, derived from Strecker and phenylalanine degradation, imparted honey-like and floral notes (Monforte et al., 2021); the latter provided a fresh mushroom-like aroma, mainly formed via enzymatic lipid oxidation and mild thermal cleavage (Su et al., 2022; Xiao et al., 2024). Isoamyl acetate and ethyl 3-methylbutanoate both with high ROAV values further reinforced the fruity characteristics. The prolonged moist-heat treatment during steaming promoted tissue softening, facilitating the release of flavor precursors and the gradual transformation of lipids and amino acids, thereby conferring an overall harmonious and mellow flavor. In contrast, the BT group exhibited a markedly reduced number of key aroma-active compounds, with consistently lower ROAV values. This trend was also confirmed by both GC–MS and GC-IMS spectra. Consistent with the key compounds identified here, 3-methylbutanal, isoamyl acetate, and 1-octen-3-one have also been reported as important aroma contributors in boiled Boletus species based on ROAV analysis (Zhang et al., 2025).

In summary, GC-IMS analysis further revealed the remodeling mechanisms of volatile flavors in L. asiatica under different thermal processing conditions. Frying facilitated the efficient transformation of multiple flavor precursors, yielding a rich and multidimensional aroma profile. Steaming generated a harmonious and mellow aroma composition, whereas boiling resulted in significant flavor migration due to loose tissue structures and limited lipid cleavage and Maillard reactions under low-temperature conditions, producing a relatively simple aroma. These results emphasize the decisive role of processing methods in the formation, enrichment, and retention of key aroma-active compounds.

3.4. Combined analysis of HS-SPME-GC–MS and GC-IMS

The combined analysis of GC–MS and GC-IMS offered complementary advantages, overcoming the limitations of single-instrument approaches, and enabling a more comprehensive and accurate characterization of volatile compounds in complex food matrices. Therefore, both techniques were applied in this study to systematically compare the aroma characteristics of L. asiatica under different thermal processing methods. A total of 104 volatile compounds were detected across the three groups, including 60 by GC–MS and 56 by GC-IMS, with 12 compounds identified by both techniques (Fig. 4A). A comparable complementary pattern was observed in cooked Grifola frondosa (Liu et al., 2026), where the combined use of HS-SPME-GC–MS and GC-IMS provided broader volatile coverage than either technique alone. Fewer compounds were detected by GC-IMS, which may be attributed to its ionization and ion-molecule interaction processes and the resulting matrix effects, leading to reduced detectability for some compounds in complex matrices compared with GC–MS (Liu et al., 2025). Further analysis was conducted by combining compounds with ROAV >1 identified from both techniques (Fig. 4B), yielding 26 key volatile aroma-active compounds. Among them, four overlapped between the two methods: nonanal (fatty, fruity), phenylacetaldehyde (honey, floral), 2,4-heptadienal (oily, oxidized), and 3-methylbutanal (nutty, malty). These four shared compounds were major contributors to flavor, exhibited significant variations under different processing methods, and served as important markers for distinguishing the flavor characteristics of L. asiatica.

Fig. 4.

Joint GC–MS and GC-IMS characterization of volatile compounds. (A) Venn diagram of volatile compound types; (B) Venn diagram of key volatile aroma-active compounds (ROAV >1); (C) Flavor wheel of Lanmaoa asiatica under thermal processing.

Based on the combined results of GC–MS and GC-IMS, a flavor wheel was constructed (Fig. 4C), systematically integrating the key volatile compounds and their sensory attributes under different processing methods. The overall aroma spectrum was divided into five categories: fatty; fruity and floral; nutty and roasted; mushroom; green and herbal. The outer circle annotated the representative compounds for each category, while the outermost layer presented the relative contributions of different processing groups, thereby achieving an intuitive unification of aroma composition and quantitative information. Overall, the FT group showed the broadest coverage, with strong contributions particularly in the fatty, fruity, floral, and mushroom categories, resulting in a rich and multilayered aroma profile. The ST group was more prominent in the nutty, roasted, green and herbal categories, presenting a milder and more harmonious flavor profile. By contrast, the BT group showed the most contracted coverage, with lower levels of key aroma compounds, leading to a weaker and more simplified aroma.

