Characterization of off-flavor compounds in ready-to-heat roasted catfish after reheating by sensomics approach

Abstract

Gas chromatography-electronic nose (GC-E-Nose), gc-ion mobility spectrometry (GC-IMS), and gc-olfactory-mass spectrometry (GC-O-MS) techniques, along with odor recombination and omission experiments, were employed to characterize the off-flavors in commercially available ready-to-heat roasted catfish after reheating. A total of 19 (DR), 11 (HR), and 21 (YR) characteristic compounds were identified across the three samples. The 19 characteristic compounds from the DR group were reconstituted to develop recombination and omission models. The resulting model demonstrated the highest similarity (84.98 %) to the original sample while retaining some odor information from the other two sample groups. The 12 key off-flavor substances were identified through the omission model and were subsequently subjected to multivariate recombined. The contribution of these off-flavors was further validated via omission testing. The results indicated that the primary off-flavors in reheated ready-to-heat roasted catfish were attributed to hexanal, heptanal, (E)-2-hexenal, octanal, 3-methyl-1-butanol, 1-octen-3-ol, and (E,E)-2,4-heptadienal. This characterization of off-flavors in ready-to-heat roasted catfish post-reheating will contribute to the regulation of the product's flavor quality.

Highlights

• •Off-flavor substances of ready-to-heat roasted catfish after reheating were characterized.
• •Combined GC-E-nose, GC-IMS, and GC-O-MS techniques identified volatile compounds.
• •Seven off-flavor substances were identified from reheating ready-to-heat roasted catfish.
• •Sulfide, allyl methyl, linalool and d-limonene may be off-flavor masking agents.

1. Introduction

Channel catfish is a high-quality freshwater fish that is well-suited for processing. Its meat is flavorful, rich in fat and various nutrients, possesses a unique taste, and is devoid of intermuscular spines. Currently, many channel catfish are processed into frozen catfish fillets, frozen catfish balls, and ready-to-heat roasted catfish (Zhong et al., 2016). Ready-to-heat roasted catfish is a semi-finished product created through a pre-treatment process that includes open back curing, roasting, and quick-freezing, allowing for consumption with simple reheating. This convenience, along with its stable taste and rich nutritional profile, has made it widely popular in the consumer market. However, during the reheating process, roasted catfish often develops an undesirable sensory characteristic known as “off-flavor”, which encompasses fishy notes and “warmed-over flavor” (WOF). This off-flavor significantly detracts from the overall olfactory enjoyment of the product and has impeded the advancement of the prepared food industry (Tims & Watts, 1958).

WOF is predominantly found in cooked-refrigerated-reheated meat products, where it was initially described as an oxidizing odor. This characterization has since been refined with descriptors such as unpleasant wet cardboard, linseed oil, paint, rancidity, unfreshness, acidity, hard-boiled egg, and fat odors (Byrne, O'Sullivan, Dijksterhuis, Bredie, & Martens, 2001; Lage et al., 2012). These odors primarily arise from lipid oxidation products, including aldehydes, alcohols, and other volatile compounds. Pathways associated with WOF production include lipid autoxidation and singlet oxygen oxidation (Kim, Li, Lim, Kang, & Park, 2016; Ramalingam, Song, & Hwang, 2019). Additionally, the oxidation of polyunsaturated fatty acids in phospholipids and phospholipid mixtures has been reported to contribute to WOF formation in cooked meat (Lepper-Blilie et al., 2014). Furthermore, the increase in lipid oxidation products, coupled with the loss of compounds responsible for meat flavor (e.g., furanone and sulfur-containing compounds) due to the Maillard reaction and protein degradation, can also enhance WOF (Kerler and Grosch, 1996; Zhang et al., 2021). During refrigeration, low temperatures inhibit fatty acid degradation, resulting in a decrease in fat degradation products. Conversely, the reheating process promotes fatty acid oxidation, leading to an increase in fatty acid oxidation products and, consequently, WOF (Zhang et al., 2022b).

Numerous researchers have investigated the formation of off-flavors in meat products. In precooked pork, the most significant contributors to fishy and WOF include 1-octen-3-ol, (Z)-2-octenal, and (E,E)-2,4-decadienal (Zang et al., 2020). In precooked beef, off-flavors are primarily attributed to the formation of hexanal, (E,E)-2,4-decadienal, pentanal, decanal, octanal, heptanal, (E)-2-octenal, (E)-2-nonenal, (Z)-2-octenal, and 1-octen-3-ol (Konopka and Grosch, 1991; Liu et al., 2024). These compounds are mainly products of lipid oxidation and are widely acknowledged as the principal sources of fishy and WOF in meat products. The development of off-flavor in prepared meat products has been extensively studied; however, the characterization of off-flavor compounds in aquatic products remains limited, with existing reports primarily focusing on surimi. Most of the identified WOF substances in surimi gels are aldehydes, including (E,E)-2,4-decadienal, heptanal, octanal, nonanal, decanal, (E)-2-nonenal, (E)-2-octenal, (E)-2-decenal, (E,E)-2,4-heptadienal, and 2,3-pentanedione (An et al., 2022). Currently, off-flavor substances have not been reported in ready-to-heat aquatic products.

