Investigating potential freshness indicators for packaged oysters through multivariate statistical analysis

Abstract

Pacific oysters (Crassostrea gigas) are globally renowned shellfish. In South Korea, oysters are commonly packaged with filling water in polyethylene bags. Previous studies have proposed various freshness and quality parameters for oysters, including pH, volatile basic nitrogen content, glycogen content, and viable cell count. We aimed to identify the objective indicators of oyster freshness during storage and distribution using metabolomic analysis and multivariate statistical techniques. Packaged oyster samples were analyzed for metabolites using gas chromatography-mass-spectrometry during 9-days storage at 5 °C or 15 °C. Additionally, the pH, turbidity, and soluble protein content of the filling water were measured. Multivariate statistical analyses, including principal component analysis, partial least squares discriminant analysis, and orthogonal projections to latent structure-discriminant analysis revealed statistically significant results, demonstrating metabolite clustering based on storage duration. In conclusion, this study introduced crucial freshness indicators for stored or distributed oysters by utilizing metabolomic analysis and multivariate statistical techniques.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10068-024-01693-y.

Introduction

Oysters are esteemed for their exquisite flavor and unique texture and are one of the most popular seafoods worldwide and are predominantly consumed raw. Crassostrea gigas is the main species consumed. Oysters are recognized for their nutritional richness, comprising essential elements such as protein, calcium, iron, and zinc (Maurya, 2021). Notably, the substantial glycogen content of oysters enhances the digestive processes and facilitates absorption. These constituents have diverse physiological activities (Son et al., 2014). Nevertheless, their susceptibility to alterations in quality and freshness is attributed to their elevated water content, delicate muscular structure, and autolytic processes. Moreover, there are alterations in freshness and quality during transportation and storage (Chen et al., 2019). Generally, the quality and freshness indicators of oysters include texture, color, flavor, pH, and adenosine triphosphate (ATP)-related compounds (Dong et al., 2023). Packaged oysters are the most easily accessible products for consumers in Korea and are typically wrapped in polyethylene films containing approximately 150 g of oyster meat and approximately 200 mL of filling water. Methods for assessing freshness or quality may slightly differ from those used for general oysters owing to the specific characteristics of the packaged product. In particular, measurements such as the turbidity of the filling water or the quantification of soluble proteins in the filling water can be employed to assess the quality of packaged oysters. Although these indicators are used to evaluate oyster quality and freshness, research identifying specific markers that clearly indicate oyster freshness remains limited. Given the predominant consumption of oysters in their raw state, verification of freshness is of paramount importance. As previously mentioned, parameters such as pH, glycogen content, turbidity, and soluble proteins in the filling water have been utilized as quality indicators of oysters in previous studies. However, few studies have identified markers representing the freshness of packaged oysters. Changes in oyster compound composition are closely associated with volatile organic compounds (VOCs) (Guan et al., 2022). In oysters with a balanced composition of proteins, fats, and carbohydrates, various volatile compounds can be detected depending on the storage temperature and duration. These VOCs have the potential to serve as novel indicators to evaluate and determine oyster freshness. Previous studies have highlighted the importance of volatile organic compounds in assessing seafood quality (Cheng et al., 2023; Dong et al., 2023; Jeong et al., 2021). Recently, integrating multivariate statistical methods and advanced analytical techniques has become an effective and powerful approach to investigate various changes in endogenous metabolites (Cai et al., 2022; Mashabela et al., 2022). Several studies have profiled seafood metabolites using gas-chromatography (GC–MS) (Cheng et al., 2023), liquid chromatography-mass spectrometry (LC–MS) (Zhuang et al., 2022), and nuclear magnetic resonance (NMR) (Bodin et al., 2022). These strategies are useful for tracing specific metabolite pathways and are efficacious for investigating novel indicators of food quality or freshness. This study aimed to suggest the potential freshness indicators of packaged oysters through GC–MS and multivariate statistical analysis. In addition, we aimed to identify latent candidate indicators through multivariate statistical analysis and present the most robust indicators by analyzing their correlation with existing quality parameters.