The combined analysis substantially improved the completeness of flavor compound profiling and clearly demonstrated the distribution patterns of aroma attributes in L. asiatica under different thermal processing methods: frying enhanced the enrichment and retention of flavors, steaming promoted aroma coordination, and boiling resulted in weaker aroma preservation. These combined results reinforced and extended the insights obtained from GC–MS and GC-IMS individual analyses, providing a comprehensive understanding and strong evidence for the mechanisms of flavor reconstruction in L. asiatica.

3.5. Amino acid analysis

Multivariate statistical analyses were used to comprehensively evaluate how different thermal processes affected amino acid composition. PCA results (Fig. 5A) showed that PC1 (98.8%) and PC2 (1.1%) effectively distinguished the three groups with good intragroup clustering, indicating that thermal processing had a significant impact on amino acid profiles. OPLS-DA further enhanced intergroup separation, and permutation tests confirmed the robustness of the model without overfitting (Fig. 5B, C). The hierarchical clustering heatmap (Fig. 5D) revealed that FT and ST clustered together, while BT formed a separate cluster. Moreover, overall amino acid expression was highest in FT, lowest in BT, and ST at intermediate levels, suggesting that boiling caused the most severe amino acid loss, whereas frying exhibited the best retention. Z-score analysis of amino acid trends across groups (Fig. 5E) further validated these findings: FT was significantly higher than BT, with ST occupying a middle position. Combined with microstructural results, it is inferred that frying rapidly formed a compact crust under high temperature, effectively preventing amino acid leaching, whereas steaming and boiling induced tissue loosening under moist heat, facilitating amino acid dissolution in water or migration with steam, thereby accelerating loss. A similar trend was reported in Lentinus edodes, where boiling reduced free amino acids by 58.6% within 10 min, accompanied by their migration into the cooking broth (Nie et al., 2020).

Fig. 5.

Amino acid profiles of Lanmaoa asiatica under thermal processing. (A) PCA score scatter plot; (B) OPLS-DA score scatter plot; (C) Cross validation plot by 200 permutation tests; (D) Cluster heatmap of amino acids; (E) Z-score plot; (F) Asp; (G) Glu; (H) Gly; (I) Ala; (J) Leu; (K) Phe.

A total of 21 amino acids were quantified in L. asiatica across different treatments, generally exhibiting an overall trend of FT > ST > BT, and detailed contents (μg/g) with statistical comparisons are provided in Table S3. To focus on pathways closely associated with flavor, six representative amino acids were selected for further analysis: umami- and sweet-taste amino acids (Asp, Glu, Gly, Ala) and precursors of key derived volatile aroma-active compounds (Leu, Phe) (Fig. 5F-K). Results showed that umami (Asp, Glu) and sweet (Gly, Ala) amino acids were significantly higher in FT than in other groups, demonstrating the advantage of frying in retaining taste precursors. In the Leu to 3-methylbutanal pathway, FT exhibited a “dual-high” pattern of precursor and product. This phenomenon resulted from frying-induced crust formation that inhibited Leu leaching, while high temperature and low water activity promoted Strecker degradation (Piornos et al., 2025), achieving a synergistic effect of “high retention and high conversion.” In contrast, both ST and BT showed low levels due to leaching and insufficient reactions. In the Phe to phenylacetaldehyde pathway, a different trend was observed: precursors were highest in FT, whereas products were highest in ST. The moderate temperature and high humidity of steaming promoted Phe degradation while suppressing further transformation of its products, allowing phenylacetaldehyde to accumulate (Lin et al., 2023). Although frying retained Phe, its high-temperature, low-moisture environment facilitated further condensation, polymerization, and reactions with Maillard intermediates to generate complex heterocycles and polymers (Marrufo-Curtido et al., 2022), thus reducing phenylacetaldehyde content. In BT, both precursor and product were at low levels due to leaching and insufficient reactions.