Therefore, this study employed three brands of commercially available ready-to-heat roasted catfish to investigate the off-flavor substances present after reheating, utilizing GC-E-Nose, GC-IMS, and GC-O-MS in conjunction with sensory omics. A multivariate recombination model was subsequently applied for further screening and validation. This research provides a theoretical foundation for flavor regulation in fish processing, thereby promoting the high-quality development of the ready-to-heat roasted catfish industry.

2. Materials and methods

2.1. Sample preparation

Three top-selling roasted catfish products were purchased from online retailers, with 10 units of each brand (each weighing 450 g ± 10 g), specifically Youyuyao (Y), Deyan (D), and Huihaiying (H). All products were made from channel catfish, with raw materials tested for quality at the factory prior to processing. The catfish were prepared as ready-to-heat roasted products by opening the back, curing at 6–10 °C for 3–5 h, roasting at 180–190 °C for 25–30 min, and subsequently freezing at −40 °C ± 5 °C. The ready-to-heat roasted catfish products, frozen for six months, were transported to the laboratory via cold chain logistics and stored in a refrigerator at −80 °C. During the measurement process, the samples were thawed in a refrigerator at 4 °C until the center temperature reached 0 °C. The samples were then reheated according to the optimal reheating method specified by the product. The thawed samples were placed in aluminum foil trays, to which 500 mL of purified water was added and boiled for 10 min. After cooling, the samples were measured and analyzed. At this stage, the reheated samples from the three brands were designated as YR, DR, and HR, respectively.

2.2. Chemicals

Standards are used to identify and quantify volatile compounds and are subjected to reconstitution and omission experiments. 3-Methylbutanal (purity, 98 %), pentanal (95 %), (E)-2-hexenal (97 %), (E,E)-2,4-heptadienal (95 %), decanal (97 %), benzaldehyde (99.5 %), 2-pentanone (98 %), ethyl acetate (99 %), ethanol (99 %), 2-methyl-butanoic acid, methyl ester (99 %), 1-butanol (99 %), 1-penten-3-ol (95 %), d-limonene (95 %), hexanoic acid, ethyl ester (99 %), 1-pentanol (99 %), 2-methyl-pyrazine (95 %), 2,3-octanedione (95 %), 1-hexanol (99 %), hydroxy-acetic acid, methyl ester (99 %), 2-ethyl-5-methyl-pyrazine (95 %), 1-heptanol (99.5 %), 2-ethyl-1-hexanol (97 %), 1-nonanol (99 %), 3-cyclohexene-1-methanol, .alpha.,.alpha.,4-trimethyl-, (S)- (95 %), benzyl alcohol (99 %), 2,6-dimethyl-pyrazine (97 %), 2,3-dimethyl-pyrazine (97 %), (E)-2-hexen-1-ol (99 %), 1-(2-furanyl)-ethanone (95 %), sulfide, allyl methyl (99 %), 3-methyl-1-butanol (99 %), 1-octen-3-ol (95 %), 3-cyclohexen-1-ol, 4-methyl-1-(1-methylethyl)-, (R)- (95 %), carbon disulfide (98 %), linalool (99 %) and the internal standard, 2-octanol (99 %), was purchased from Aladdin (Shanghai, China). The n-alkanes external standard (C7-C30), hexanal (95 %), heptanal (97 %), octanal (99 %), and nonanal (99.5 %) was purchased from Sigma-Aldrich Chemical Co. (St. Louis, MO, USA).

2.3. Determination of GC-E-Nose

A 2.0 g chopped sample and 2 mL of saturated NaCl solution were weighed into a headspace vial, which was subsequently sealed using a PTFE spacer. The prepared sample was then placed in a GC-E-Nose Heracles NEO autosampler unit. The sample heating and oscillation temperature were set to 65 °C, with a duration of 30 min. An injection volume of 5000 μL was utilized at an injection rate of 250 μL/s, with the inlet port temperature maintained at 200 °C and an injection duration of 25 s. Hydrogen served as the carrier gas. Following this, a purge and trap system was employed to adsorb, concentrate, and aggregate the volatiles present in the headspace. The initial and final trap temperatures were set at 30 °C and 240 °C, respectively, with a trap duration of 30 s. The analytes were separated and transferred simultaneously to two columns. The column oven contained two parallel capillary columns (MXT-5 and MXT-1701) of differing polarities. The temperature program for the oven commenced at 40 °C for 30 s, followed by an increase of 1.0 °C/s to 80 °C, and then 1.5 °C/s to 260 °C for 10 s. Volatile compounds were qualitatively analyzed using the AroChemBase database (Alpha M.O.S., Toulouse, France) and n-alkanes (C6-C16) within the instrument software. The relative odor activity values (ROAV) of the volatiles in the samples were calculated according to the method described by Xu et al. (2023).