Materials and methods

Sample preparation and storage conditions

Packaged Pacific oysters (Crassostrea gigas) were purchased from a seafood company (DAIONE FOOD Co., Ltd) in Tong-Yeong, Republic of Korea. The packaged oyster products contained 150 g oyster meat and approximately 200 mL filler water. The filler water was filtered and mixed with seawater and tap water in a 1:1 volume ratio. Each oyster weighed approximately 10–15 g and each product contained 10–15 oysters on average. Packaged oyster samples were prepared on the same day as the beginning of storage, and kept below 10 °C for transport to the laboratory. One packaged oyster was used for each experiment condition and the sample was stored in an incubator at 5 °C or 15 °C for 9 days. During storage, oyster samples were collected on days 0, 1, 2, 3, 4, 7, and 9 days. The packaged oysters were opened and separated into filling water and oysters, and the filling water was used to measure turbidity and soluble protein content. The separated oysters were drained for 10 min before use.

pH, turbidity, and soluble protein

The pH of the oysters, turbidity, and soluble protein content of the filling water were measured to determine general quality changes in packaged oysters during storage at different temperatures. A pH meter (Orion Star A211; Thermo Fisher Scientific, Waltham, MA, USA) was used to measure the change in the pH of the raw oysters according to the storage temperature and duration. One piece of raw oyster (10–15 g) was mixed with 9-times the volume of distilled water, homogenized, and used for pH measurements. The soluble protein content of the filling water was determined using the biuret assay (Jiang et al., 2019). One milliliter of water was mixed with 4 mL of Biuret reagent (Sigma Aldrich, St. Louis, MO, USA) and kept at room temperature(25.0 ± 1.0 °C) for 30 min. The absorbance was measured at 540 nm using a microplate reader (Biotek, Synergy HTX). The soluble protein content was quantified using a standard curve prepared with bovine serum albumin (Sigma-Aldrich) as the standard protein. The turbidity of the filling water was measured using a portable turbidity meter (Hanna HI 93414; Hanna Instruments, Woonsocket, RI, USA), dispensing approximately 10 mL of the filling water into a dedicated container, and expressing the result as 0–1000 nephelometric turbidity units (NTU).

Metabolite extraction

One mL of tertiary distilled water containing the internal standard 2-methylpentanoic acid and 1 g of NaCl were added to 500 μL of the sample solution. The sample solution was prepared by homogenizing five oysters, each weighing approximately 10–15 g, for 2 min using a stomacher (BagMixer 400 CC; Interscience, St. Nom, France). Subsequently, the solution was placed in a 20 mL headspace glass vial with a magnetic bar, sealed with a cap, and reacted at 550 rpm for 10 min at 20 °C to extract the gas components. The extracted components were captured using solid phase microextraction (SPME) fibers (50/30 μm, DVB/ CAR/PDMS Stableflex, Dupleco Inc., Bellefonte, PA, USA) for 10 min and injected into the GC/MS for desorption by exposing the fibers at the injection port for 10 min at 250 °C.

Gas chromatography-mass spectrometry

Volatile metabolite extracts were analyzed using a Shimadzu GC-2010 Plus instrument (Shimadzu, Kyoto, Japan). The volatile metabolites absorbed in the SPME fiber and derivatized samples were injected into a DB-WAX capillary column (30 m × 0.25 mm i.d. × 0.25 μm film thickness, J&W scientific, Calif, USA) with splitless mode. For volatile metabolites, the injector temperature was set to 230 °C, the oven temperature program was initiated at 40 °C for 2 min, followed by increasing the temperature at 7 °C/min to 100 °C, increasing the temperature by 20 °C/min to 240 °C, and holding the temperature at 240 °C for 4 min. Helium was used as a carrier gas at a flow rate of 1 mL/min. The GC column effluent was analyzed using a Shimadzu GCMS-TQ 8030 mass spectrometer (Shimadzu, Kyoto, Japan). The GCMS-TQ 8030 MS/MS system consists of a Shimadzu GC-2010 Plus gas chromatograph, a Shimadzu TQ 8030 triple quadrupole mass spectrometer, and a Shimadzu AOC-5000 Plus sample handling system (Shimadzu Europa GmbH, Duisburg, Germany) A value of 70 eV was used for ionization. The ion source and interface temperatures were 200 °C and 250 °C, respectively. Data monitoring was performed in full-scan mode (m/z 40–550). The scan event time and velocity were 0.03 s and 2,000 amu/s, respectively. For quality control, a mixture of all non-volatile samples was injected every five samples.