Overall, frying achieved both high retention and efficient conversion of amino acids through crust densification and coupled thermal reactions. Steaming favored the stable accumulation of aromatic amino acid products, while boiling caused severe losses of amino acids and derived volatile aroma-active compounds due to extensive leaching and insufficient reactions. These patterns were highly consistent with microstructural observations, water distribution, and GC–MS/GC-IMS results, further validating the proposed “structure-precursor-product” synergistic mechanism.

3.6. E-nose, E-tongue, and correlation analysis

3.6.1. E-nose and E-tongue analysis

The E-nose and E-tongue simulate human olfaction and taste, enabling objective characterization of volatile and non-volatile components. The E-nose radar plots (Fig. 6A) and PCA results (Fig. 6B) revealed significant differences in odor profiles among the treatment groups, with PC1 and PC2 together explaining 93.8% of the total variance. Prominent responses were observed in W5S (broad-spectrum components), W1S/W1W (terpenes and aromatic compounds), and W2S/W2W (alcohols, aldehydes, and ketones), with the ST group exhibiting overall stronger signal intensities than the FT group, particularly in channels associated with polar small molecules. Processing dependent E-nose differentiation has also been reported in cooked Boletus edulis, where different cooking methods markedly altered sensor responses and produced distinct odor profiles (Yang et al., 2026). It should be noted that these results are not contradictory to the GC–MS and GC-IMS findings indicating a higher total abundance of volatile compounds in the FT group. GC–MS and GC-IMS primarily reflect the overall composition and potential generation capacity of volatile compounds in the samples, whereas the E-nose, operating under relatively mild and static conditions, mainly captures the instantaneous release of volatiles into the gas phase that are detectable by the sensors (Burlachenko et al., 2016). In the ST group, steaming resulted in a relatively loose microstructure and higher free water content, which facilitated the release of polar small molecules such as ketones and alcohols into the gas phase. In addition, elevated moisture levels exerted a signal-amplifying effect on MOS-type sensors (Liu et al., 2021). Together with GC-IMS results showing higher relative proportions of ketones and alcohols in the ST group, these factors further enhanced the responses of the corresponding E-nose channels. By contrast, although the FT group exhibited a higher overall level of volatile compounds, it was enriched in compounds with relatively higher molecular weight and lower polarity, including aldehydes, alkanes, and advanced Maillard reaction products. The compact structure formed during frying restricted the rapid release of these compounds under E-nose detection conditions, resulting in comparatively weaker sensor responses (Liu et al., 2023). Therefore, chemical analyses revealed differences in the overall volatile composition among treatments, whereas the E-nose reflected the instantaneous release and perception of volatile compounds under specific detection conditions. The two approaches provided complementary information, and their trends were consistent with the observed microstructural differences.

Fig. 6.

Analysis of electronic nose (E-nose), electronic tongue (E-tongue), and mantel test correlation. (A) Radar chart of e-nose; (B) PCA of e-nose; (C) Radar chart of e-tongue; (D) PCA of e-tongue; (E) Correlation between e-nose and key volatile compounds; (F) Correlation between e-tongue and Amino acids.

The E-tongue radar plots (Fig. 6C) showed significant differences among the three groups in both basic tastes and aftertaste attributes. The FT group exhibited stronger responses in saltiness, umami, aftertaste, and astringency (umami >9, astringency >7), whereas the ST and BT groups showed stronger responses in sweetness. Similar processing induced changes in E-tongue responses have been observed in Lentinus edodes, where thermal treatments altered taste attributes together with taste active amino acids and nucleotides (Li et al., 2017). These differences were consistent with previous findings: the FT group retained more taste precursors (Asp, Glu, Gly, Ala) and enriched aldehydes and advanced reaction products, thereby enhancing aftertaste and astringency. Moreover, the compact crust and low water activity formed during frying suppressed amino acid leaching, favoring the expression of saltiness and umami. In contrast, the looser structures and higher free water content of ST and BT facilitated the dissolution of water-soluble sweet compounds, which were readily captured by the E-tongue, resulting in stronger sweetness responses. PCA analysis (Fig. 6D) showed that the first two principal components explained 87.4% of the total variance, clearly separating the samples. These results indicate that E-tongue phenotypes closely aligned with microstructural and chemical composition changes, and that different thermal processes reshaped taste characteristics through the combined effects of retention, leaching, and reaction transformations.