2.4. Determination by GC-IMS

The volatile profiles of reheated roasted catfish were characterized using GC-IMS. The sample preparation and detection methods were adapted from An et al. (2022). A 2.0 g chopped sample was placed into a 20 mL headspace bottle and maintained at a temperature of 80 °C for 15 min. GC was conducted with a column (MXT-5, ID: 0.53 mm, FT: 1 μm) to separate the volatiles. The chromatographic program was as follows: 2 mL/min for 2 min, 10 mL/min for 8 min, 100 mL/min for 10 min, and 150 mL/min for 10 min. Nitrogen served as the drift gas for IMS at a flow rate of 150 mL/min, with the temperature set at 45 °C. Two-dimensional topographic plots and fingerprints of the volatiles from the samples were constructed using the Reporter and Gallery plug-ins. Qualitative analysis of the volatiles was performed using the National Institute of Standards and Technology (NIST) and IMS databases within the instrument software, in conjunction with n-ketones (C4-C9). The ROAV of the volatiles in the samples was calculated following the method described by Xu et al. (2023).

2.5. Determination of GC–MS

2.5.1. Determination of volatile compounds

A solid-phase microextraction (SPME) - GC–MS system (7890 A-5975C, Agilent, USA) was employed to collect, separate, and detect sample volatiles. A 2.0 g chopped sample was placed in a 20 mL headspace vial. Subsequently, 50 μL of 2-octanol (0.86 μg/mL) was rapidly injected, and the vial was sealed with a PTFE silicone stopper. The sealed vials were equilibrated in a water bath at 45 °C for 15 min. Extraction was conducted continuously for 40 min at 60 °C using SPME fibers (50/30 μm DVB/CAR/PDMS, Supelco, USA). Following extraction, desorption was performed at 250 °C for 5 min.

The volatile compounds were separated using a DB-WAX capillary column (60 m × 250 μm × 0.25 μm, Agilent, USA). The initial column temperature for gas chromatography was set at 40 °C, which was subsequently increased to 100 °C at a rate of 2 °C/min. The temperature was further raised to 180 °C at a rate of 5 °C/min, followed by an increase to 240 °C at a rate of 8 °C/min, with a holding time of 5 min. Helium served as the carrier gas at a flow rate of 1.0 mL/min, which the forward sample port temperature was maintained at 250 °C. Mass spectra were acquired using electron collision scanning within a mass scan range of 30–400 m/z. The split ratio of the sniffer port ODP3 to the mass spectrometry detector was established at a 1:1 volume ratio at the capillary column outlet. Odor descriptions were recorded concurrently as the samples were analyzed on the column (Yang et al., 2021a). Three replicates were employed for the GC-O-MS analysis in this study.

2.5.2. Identification of volatile compounds

The volatile compounds present in reheated catfish were identified utilizing the NIST 17 mass spectrometry database. Retention indices (RIs) were calculated for n-alkanes ranging from C7 to C30, along with corresponding odor descriptions and standards. Experienced panelists, who were familiar with odor descriptions and artificial odorant solutions, assessed the odor profiles of the volatile compounds. Three seasoned panelists conducted olfactory evaluations and were instructed to document the characteristics of the odors (Refer to the sensory method described in Section 2.6).

2.5.3. Quantification of volatile compounds

Volatile compounds in reheated catfish were semiquantitated using 2-octanol as an internal standard. The concentration of these volatile compounds was determined based on the ratio of the peak area to the concentration of the internal standard (2-octanol).

Prior to quantitative analysis, the Deyan brand of grilled fish underwent further processing to create an artificial odorless matrix. Among various brands, Deyan's roasted fish was the most popular. The DR samples were prepared according to the methodology outlined in section 2.1. The samples were quickly minced and subsequently rinsed with organic solvents to eliminate all volatile compounds until no odor was detectable by SPME-GC-O-MS (Chen et al., 2023; Liu et al., 2019). This rinsing process involved the addition of diethyl ether and pentane in a 2:1 (v/v) ratio to the minced roasted fish. The mixture was shaken for 8 h, after which the organic solvent was removed through five cycles of rinsing. Finally, the samples were treated with liquid nitrogen and frozen at −60 °C for 48 h using an LGJ-10C freeze dryer (Beijing Sihuan Scientific Instrument Factory Co., Ltd., Beijing, China).

Based on both identification and semi-quantitation, volatile compounds with odor-activity values (OAVs) greater than 1 were quantified after reheating catfish. The standards were diluted with methanol to create seven different concentrations (Yang et al., 2021b). 2-Octanol served as a calibrated internal standard for quantifying external standards. In the selected ion monitoring (SIM) mode, the concentrations of the flavor standards were determined using the characteristic ionic fragments of the standard compounds. The SPME method was employed to extract volatile compounds by introducing mixed standard and internal standard solutions into an artificial odorless matrix. The obtained extracts were then injected into the GC–MS instrument, and the area ratio of each target compound to its corresponding internal standard was calculated. Peak area ratios and mass ratios were utilized to establish standard curves. Detailed quantitative information and standard curves for each standard are presented in Table 1, with R² values exceeding 0.99, indicating a strong linear correlation. The standard curve was subsequently used to quantify the volatile components of roasted catfish.

Table 1.

Internal Standard (IS), Scanned ions, calibration equations, and odor descriptions in the determination of volatile compounds in selected-ion-monitoring (SIM) Mode.