Data processing for multivariate statistical analysis and correlation analysis

The peak areas of the metabolic compounds in each sample were imported into SIMCA 13 software (Umetrics, Umea, Sweden). Data were normalized (UV scaling) for statistical analysis using the normalization function provided by SIMCA 18. Principal component analysis (PCA) was performed to assess inherent variations among samples. Orthogonal partial least squares discriminant analysis (OPLS-DA) was employed as a supervised approach to analyze the maximum variance between samples. The quality of the models was characterized by the R₂ and Q₂ values. R₂ represents the proportion of variance in the data explained by the models and signifies goodness-of-fit, whereas Q₂ denotes the proportion of variance predicted by the model, indicating predictability (Mahadevan et al., 2008). The model was validated using 200 random permutations. In the OPLS-DA model, metabolites with variable importance in projection (VIP) scores exceeding 1.0 and p-values below 0.05 were deemed indicative of significant changes. Pearson’s correlation coefficient is a statistical measure that quantifies the linear relationship between two variables. It is calculated based on data assumed to follow a Gaussian (normal) distribution and ranges from − 1 to 1, with higher absolute values indicating a stronger correlation. Pearson’s correlation analysis between the identified compounds and the physicochemical results was performed using the Statistical Package for the Social Sciences program (ver. 17.0, SPSS Inc., Chicago, IL, USA).

Statistical analysis

SIMCA version 18.0.1 (Umetrics, Umeå, Sweden) was used to analyze the GC–MS data, and PLS-DA was used to visualize the differences between the sample groups. The quality of the PLS-DA model was evaluated using three parameters (R₂X and R₂Y: goodness-of-fit measures, Q2: predictive ability, and p-value) and validated by a permutation test (n = 200). All data were analyzed by one-way analysis of variance (ANOVA) with Duncan’s test (p < 0.05) using SPSS 17.0 (SPSS Inc., Chicago, IL, USA).

Results and discussion

Changes in quality of packaged oyster (pH, turbidity, and soluble protein)

Variations in oyster pH (Fig. 1A), filling water turbidity (Fig. 1B), and soluble protein content in the filling water (Fig. 1C) were assessed to measure changes in oyster freshness and quality during storage. Various quality indicators in oysters during the storage period occurred linearly and significantly, as reported in previous studies (Jeong et al., 2021; Lee et al., 2020; Min et al., 2020; Son et al., 2014). The pH consistently decreased during the storage period, with a minimal decrease to 5.360 after 9 days at 5 °C, and 4.515 at 15 °C by the 7th day. Oyster pH levels are indicative of freshness; for instance, pH 6.3 or higher is considered "very good," 6.2–5.9 is "good," 5.8 is "off," 5.7–5.5 is "musty," and 5.2 or less is judged "sour" or "putrid" (Pottinger, 1948; Son et al., 2014). Initially, packaged oysters were in the "very good" or "good" stage, maintaining quality with a pH above 6.0 until the 5th day at 5 °C. However, the pH reached 5.665 following storage at 15 °C on the 2nd day and indicated significant quality deterioration with temperature. The turbidity of the filling water changes in packaged oysters stored at 5 °C was minimal; however, those stored at 15 °C sharply increased, exceeding 5,000 NTU on the 7th day. Turbidity in the filling water is closely related to oyster freshness and is highly correlated with other quality indicators (Son et al., 2014). Soluble protein contents in filling water linearly increased during the storage period, with a more pronounced increase and rapid quality change observed at 15 °C storage condition. Changes in these fundamental quality indicators have been reported in various studies, and similar trends were observed in the current study. These results indicate that the oyster samples used in the study were fresh and exhibited minimal variation between individual samples.

Fig. 1.

Changes in pH of oyster (A), turbidity of filling water (B), and soluble protein of filling water (C) during storage at 5℃ and 15℃ (●, stored at 5 °C; ■, stored at 15 °C