3.6.2. Correlation analysis

The Mantel test was used to evaluate associations between molecular features and sensory phenotypes. In the network, line width reflects correlation strength (Mantel's r), and color indicates statistical significance (Mantel's p). Correlation analysis between E-nose responses and key volatiles showed (Fig. 6E), channels W2W, W2S, W1W, W1S, W3C, and W5S showed significant positive correlations with 3-methylbutanal, phenylacetaldehyde, isoamyl acetate, butanal, 1-octen-3-ol, 2-amylfuran, and hexanal. Combined with chemical analysis, it was observed that the FT group was characterized by short-chain aldehydes (3-methylbutanal, butanal, hexanal) and heterocyclic compounds (2-amylfuran), which markedly enhanced responses in W1S, W5S, and W1W channels. In contrast, the ST group was dominated by alcohols and esters (1-octen-3-ol, isoamyl acetate) and aromatic aldehydes (phenylacetaldehyde), which strengthened W2S, W2W, and W1W signals. This pattern was consistent with the differences in aldehyde, alcohol, and ketone proportions detected by GC–MS and GC-IMS, as well as with the regulatory effects of microstructural changes on release rates. The correlation between E-tongue responses and amino acids (Fig. 6F) further supported this flavor reconstruction model: umami, saltiness, and astringency were significantly correlated with Phe, Glu, Asp, and Leu, while frying inhibited leaching and promoted Maillard reactions, thereby enriching Glu, Asp, and Leu and enhancing both umami intensity and fullness. Steaming, on the other hand, promoted the conversion of Phe into aromatic compounds and retained high-sweetness amino acids such as Ala and Gly, shaping a flavor profile dominated by floral-fruity and sweet attributes. Associations between flavor compounds and sensory responses have also been reported in other complex food systems using Mantel analysis, involving volatile compounds, amino acids, E-nose, E-tongue, and sensory attributes (Ping et al., 2025).

Taken together, the correlation networks not only validated the strong consistency between key chemical constituents and E-nose/E-tongue sensory features under different thermal processes, but also revealed the molecular basis of aroma and taste reconstruction in L. asiatica from the perspectives of dynamic release mechanisms and precursor transformation pathways.

3.7. Molecular docking analysis of key volatile compounds and olfactory receptors

To gain in silico insights into potential ligand-receptor interactions that may be associated with aroma differences of L. asiatica during thermal processing, we docked four key volatile compounds (nonanal, phenylacetaldehyde, 2,4-heptadienal, and 3-methylbutanal) with six representative human olfactory receptors (OR1A1, OR1G1, OR2J3, OR1D2, OR5K1, and OR52D1) (Fig. 7). Predicted binding affinities (kcal/mol) were used to evaluate binding feasibility and relative interaction strength. Typically, binding energies below −4.25 kcal/mol indicate binding potential, while values below −5.0 kcal/mol suggest more favorable predicted binding (Tian et al., 2025). The docking analysis estimated binding energies and predicted hydrophobic and hydrogen bonding interaction sites (Table 3). The results showed that phenylacetaldehyde exhibited the lowest average binding energy (−5.28 kcal/mol), with comparatively stronger affinities toward OR1A1 (−6.0), OR1D2 (−5.5), OR2J3 (−5.4), OR1G1 (−5.2), and OR52D1 (−5.0). 2,4-Heptadienal displayed stronger affinities with OR1A1 (−5.4) and OR5K1 (−5.0). Nonanal bound most strongly to OR1D2 (−4.8) and OR52D1 (−4.7). Although 3-methylbutanal displayed weaker overall affinity, it still showed binding potential with OR1A1 (−4.7). From the receptor perspective, OR1A1 presented the lowest average binding energy (−5.18), suggesting a relatively higher binding propensity toward both aromatic and aliphatic aldehydes among the receptors tested.

Fig. 7.

Molecular docking of human olfactory receptors with key volatile aroma-active compounds.

Table 3.

Summary of the interaction forces between ligands and olfactory receptors.