Compound	Ions (m/z)a	Calibration equationsb	R2	Linear range (μg/mL)	Odor descriptionc
Hexanal	39, 44, 56	y = 0.0094× + 2.1503	0.9920	0.04–3456	Almond, malt, pungent
Heptanal	41, 55, 70	y = 0.0361× + 0.821	0.9939	0.02–2216	Creamy，fatty, green, metallic
Octanal	41, 57, 84	y = 0.0547× + 0.4329	0.9994	0.02–2252	Citrus, fatty, warmed-over
Pentanal	44, 71, 86	y = 0.0039× + 1.2747	0.9999	0.22–3376	Almond, malt, pungent
Decanal	57, 70, 82	y = 0.0018× + 0.1466	0.9990	0.01–956	Fatty
Nonanal	41, 57, 70	y = 0.0272× + 0.334	1	0.01–1260	Citrus, fatty, warmed-over
Benzaldehyde	51, 77, 106	y = 0.0014× + 0.2012	0.9999	0.01–1120	Green, rosy
3-Methylbutanal	58, 71, 86	y = 0.0002× + 0.0812	0.9936	0.02–1712	Malty，fermented, fruity, nutty
(E)-2-Hexenal	41, 55, 69	y = 0.0133× + 0.5531	0.9988	0.02–2196	Nutty，pungent, marzipan
(E,E)-2,4-Heptadienal	53, 81, 110	y = 0.0004× + 0.1249	0.9956	0.7–2324	Fatty, fishy
Ethanol	31, 41, 45	y = 0.0001× + 0.3254	0.9945	1.11–3480	Pungent odor
1-Pentanol	42, 55, 70	y = 0.0031× + 0.5627	1	1.22–3800	Musty, nutty
1-Hexanol	43, 56, 69	y = 0.0084× + 0.2637	0.9999	1.42–4440	Mossy，oil, fruity
1-Octen-3-ol	43, 57, 72	y = 0.0046× - 0.1739	1	0.72–2256	Mushroom-like
1-Heptanol	41, 56, 70	y = 0.0009× - 0.0755	0.9999	0.69–2168	Fatty
2-Ethyl-1-hexanol	41, 57, 83	y = 0.003× + 0.4073	1	0.73–2272	Floral
Linalool	43, 71, 93	y = 0.1519× - 0.0087	0.9995	0.64–2000	Flower, lavender
3-Cyclohexen-1-ol, 4-methyl-1-(1-methylethyl)-, (R)-	71, 93, 111	y = 0.1043× - 4.8362	1	0.39–1208	–
3-Cyclohexene-1-methanol, .alpha.,.alpha.,4-trimethyl-, (S)-	59, 93, 121	y = 0.1036× - 0.9422	0.9995	0.01–220	Oil, anise, mint
1-Penten-3-ol	41, 57, 71	y = 0.0014× + 0.1346	1	1.28–4000	Pungent, butter， grassy, fishy
3-Methyl-1-butanol	39, 55, 70	y = 0.0012× + 0.0682	1	1.21–3772	Musty, nutty
(E)-2-Hexen-1-ol	41, 57, 82	y = 0.0036× + 0.1592	1	1.14–3576	Floral
1-Nonanol	41, 56, 70	y = 0.2968× - 11.918	1	0.25–784	Floral, rosy
Benzyl Alcohol	51, 79, 108	y = 7E-06× + 0.0136	0.9941	0–9250	Green, floral
1-Butanol	31, 41, 56	y = 0.0012× - 0.0513	1	1.20–3760	Floral
2-Pentanone	43, 58, 86	y = 0.0278× + 0.0554	0.9958	0.03–2000	–
2,3-Octanedione	43, 71, 99	y = 0.0014× + 1.4421	0.9994	0.06–4760	Warmed-over
2-Methyl-butanoic acid, methyl ester	41, 57, 88	y = 0.0171× + 4.1679	0.9909	0.04–3688	Floral
Ethyl Acetate	43, 61, 70	y = 0.0108× + 0.0965	0.9990	0.16–40,000	–
Hydroxy-acetic acid, methyl ester	31, 59, 90	y = 8E-05× - 0.0548	1	0.33–25,800	Floral
Hexanoic acid, ethyl ester	43, 60, 99	y = 0.079× + 0.8364	0.9997	0.01–1272	Sweet, fruity
Methyl-pyrazine	67, 53, 94	y = 0.0005× + 0.0837	1	0.06–4440	Nutty
2,6-Dimethyl-pyrazine	39, 42, 108	y = 0.0068× - 0.0769	1	0.03–2020	Potato-like
2,3-Dimethyl-pyrazine	40, 67, 108	y = 0.0002× + 0.0347	1	0.06–4892	Potato-like, nutty
2-Ethyl-5-methyl-pyrazine	56, 94, 121	y = 0.0014× + 0.3643	0.9999	0.05–3648	Nutty
Sulfide, allyl methyl	73, 88, 61	y = 0.0085× + 3.282	0.9997	0.02–1220	Garlicky
1-(2-furanyl)-Ethanone	39, 95, 110	y = 1.0338× + 0.0375	0.9975	0.26–20,000	–
d-Limonene	68, 79, 93	y = 0.0064× + 1.9621	0.9956	0.12–1844	Citrussy, flowery
Carbon disulfide	32, 44, 76	y = 0.0007× + 8.6283	0.9915	0.51–8000	–

^aMonitored ions used for quantitation. ^b Variables: x is the peak area relative to that of the internal standard (2-octanol), and y is the concentration (μg/kg) in the sample relative to that of the internal standard (2-octanol). ^c Sniffed odors determined by GC-O analysis.