Metabolic compound identification by GC–MS

The results of the identification of changes in volatile compounds in packaged oysters during storage using GC–MS are presented in Table 1 and Suppl. Data 1. Gas chromatography–mass spectrometry analysis of volatile compounds during the storage period identified 27 compounds. These substances are classified as follows: 13 belong to organic oxygen compounds, 4 to lipids and lipid-like molecules, 3 to organosulfur compounds, 2 to organic nitrogen compounds, 2 to organic acids and derivatives, 1 to fatty acids, 1 to benzenoids, and 1 to hydrocarbons. Various subclasses of substances were identified (Table 1), including seven carbonyl compounds, five fatty alcohols, five alcohols and polyols, two alkyl thioethers, one amine, one organic cyanide, one toluene, one aldehyde, one carboxylic acid, one branched unsaturated hydrocarbon, one amino acid, peptides and analogs, and one sulfoxide. Reflecting on the characteristics of GC–MS and SPME analyses, many substances related to aroma and flavor were identified, and the flavor DB (http://cosylab.iiitd.edu.cn/flavordb) was used to identify the aroma components of the substances (Garg et al., 2018). The results are summarized in Suppl. data 2. Carbonyl compounds (hexanal, 3-octanone, 1-octen-3-one, trans-2-octenal, and 2,6-nonadienal) primarily represent the aromatic components in plant categories such as herbal, fresh, fruity, and leafy. Alcohol and polyols (1-propanol, 1-penten-3-ol, and 1-pentanol) indicated aroma components in the pungent, fusel, and musty categories. Fatty alcohol (3-octen-2-ol, 1-tetradecanol, and 1-octen-3-ol) were associated with aroma components in the waxy, coconut, melon, and mushroom categories, particularly 1-octen-3-ol, which was related to aroma components in raw, fishy, fungal, and oily seafood. These results were similar to those reported in various studies that analyzed volatile compounds in oysters using GC/MS-SPME (Fratini et al., 2012; Fu et al., 2023; Zhang et al., 2009). In particular, volatile compounds such as 1-penten-3-ol, 2,6-nonadienal, 3-octanone, propionic acid, and dimethyl sulfide were identified in a comparison of fresh and non-fresh oysters (Zhang et al. 2009), which is similar to the results of this study. GC/MS with SPME analysis did not involve direct sample injection into the GC-injector. Therefore, only volatile compounds are concentrated in SPME and produced as a result. (Zhang et al., 2009). The volatile compounds identified in this study have a high potential to be utilized as biomarkers to indicate the freshness of packaged oysters. Furthermore, multivariate statistical analysis was conducted to discern and ascertain robust freshness indicators with greater reliability.

Table 1.

Identification of major metabolites contributing the separation among sample groups

No	RT (min)	Compound	Class		p-value
			Super class	Sub class
1	1.42	2-Pentanamine	Organic nitrogen compounds	Amines	5.32E−02
2	1.81	Dimethyl sulfide	Organosulfur compounds	Dialkyl thioethers	1.70E−06
3	3.1	3,5-Octadiene / 3-Octen-2-ol	fatty acid	Fatty alcohols	1.61E−05
4	3.22	Ethanol	Organic oxygen compounds	Alcohols and polyols	1.69E−13
5	3.73	3-Pentanone	Organic oxygen compounds	Carbonyl compounds	1.36E−08
6	4.06	Methyl cyanide	Organic nitrogen compounds	Organic cyanides	2.14E−10
7	4.68	Methylbenzene	Benzenoids	Toluenes	4.57E−03
8	4.86	1-Propanol	Organic oxygen compounds	Alcohols and polyols	4.68E−15
9	5.29	Methyl disulfide	Organosulfur compounds	Dialkyl thioethers	2.33E−11
10	5.47	Hexanal	Organic oxygen compounds	Carbonyl compounds	6.55E−05
11	5.79	3,5-Octadien-2-ol	Lipids and lipid-like molecules	Fatty alcohols	3.07E−04
12	6.41	2-Pentenal	Organic oxygen compounds	Carbonyl compounds	2.75E−04
13	6.7	3-Hexenal	Organic oxygen compounds	Aldehydes	8.08E−03
14	7.19	1-Penten-3-ol	Organic oxygen compounds	Alcohols and polyols	1.91E−07
15	9.08	3-Octanone	Organic oxygen compounds	Carbonyl compounds	8.36E−10
16	9.08	1-Pentanol	Organic oxygen compounds	Alcohols and polyols	2.42E−14
17	10.05	1-Octen-3-one	Organic oxygen compounds	Carbonyl compounds	3.63E−03
18	10.5	2-pentenol	Organic oxygen compounds	Alcohols and polyols	7.08E−05
19	12.25	2-Octenal	Organic oxygen compounds	Carbonyl compounds	2.90E−04
20	12.34	1-Tetradecanol	Lipids and lipid-like molecules	Fatty alcohols	7.52E−04
21	12.52	Acetic acid	Organic acids and derivatives	Carboxylic acids	2.94E−20
22	12.54	1-Octen-3-ol	Lipids and lipid-like molecules	Fatty alcohols	5.05E−06
23	12.95	2,5,5-Trimethyl-2-hexene	Hydrocarbons	Branched unsaturated hydrocarbons	1.47E−05
24	13.5	Propionic acid	Organic acids and derivatives	Amino acids, peptides, and analogues	1.31E−14
25	14	2,6-Nonadienal	Organic oxygen compounds	Carbonyl compounds	7.48E−11
26	14.02	Dimethyl sulfoxide	Organosulfur compounds	Sulfoxides	5.62Ev02
27	15.31	3,6-Nonadien-1-ol	Lipids and lipid-like molecules	Fatty alcohols	5.12E−11