Ligands	Reporter	Docking score (kcal/mol)	Hydrophobic interactions	Hydrogen bonds
nonanal	OR1A1	−4.6	LEU262	ASN176
phenylacetaldehyde	OR1A1	−6	TYR258, VAL203	/
2,4-heptadienal	OR1A1	−5.4	TYR258, PHE206, VAL203, VAL254	TYR250
3-methylbutanal	OR1A1	−4.7	VAL254, TYR258, VAL203, PHE206	ASN109, ASN155
nonanal	OR1G1	−4.3	ILE142, PHE61	/
phenylacetaldehyde	OR1G1	−5.2	PHE252	/
2,4-heptadienal	OR1G1	−4.1	VAL211, PHE260, PHE256	SER255
3-methylbutanal	OR1G1	−3.6	PRO58, PRO138, ILE142, CYS141, PHE61
nonanal	OR2J3	−4.5	TYR281, PHE254, ALA258	THR112
phenylacetaldehyde	OR2J3	−5.4	LEU202, PRO185, PRO266	ARG189
2,4-heptadienal	OR2J3	−4.3	PHE254, ALA258, TYR281	/
3-methylbutanal	OR2J3	−4.1	VLA184, LEU188	HIS162, THR166
nonanal	OR1D2	−4.8	CYS256, LEU255	LEU199
phenylacetaldehyde	OR1D2	−5.5	PHE207, TYR259, LEU255	CYS256
2,4-heptadienal	OR1D2	−4.8	LEU255, TYR252, CYS256, LEU208, PHE207	GLY203, TYR155
3-methylbutanal	OR1D2	−3.9	LEU255, TYR259, PHE207	/
nonanal	OR5K1	−4.2	LEU264, PHE200, LEU199	LEU264, LEU265
phenylacetaldehyde	OR5K1	−4.6	PHE245, LEU222	/
2,4-heptadienal	OR5K1	−5	TYR259, LEU199, ILE181, PHE101	PHE200, SER203, GLY204
3-methylbutanal	OR5K1	−4.3	LYS90, PHE17, PHE85	PHE177, ASN84
nonanal	OR52D1	−4.7	PHE262, LEU207, ALA206	SER261
phenylacetaldehyde	OR52D1	−5	TYR111, ALA258	TYR111
2,4-heptadienal	OR52D1	−4.9	PHE262, PHE259, LEU207	ALA258
3-methylbutanal	OR52D1	−3.9	MET210, ALA258, PHE262, PHE259	/

Predicted binding mode analyses indicated that the docking poses of the four key volatiles were mainly characterized by hydrophobic interactions and aromatic stacking, with hydrogen bonding also contributing to the predicted interactions. For phenylacetaldehyde-OR1A1 (−6.0 kcal/mol), the benzene ring formed aromatic-hydrophobic interactions with Tyr258, while Val203 contributed to hydrophobic confinement. In 2,4-heptadienal-OR5K1, the carbon chain fitted into a hydrophobic channel, while the aldehyde group formed hydrogen bonds with Gly204, Phe200, and Ser203. In nonanal-OR1D2, the long chain was embedded within hydrophobic residues (Cys256 and Leu255), with the carbonyl group forming a weak hydrogen bond with Leu199. In 3-methylbutanal-OR2J3, the ligand was located in a hydrophobic pocket formed by Val184 and Leu188, and the carbonyl group formed hydrogen bonds with His162 and Thr166. Overall, these results suggest that hydrophobic embedding, aromatic interactions, and hydrogen bonding may jointly contribute to the stability of the predicted complexes. Comparable computational studies of food odorants have likewise identified hydrogen bonding and aromatic interactions as important contributors to predicted odorant receptor interactions (Zeng et al., 2023).