2.6. Recombination and omission experiments

Recombination and omission experiments for volatile compounds were conducted using sensory evaluation methods (Liu et al., 2019). The sensory evaluation took place in an odorless room maintained at 25 ± 1 °C. Nine healthy subjects were selected from a pool of 25 individuals at the Institute of Agricultural Products Processing and Nuclear Agriculture Technology, Chinese Academy of Agricultural Sciences. All participants consented to take part in the sensory experiment and permitted the use of their information. Each sensory panelist had prior experience in food sensory evaluation and was trained in various flavor criteria to accurately describe odor types and intensities. The training involved the use of aqueous solutions of reference compounds, including (E,E)-2,4-heptadienal, which exhibits a fishy and fatty flavor; hexanal, which has a grassy flavor; linalool characterized by a floral flavor; and 1-octen-3-ol, known for its mushroom flavor.

Recombination models were developed in odorless matrices based on the quantitative concentrations of characteristic compounds found in various samples (Table 2). The original models DR, HR, and YR denote the samples after reheating the original baked fish. Recombination models A, B, and C consist of an odorless matrix, ultrapure water, and characteristic volatiles with an OAV ≥ 1 from DR, HR, and YR, respectively. A nine-member sensory panel conducted odor evaluations at room temperature (25 ± 1 °C). The intensity of earthy, fishy, grassy, warmed-over, metallic, fatty, and nutty flavors was assessed using the following scale: 0, no flavor; 1, very weak; 2, weak; 3, medium; 4, strong; 5, very strong. Additionally, the similarity of each recombinant model to the original model was recorded as a percentage, ranging 0 to 100 %.

Table 2.

Volatile compound concentrations, identification methods, and odor activity values (OAVs) during catfish processing.