Multivariate analysis according to storage: PLS-DA & OPLS-DA analysis

Volatile compounds of packaged oysters stored at different temperatures (5 °C and 15 °C) were analyzed, and the differences were visualized through PLS-DA and OPLS-DA plots (Figs. 2, 3). Partial least squares discriminant analysis is based on PLS regression analysis and is a suitable method to classify samples with diverse variables and patterns (Yue et al., 2023). All samples were statistically separated, indicating distinct quality parameters according to the PLS-DA score plot (Fig. 2) for the volatile profile data concerning packaged oysters stored at different temperatures (Total group: R₂X = 0.53, R₂Y = 0.123, Q₂ = 0.106/5 °C group: R₂X = 0.554, R₂Y = 0.141, Q₂ = 0.247/15 °C group: R₂X = 0.587, R₂Y = 0.193, Q₂ = 0.263). The permutation test (n = 200) confirmed the statistical acceptance of the PLS-DA model (p < 0.05; permutated R₂ < 0.4 and Q₂ < -0.1), signifying model accuracy and robustness (Liu et al., 2023a, 2023b). However, the overall predictive ability of the PLS-DA model for t1 and t2 was low. t1 and t2 represent the first and second latent variables in all the multivariate statistical models, respectively. These are linear combinations of the original variables (e.g., volatile compounds concentrations) that are computed to capture the maximum variance in the predictor matrix (X) while also achieving the best separation of the response variable (Y), which often represents different classes or conditions (Barker and Rayens 2003; Eriksson et al., 2001; Trygg and Wold, 2002). In PLS-DA, these latent variables are used to model the relationship between the predictor variables (X) and the response variables (Y), ensuring that the components extracted help in distinguishing between the predefined classes. Additionally, it is challenging to establish a reliable predictive model when the cumulative contribution to the variance of t1 and t2 was below 85% (Han et al., 2022). The PLS-DA score plot for packaged oysters stored at 5 °C (Fig. 2B) distinguished the initial storage (0–72 h) from the late storage (96–168 h) based on t1. However, clustering is shared between 0 and 96 h owing to the lack of significant changes in volatile compounds during storage at 5 °C. The PLS-DA score plot for packaged oysters stored at 15 °C exhibits clearer clustering compared to that at 5 °C. Specifically, the initial storage (0 h), 48 h, and 72 h were distinctly separated by t1. This indicates that changes in metabolites during the storage period form clusters; however, the differences are unclear, similar to that of previous findings (Kim et al., 2020).

Fig. 2.

Partial least squares discriminant analysis (PLS-DA) score plot of volatile compounds analyzed by gas chromatography-mass spectrometry (GC–MS). The quantification of the PLS-DA models was evaluated with R2X, R2Y, and Q2 and validated by cross-validation with a permutation test (n = 200). A PLS-DA plot using the volatile compounds of packaged oysters stored at 5 °C and 15 °C; B PLS-DA plot of volatile compound stored at 5 °C, PLS-DA plot of volatile compound stored at 15 °C

Fig. 3.