When interpreted together with the chemical and sensory-related measurements, the docking results provide supportive, in silico indications of potential interactions between several aldehydes and selected human olfactory receptors. For example, ST was associated with higher phenylacetaldehyde levels, which showed relatively strong binding affinities to multiple receptors (e.g., OR1A1) in docking simulations; this may be linked to the floral-fruity impressions reflected by the e-nose responses and the flavor wheel. Similarly, FT showed increased levels of 2,4-heptadienal, nonanal, and 3-methylbutanal, which exhibited binding potential toward receptors such as OR1A1, OR5K1, OR1D2, and OR52D1 in silico, potentially consistent with the fatty, nutty, and roasted notes observed. In contrast, BT presented lower levels of these volatiles, which may partially explain its relatively weaker aroma intensity. Overall, molecular docking provided computational predictions of potential interactions between key volatile compounds and selected human olfactory receptors, which may help interpret the observed aroma differences. It should be noted that these predictions do not demonstrate actual receptor binding, activation, or sensory perception in biological systems. Accordingly, the docking analysis was used solely as an exploratory and supportive approach to complement the chemical and sensory related findings. Future studies should incorporate experimental receptor validation to further substantiate the predicted ligand receptor interactions and their relevance to aroma perception.

4. Conclusion

This study systematically elucidated the cross-level regulatory mechanisms by which frying, steaming, and boiling reshape the structure, water status, generation of key volatile aroma-active compounds, and sensory perception in L. asiatica. The results demonstrated that frying formed a compact outer crust and increased the proportion of immobilized water, whereas steaming and boiling led to looser structures and elevated free water levels. Structural integrity and water status jointly influenced the retention and release behavior of flavor precursors. GC–MS and GC-IMS individually profiled the volatile spectra under different processing methods and constructed flavor wheels, revealing significant differences in compound categories and relative abundance distributions. Their complementary data enabled the joint identification of four common key volatile aroma-active compounds—nonanal, phenylacetaldehyde, 2,4-heptadienal, and 3-methylbutanal. These compounds highlighted that frying intensified lipid cleavage and Maillard reactions, steaming facilitated the formation of aromatic aldehydes, while boiling exhibited a “dual-low” feature of both precursors and flavor products. The correlation analysis linked volatile and amino acid profiles with the responses of the E-nose and E-tongue, confirming the strong consistency between chemical composition and sensory phenotypes. Furthermore, molecular docking suggested favorable binding interactions between the four key compounds and specific olfactory receptors (OR1A1, OR1G1, OR2J3, OR1D2, OR5K1, and OR52D1), primarily through aromatic stacking, hydrophobic embedding, and limited hydrogen bonding. Overall, this study established an integrated framework linking structural characteristics and water distribution with chemical composition changes and potential receptor interactions. These findings provide scientific support and technical insights for the quality optimization of edible mushrooms and the improvement of thermal processing techniques, with particular relevance to structure-related effects on flavor. The absence of an uncooked control limits the direct assessment of changes relative to the native state. In addition, because frying, steaming, and boiling differed simultaneously in temperature, heating time, processing medium, and moisture, the observed changes reflect the integrated effects of each processing method rather than any single factor. Molecular docking provides only in silico evidence of potential interactions between aroma compounds and olfactory receptors and requires further experimental validation. Future studies should include matched fresh samples from the same batch to trace structural and flavor evolution, isolate key processing variables, and experimentally validate receptor interactions to strengthen the mechanistic interpretation.

CRediT authorship contribution statement

Jingchuan Zheng: Writing – review & editing, Writing – original draft, Visualization, Methodology, Investigation, Formal analysis. Lei Peng: Writing – review & editing, Investigation, Formal analysis, Data curation. Muhammad Aaqil: Resources, Investigation, Data curation. Xinyu Wei: Methodology, Investigation, Formal analysis. Feng Zhang: Software, Formal analysis, Data curation. Yang Tian: Supervision, Project administration, Investigation, Funding acquisition, Conceptualization. Cunchao Zhao: Supervision, Project administration, Methodology, Investigation, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors affirm that they have no financial, personal, or other relationships that could be perceived as influencing the outcomes of this study.

Appendix A. Supplementary data

Supplementary material 1

(Detailed information on all compounds detected by HS-SPME-GC-MS.).

Supplementary material 2

(Detailed information on all compounds detected by GC-IMS.).

Supplementary material 3

(Contents (μg/g) of 21 amino acids with statistical comparisons among groups.).

Data availability

Data will be made available on request.