Compounds	ID	LRI	Concentration (μg/kg)			Thresholds (μg/kg)	OAV
			HR	YR	DR		HR	YR	DR
Carbon disulfide	MS, LRI, O, S	–	34,210.04 ± 9977.04	25,370.59 ± 6565.18	–	–
2-Pentanone	MS, LRI, O, S	792.66	94.52 ± 9.09	181.25 ± 36.80	348.64 ± 216.02	–
Ethyl Acetate	MS, LRI, O, S	927.88	129.47 ± 35.98	150.28 ± 24.41	648.26 ± 116.02	3300.00	0.04	0.05	0.20
3-Methylbutanal	MS, LRI, O, S	961.67	–	23,881.39 ± 1235.15	3890.52 ± 666.23	1.10	0.00	21,710.36	3536.84
Ethanol	MS, LRI, O, S	990.19	326,822.52 ± 66,758.10	8668.83 ± 1540.01	812,785.87 ± 7999.94	52,000.00	6.29	0.17	15.63
Pentanal	MS, LRI, O, S	1027.90	–	236.94 ± 0.00	–	20.00	0.00	11.85	0.00
2-Methyl-butanoic acid, methyl ester	MS, LRI, O, S	1076.29	–	99.29 ± 0.00	–	0.25	0.00	397.15	0.00
Hexanal	MS, LRI, O, S	1207.37	875.28 ± 166.09	4258.25 ± 1152.08	8673.53 ± 1272.54	4.50	194.51	946.28	1927.45
1-Butanol	MS, LRI, O, S	1318.10	246.95 ± 15.39	421.38 ± 133.32	292.27 ± 89.20	500.00	0.49	0.84	0.58
1-Penten-3-ol	MS, LRI, O, S	1338.13	449.97 ± 83.56	967.05 ± 102.88	907.58 ± 164.39	400.00	1.12	2.42	2.27
Heptanal	MS, LRI, O, S	1368.86	6.19 ± 0.00	51.42 ± 16.12	60.25 ± 19.61	2.80	2.21	18.36	21.52
d-Limonene	MS, LRI, O, S	1386.98	–	208.21 ± 50.25	139.34 ± 27.91	10.00	0.00	20.82	13.93
(E)-2-Hexenal	MS, LRI, O, S	1411.87	10.65 ± 1.31	0.31 ± 0.00	1252.32 ± 357.91	19.20	0.55	0.02	65.22
Hexanoic acid, ethyl ester	MS, LRI, O, S	1444.61	–	–	34.62 ± 4.39	5.00	0.00	0.00	6.92
1-Pentanol	MS, LRI, O, S	1475.32	429.27 ± 121.19	329.90 ± 23.35	1308.02 ± 249.35	42.00	10.22	7.85	31.14
Methyl-pyrazine	MS, LRI, O, S	1495.54	293.01 ± 0.00	7890.15 ± 989.01	61.01 ± 0.00	1900.00	0.15	4.15	0.03
Octanal	MS, LRI, O, S	1523.32	–	17.06 ± 3.49	14.15 ± 2.95	0.50	0.00	34.12	28.30
2,3-Octanedione	MS, LRI, O, S	1569.51	–	1802.85 ± 6.49	3475.68 ± 828.03	2.52	0.00	715.42	1379.24
1-Hexanol	MS, LRI, O, S	1613.03	2395.64 ± 629.82	896.60 ± 136.26	304.03 ± 51.82	5.60	427.79	160.11	54.29
Hydroxy-acetic acid, methyl ester	MS, LRI, O, S	1638.16	14,131.38 ± 1662.63	21,360.74 ± 3261.12	11,446.04 ± 348.96	–
2-Ethyl-5-methyl-pyrazine	MS, LRI, O, S	1651.74	–	782.49 ± 189.85	15.49 ± 0.00	320.00	0.00	2.45	0.05
1-Heptanol	MS, LRI, O, S	1707.19	1229.95 ± 109.57	1470.60 ± 167.97	2052.91 ± 43.63	1.10	1118.14	1336.91	1866.28
(E,E)-2,4-Heptadienal	MS, LRI, O, S	1739.48	–	44.39 ± 0.00	526.40 ± 106.32	15.40	0.00	2.88	34.18
2-Ethyl-1-hexanol	MS, LRI, O, S	1745.12	23.93 ± 5.09	179.85 ± 63.12	10.86 ± 3.71	270,000.00	0.00	0.00	0.00
Decanal	MS, LRI, O, S	1750.84	5.22 ± 1.47	59.05 ± 11.17	–	0.10	52.20	590.52	0.00
Benzaldehyde	MS, LRI, O, S	1780.89	142.63 ± 24.09	664.63 ± 77.17	–	41.70	3.42	15.94	0.00
Linalool	MS, LRI, O, S	1822.48	0.89 ± 0.18	0.84 ± 0.20	20.13 ± 2.15	2.20	0.40	0.38	9.15
1-Nonanol	MS, LRI, O, S	1982.12	40.99 ± 0.21	41.40 ± 0.44	40.57 ± 0.03	41.00	1.00	1.01	0.99
3-Cyclohexene-1-methanol, .alpha.,.alpha.,4-trimethyl-, (S)-	MS, LRI, O, S	2112.02	22.49 ± 2.65	54.14 ± 15.79	14.83 ± 3.84	330.00	0.07	0.16	0.04
Benzyl Alcohol	MS, LRI, O, S	–	12,036.21 ± 1267.92	–	4066.18 ± 364.08	120.00	100.30	0.00	33.88
2,6-Dimethyl-pyrazine	MS, LRI, O, S	1572.09	–	643.78 ± 80.00	–	250.00	0.00	2.58	0.00
2,3-Dimethyl-pyrazine	MS, LRI, O, S	1603.11	–	5502.42 ± 593.52	–	2500.00	0.00	2.20	0.00
Nonanal	MS, LRI, O, S	1650.20	–	45.41 ± 13.21	–	1.1	0.00	41.28	0.00
(E)-2-Hexen-1-ol	MS, LRI, O, S	1665.62	–	26.74 ± 2.49	–	100.00	0.00	0.27	0.00
1-(2-furanyl)-Ethanone	MS, LRI, O, S	1761.62	–	0.09 ± 0.02	–	6	0.00	0.02	0.00
Sulfide, allyl methyl	MS, LRI, O, S	1017.85	–	–	421.68 ± 99.03	2.00	0.00	0.00	210.84
3-Methyl-1-butanol	MS, LRI, O, S	1404.93	–	–	3067.73 ± 217.69	1000.00	0.00	0.00	3.07
1-Octen-3-ol	MS, LRI, O, S	1700.54	–	–	62.25 ± 0.31	1.00	0.00	0.00	62.25
3-Cyclohexen-1-ol, 4-methyl-1-(1-methylethyl)-, (R)-	MS, LRI, O, S	1882.30	–	–	66.18 ± 2.01	340.00	0.00	0.00	0.19

Abbreviations: MS, identified by NIST mass spectral database; LRI, agreed with the published linear retention index; O, sniff odors from ODP; S, standard. Blank, no detect. The thresholds of volatile compounds in water was obtained from a book “compilations of flavor threshold values in water and other media (second enlarged and revised edition)”. Data: mean ± standard deviation.

The recombination model A was selected as the representative model due to its comprehensive information on odor. To assess the contribution of specific compounds to off-flavor, odor omission experiments were conducted. Compounds were categorized based on their odor descriptions and chemical properties, and subsequently omitted from recombinant model A. In total, 19 omission models were developed. Each model was evaluated against recombinant model A using a triangulation test, following the methodology outlined in ISO 4120:2021 (ISO 4120, 2021). All test (19 omission models) samples were arranged in sequence, and panelists were instructed to sniff the samples. Additionally, panelists were asked to describe their perceived differences and to indicate whether the omitted compounds had a positive or negative impact on off-flavor.