Orthogonal partial least squares discriminant analysis (OPLS-DA) score plot of volatile compounds analyzed by gas chromatography-mass spectrometry (GC–MS). Quantification of the OPLS-DA models was evaluated with R2X, R2Y, and Q2 and validated by cross-validation with a permutation test (n = 200). A OPLS-DA plot using the volatile compounds of packaged oysters stored at 5 °C and 15 °C; B OPLS-DA plot of volatile compound stored at 5 °C; C OPLS-DA plot of volatile compound stored at 15 °C

Orthogonal projections to latent structure-discriminant analysis was performed to further clarify the resulting patterns and systematically reduce noise by minimizing overfitting (Ma et al., 2022). The OPLS-DA score plot showed that all samples were statistically separated, indicating distinct quality parameters (Total group: R2X = 0.53, R2Y = 0.12, Q2 = 0.096/5 °C group: R2X = 0.487, R2Y = 0.103, Q2 = 0.232/15 °C group: R2X = 0.587, R2Y = 0.193, Q2 = 0.242 (Fig. 3). The permutation test (n = 200) confirmed the statistical acceptance of the OPLS-DA model (p < 0.05; permutated R2 < 0.4 and Q2 < -0.1). Similar to the PLS-DA model, the OPLS-DA plot including results for 5 °C and 15 °C does not exhibit clear clustering. However, distinct clustering was observed when the results were represented by storage temperature in the OPLS-DA plot. The OPLS-DA plot for packaged oysters stored at 5 °C showed clear changes based on storage days along t1, indicating clustering. In the OPLS-DA plot, the x-axis represents the predictive components explaining the maximum variation between groups, whereas the y-axis captures the orthogonal components unrelated to group differentiation (Saei et al., 2021). PLS-DA models the relationship between predictor variables (X) and response variables (Y) to discriminate between classes, extracting latent variables that mix predictive and non-predictive variations. In contrast, OPLS-DA separates the variation in X into predictive components related to Y and orthogonal components unrelated to Y, enhancing model clarity. OPLS-DA typically requires fewer predictive components, simplifying interpretation and visualization of class separation. This method provides clearer insights by focusing on variations directly correlated with the response variable, reducing noise (Trygg and Wold, 2002). Additionally, the scatters of each sample by storage day in this plot are clustered to the right and left, indicating separation based on the predicted component. Furthermore, the distances between scatters for packaged oysters stored at 15 °C were narrower than those in the 5 °C score plot, and clear clustering was evident, implying significant differences in metabolites among the samples and changes in quality. Significant and distinct metabolic changes owing to storage temperature were observed through GC–MS analysis of volatile compounds in packaged oysters stored at different temperatures. Moreover, statistically significant indicators can be selected based on changes in the storage temperature. The VIP values representing the weighted squares in the OPLS-DA model were used to select potential indicators of freshness during the storage of packaged oysters.

Identifying differential metabolites

The identification of potential compounds of interest was performed using OPLS-DA and specific volatile compounds with significant variation by VIP values represented the weighted squares of the OPLS-DA model. The VIP scores calculated the weighted sum of each feature's contribution and assessed the proportion of the y-variance explained in each dimension. The VIP values assess the collective impact of variables on the sample in the OPLS-DA model and are ranked based on their contributions (Moser et al., 2023). The potential freshness indicators for each storage temperature were identified using the OPLS-DA model and VIP screening (VIP > 1.0, p < 0.05) (Table 2). Ten volatile compounds with VIP scores exceeding 1.0 were selected from packaged oysters stored at 5 °C: methyl disulfide (1.7394), acetic acid (1.2413), ethanol (1.1573), propionic acid (1.1557), 1-propanol (1.1420), methyl cyanide (1.0751), dimethyl sulfide (1.0499), hexanal (1.0175), 2,6-nonadienal (1.0119), and 2-pentanal (1.0055). These 10 candidate compounds are graphically presented in Fig. 4A using normalized intensity values and storage times. The statistical significance of each compound was analyzed and displayed above the bars in the graph. Four compounds were selected based on their significant and linear changes depending on the storage time: propionic acid, 1-propanol, hexanal, and 2-pentanal. These four compounds were selected as the final freshness indicators based on statistically significant differences observed at each storage time point, as illustrated in Fig. 4A. Acetic acid and ethanol were excluded from the final selection because they did not show statistically significant differences (p > 0.05) until 168 h of storage. In contrast, propionic acid and 1-propanol exhibited statistically significant differences at each storage stage, warranting their selection as freshness indicators. Hexanal and 2-pentanal were also selected due to their high initial concentrations, which decreased significantly over time, thereby serving as reliable markers for oyster freshness. In brief, these volatile compounds exhibited statistical significance over time, showing an increase in intensity during the initial storage period when packaged oysters were stored at 5 °C, followed by a decrease or no measurement as the storage time progressed.