2.7. Validation experiments on the importance of key volatile compounds for off-flavor

To confirm the significance of the key compounds identified through the recombination deletion experiment in relation to off-flavor, a validation test was conducted. The key off-flavor substances selected in Section 2.6 were subjected to multivariate reconstitution based on quantitative results. These reconstituted substances were then incorporated into a flavorless matrix to simulate the incubation temperature and duration outlined in Section 2.5.1 for sensory evaluation of the model. The selected key off-flavor compounds were reconstituted to recreate the original model. Initially, one compound was omitted, resulting in what is referred to as Reconstituted Model 1. By analyzing the similarity with the original model, the significance of the omitted compound to the off-flavor was assessed; a high similarity (greater than 90 %) indicated that the compound had a minimal contribution to the off-flavor and could be excluded from the construction of Reconstituted Model 2. Conversely, a low similarity (less than 90 %) suggested that the compound played a significant role in the off-flavor, necessitating its inclusion in Reconstituted Model 2. A total of 12 multivariate reconstituted models were constructed (Table 3), and through similarity analysis, the critical off-flavor substances in the reheated grilled catfish were identified.

Table 3.

Construction of a multivariate recombination model.

Model	Recombinant substances
Original	add 1–12
Recombination model 1	add 1–11, missing 12
Recombination model 2	add 1–10, missing 11 and 12
Recombination model 3	add 1–9 and 11, missing 10 and 12
Recombination model 4	add 1–8 and 11, missing 9–10 and 12
Recombination model 5	add 1–7, 9 and 11, missing 8, 10 and 12
Recombination model 6	add 1–6, 9 and 11, missing 7, 8, 10 and 12
Recombination model 7	add 1–5, 7, 9 and 11, missing 6, 8, 10 and 12
Recombination model 8	add 1–4, 7, 9 and 11, missing 5, 6, 8, 10 and 12
Recombination model 9	add 1–3, 5, 7, 9 and 11, missing 4, 6, 8, 10 and 12
Recombination model 10	add 1, 2, 4, 5, 7, 9 and 11, missing 3, 6, 8, 10 and 12
Recombination model 11	add 1, 3–5, 7, 9 and 11, missing 2, 6, 8, 10 and 12
Recombination model 12	add 2–5, 7, 9 and 11, missing 1, 6, 8, 10 and 12

Number 1: 3-methylbutanal; 2: hexanal; 3: heptanal; 4: 3-methyl-1-butanol; 5: (E)-2-hexenal; 6: hexanoic acid, ethyl ester; 7: octanal; 8: 2,3-octanedione; 9: 1-octen-3-ol; 10: 1-heptanol; 11: (E,E)-2,4-heptadienal; 12: benzyl alcohol.

2.8. Statistical analysis

A one-way analysis of variance (ANOVA) was conducted to identify significant differences, utilizing data processing system (DPS) software. The data were visualized using Origin 2021 software (Microcal Software, Inc., Northampton, MA, USA).

3. Results and discussion

3.1. GC-E-nose assessment of the odor profile of ready-to-heat roasted catfish after reheating

Conventional electronic noses primarily utilize various metal oxide gas sensors to fingerprint responses to volatile compounds (Di-Rosa, Leone, Cheli and Chilfalo, 2017). However, this approach is susceptible to sensor drift and contamination, leading to highly ambiguous signal information (Lutfi, Coradeschi, Mani, Shankar, & Rayappan, 2015). The GC-E-Nose integrates two chromatographic columns of differing polarity arranged in parallel, simultaneously generating two chromatograms to monitor fingerprint information in real time, thereby providing more comprehensive data on volatile components (Chen et al., 2022). A total of 23 volatile substances were identified in different grilled fish after reheating, including 7 aldehydes, 6 ketones, 5 alcohols, 2 olefins, 1 ester, 1 pyrazine and 1 amine. Except for 2-propanol and vinylbenzene, 21 volatiles were detected in all three reheated roasted fish, although variations in content were observed (Fig. 1A). Additionally, DR exhibited partial odor information that was also present in the other two roasted fish products.

Fig. 1.

Gas chromatography-electronic nose and gas chromatography-ion mobility spectrometry determination of odor characteristics of ready-to-heat roasted catfish after reheating. (A) the content of volatile compounds after reheating the three products. (B) the volatile compounds with ROAV ≥1 and their scores. (C) the compositional profile of volatiles in samples. (D) the fingerprints of volatile constituents in samples. (E) the volatile compounds with ROAV ≥1 and their scores.