Table 2.

VIP score of identified compounds by the Multivariate Analysis

Storage temperature	Primary ID	Identified compound	VIP score
5℃	Var_10	Methyl disulfide	1.7394
	Var_22	Acetic acid	1.2413
	Var_5	Ethanol	1.1573
	Var_25	Propionic acid	1.1557
	Var_9	1-Propanol	1.1420
	Var_7	Methyl cyanide	1.0751
	Var_2	Dimethyl sulfide	1.0499
	Var_11	Hexanal	1.0175
	Var_26	2,6-Nonadienal	1.0119
	Var_13_	2-Pentenal	1.0055
15℃	Var_7	Methyl cyanide	1.2315
	Var_22	Acetic acid	1.1368
	Var_25	Propionic acid	1.1199
	Var_18	1-Octen-3-one	1.0928
	Var_17	1-Pentanol	1.0865
	Var_10	Methyl disulfide	1.0850
	Var_27	Dimethyl sulfoxide	1.0624
	Var_23	1-Octen-3-ol	1.0618
	Var_3	3,5-Octadiene/3-Octen-2-ol	1.0598
	Var_9	1-Propanol	1.0459
	Var_24	2,5,5-Trimethyl-2-hexane	1.0272
	Var_5	Ethanol	1.0251
	Var_11	Hexanal	1.0187
	Var_13	2-Pentenal	1.0117
	Var_20	2-Octenal	1.0055

Fig. 4.

The relative abundance of identified compounds with a high variable importance in projection (VIP) score (> 1.0) in packaged oysters (A, stored at 5 °C; B, stored at 15 °C). The vertical axis represents normalized intensity while the horizontal axis depicts the storage period. Different letters on the bars indicate significant differences (p < 0.05)

Kawabe et al. (2019) reported alterations in the volatile compound profiles of live Pacific oysters during storage. Their investigation revealed the presence of specific volatile compounds in the initial stages, which were undetectable during storage. Notably, 1-penten-3-one decreased to undetectable levels, whereas volatile carboxylic acids originating from anaerobic end-products and 1-propanol increased during air-exposed storage. Zhang et al. (2009) and Zhang et al. (2021) reported similar trends in the changes in volatile compounds in oysters during storage. Moreover, 15 volatile compounds with a VIP score exceeding 1.0 were identified from the packaged oysters stored at 15 °C (Table 2). These volatiles and their respective VIP scores are as follows: methyl cyanide (1.2315), acetic acid (1.1368), propionic acid (1.1199), 1-octen-3-one (1.0928), 1-pentanol (1.0865), methyl disulfide (1.0850), dimethyl sulfoxide (1.0624), 1-octen,3-ol (1.0618), 3,5-octadiene/3-octen-2-ol (1.0598), 1-propanol (1.0459), 2,5,5-trimethyl-2-hexane (1.0272), ethanol (1.0251), hexanal (1.0187), 2-pentenal (1.0117), 2-octenal (1.0055). The visualized graph utilizing the normalized intensity values and storage times of the 15 candidate volatiles is presented in Fig. 4B. Similar to the results from oysters stored at 5 °C, 6 volatiles demonstrated statistical significance for storage duration among the detected compounds in packaged oysters stored at 15 °C: acetic acid, propionic acid, ethanol, hexanal, 2-pentanal, and 2-octenal.

Changes in volatile compounds during the storage of packaged oysters are closely associated with metabolic processes, and compounds such as acetic acid, propionic acid, and ethanol are closely related to carbohydrate metabolism and fermentative hydrogen production (Antonopoulou et al., 2008; Horecker & Mehler, 1955; Maughan, 2009; Wang & Wan, 2008a, 2008b, 2008c). Packaged oysters stored at 5 °C exhibit a slower metabolic process owing to refrigeration effects compared to those stored at 15 °C, resulting in the predominance of more volatile compounds and carbohydrate metabolism-related compounds in 15 °C storage. The VIP scores derived from OPLS-DA were used to evaluate the impact of each compound on the overall model. The VIP scores exceeded 1.0, suggesting that these compounds may serve as key indicators of changes during packaged oyster storage. Finally, correlation analysis was performed between the volatile compounds, storage duration, and physicochemical indicators to select key indicators with high reliability.

Correlation with volatile compounds and physicochemical indicators

A correlation analysis was conducted between storage duration, physicochemical indicators, and volatiles to explore the freshness indicators associated with the storage of packaged oysters (Fig. 5). The correlation coefficient of storage duration signified a linear change in quality over time, indicating a higher correlation with the freshness of packaged oysters as the correlation coefficient increased (Schober et al., 2018). The results of the correlation analysis for packaged oysters stored at 5 °C are presented in Fig. 5A, and those for packaged oysters stored at 15 °C are shown in Fig. 5B. According to Ratner (2009), values between 0.3 and 0.7 (0.3 and − 0.7) imply a moderate positive (negative) linear relationship based on a fuzzy-firm linear rule, and values between 0.7 and 1.0 (− 0.7 and − 1.0) indicate a strong positive (negative) linear relationship based on a firm linear rule. Correlation analysis of packaged oysters stored at 5 °C revealed highly negative correlation coefficients for storage duration with pH (− 0.833) and turbidity (− 0.878) (Fig. 5A). According to Son et al. (2014), pH, glycogen, and soluble protein exhibit high correlation coefficients of 0.8 (or -0.8) or higher with each other. In this study, volatiles exhibiting a high positive correlation with storage duration included 1-propanol (0.755), ethanol (0.706), and propionic acid (0.702), while volatiles with a high negative correlation included 2-pentanal (− 0.709), and hexanal (− 0.704). Notably, the volatile with the highest VIP score in packaged oysters stored at 5 °C (methyl disulfide) did not show a significant correlation with the existing physicochemical indicators or storage duration in the correlation analysis. This phenomenon is attributed to the subtle differences in the VIP scores among individual volatiles. Although methyl disulfide exhibited the highest VIP score, its diminished correlation coefficient can be ascribed to the substantial deviation observed in the normalized intensity. The correlation analysis results indicated a high correlation between storage duration and pH (− 0.976), and turbidity (+ 0.924) for packaged oysters stored at 15 °C (Fig. 5B). Volatiles exhibiting a strong positive correlation with storage duration were ethanol (0.936) and propionic acid (0.882), whereas those with a strong negative correlation were 2-octenal (− 0.844), hexanal (− 0.770), and 2-pentanal (− 0.744). The rapid and substantial changes in metabolic compounds at 15 °C likely suggest the high correlation coefficients with storage duration.

Fig. 5.

Correlation analysis between identified volatile compounds, storage period, and physicochemical indicator (A stored at 5 °C; B stored at 15 °C). Red (+ 1.0) and green (− 1.0) color gradients indicate a positive or negative correlation coefficient. Bold letters mean statistical significance (p < 0.05)

This study performed a GC–MS analysis of volatile compounds in packaged oysters during storage, followed by multivariate statistical analysis to examine the differences between groups. Potential freshness indicators of the packaged oysters were selected based on the derived VIP scores. The selected candidate compounds were further analyzed for their correlation coefficients with storage duration and existing quality indicators (pH, turbidity, and soluble protein) to identify those that underwent significant changes during storage. Consequently, compounds with a negative correlation coefficient with storage duration were selected as potential indicators since packaged oyster freshness indicators should ideally be representative of their freshest state during the initial storage period. Here, we suggest hexanal and 2-pentanal as potential indicators of freshness of packaged oysters using multivariate and correlation analyses. These compounds showed statistically significant decreases during the storage period initially measured in the fresh state and were subsequently measured in very small amounts or not detected after the 3rd day of storage. The results of this study provide insights into tracking changes during the storage of packaged oysters to identify indicators representative of freshness. These results may be crucial for the future development of technologies related to oyster freshness measurements.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ATP

Adenosine triphosphate

GC–MS

Gas chromatography-mass-spectrometry

LC–MS

Liquid chromatography-mass spectrometry

NMR

Nuclear magnetic resonance

NTU

Nephelometric turbidity unit

OPLS-DA

Orthogonal projections to latent structure-discriminant analysis

PCA

Principal component analysis

PLS-DA

Partial least squares discriminant analysis

SPME

Solid phase microextraction

VBN

Volatile basic nitrogen

VIP

Variable importance in projection

VOC

Volatile organic compound

Funding

This study was supported by the Korea Institute of Marine Science & Technology Promotion (KIMST) funded by the Ministry of Oceans and Fisheries (20210695).

Declarations

Conflict of 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.