In addition to the concentration of volatile compounds, the odor threshold significantly influences the odor contribution of a product. Consequently, we calculated the ROAV of the measured volatile compounds, considering those with ROAV ≥1 as characteristic volatiles of the samples (Fig. 1B). For the DR samples, the characteristic flavor compounds identified were propanal (ROAV = 100) and (E)-2-butenal (ROAV = 1.88). In the HR samples, propanal (ROAV = 100) was also identified as a characteristic flavor compound. The YR samples exhibited characteristic flavor compounds including propanal (ROAV = 2.44), (E)-2-butenal (ROAV = 100), and 1-penten-3-one (ROAV = 27.86). Propanal has been reported to be associated with A. sobria (Tan et al., 2024). (E)-2-Butenal, which is derived from the oxidative degradation of triglycerides, possesses a pungent odor (Hu et al., 2023). Additionally, 1-penten-3-one is an unsaturated ketone that produces an unpleasant fishy and sour odor (Hartvigsen, Lund, Hansen, & Holmer, 2000; Venkateshwarlu, Let, Meyer, & Jacobsen, 2004). The production of both (E)-2-butenal and 1-penten-3-one is linked to the oxidative degradation of lipids, resulting in the formation of off-flavor substances. Therefore, these characteristic volatiles may contribute to the development of off-flavor substances upon reheating roasted fish. In summary, the GC-E-Nose technique can rapidly and effectively capture the odor profiles of various ready-to-heat roasted catfish after reheating, serving as a valuable complement to sensory evaluation.

3.2. GC-IMS analysis of the odor profile of ready-to-heat roasted catfish after reheating

To further identify the key volatile compounds after reheating various roasted catfish samples, we employed GC-IMS. Each point to the right of the reactive ion peak corresponds to a volatile compound, with the intensity of the red and blue colors indicating the content of each compound; higher intensity signifies greater content (Fig. 1C). To visualize the differences in odor composition and content among the three sample groups, we created a two-dimensional differential topographic map using DR as a reference. In this map, regions are colored white when the sample content is similar to that of the reference samples. A red area signifies that the volatile compound content exceeds that of DR, while a blue area indicates lower content compared to DR. The figure reveals a significant difference in the volatile compound content among the three sample groups. Consequently, a fingerprint was constructed using the Gallery Plot plug-in to elucidate the distinct compounds present in the three groups. Each row of the graph represents the signal peak of an individual sample, and each column corresponds to the signal peak of the same volatile compound, with darker red colors indicating higher substance content (Fig. 1D). It is noteworthy that when volatile compound molecules share a proton or electron, they may form dimers or polymers (Liu, Guo, Gao, Bao and Lin, 2024). Therefore, in the fingerprint, substances with identical names are differentiated as monomers (M) and dimers (D).

A total of 65 volatile compounds were identified across the three groups of samples, which included 24 aldehydes, 11 alcohols, 13 ketones, 10 esters, 4 olefins, 1 acid, and 2 furans. These compounds were detected in all three sample groups; however, their concentrations exhibited significant differences, resulting in distinct odors among the samples. As illustrated in the figure, the volatile content within the red-framed areas of DR, YR, and HR was significantly higher than that of the other two groups. Certain volatile compounds, such as 2-hexanone, acetic acid, ethyl acetate-M, 2,3-pentanedione, and hexanal-M, displayed elevated levels in all three sample groups, suggesting a potential association with off-flavor. To investigate the compounds related to off-flavor among the 65 substances, the ROAV was calculated, leading to the identification of 22 characteristic volatiles (Fig. 1E). Specifically, 21, 19, and 14 characteristic volatiles were identified in DR, HR, and YR, respectively, indicating that DR may contain some of the odor profiles from the other two roasted fish products when reheated simultaneously. Furthermore, there were 12 characteristic compounds shared among the three sample groups, including nonanal, octanal (M), heptanal (M), hexanal (M), hexanal (D), (Z)-4-heptenal, 3-methylbutanal (M), 3-methylbutanal (D), 2-methylbutanal (D), 1-octen-3-ol, 1-penten-3-ol, and 1-octen-3-one.

Nonanal, octanal, heptanal, hexanal, (Z)-4-heptenal, 1-octen-3-ol, and 1-penten-3-ol are primarily derived from the oxidative decomposition of oleic, linoleic, linolenic, and arachidonic acids, contributing grassy, fatty, and fishy flavors. Additionally, 3-methylbutanal and 2-methylbutanal are aldehydes produced during drying, resulting from the degradation of leucine and isoleucine, and impart nutty, cheesy, and savory flavors (Zhang et al., 2020). 1-Penten-3-one, a product of fat oxidation, is characterized by a fishy and sour odor. With the exception of 3-methylbutanal and 2-methylbutanal, all the aforementioned compounds are lipid oxidation products and are widely recognized as the primary contributors to fishy and WOF in meat products (Byrne, Bredie, Mottram, & Martens, 2002; Cheng and Ockerman, 2013). Previous studies indicate that lipid oxidation is the principal pathway for the formation of fishy and WOF (García-Lomillo, Gonzalez-SanJose, Pino-García, Ortega-Heras, & Muniz-Rodríguez, ˜ P., 2017; Zhang et al., 2021). Furthermore, secondary lipid oxidation products, such as hexanal, octanal, nonanal, and 1-octen-3-ol, serve as indicators for evaluating WOF (Zhang et al., 2022a).

3.3. GC-O-MS analysis of odor characteristics of ready-to-heat roasted catfish after reheating

To more accurately screen the off-flavors produced after reheating ready-to-heat roasted catfish, the characteristic volatiles identified through GC-E-Nose and GC-IMS were combined with GC-O-MS for absolute quantification in SIM mode. The calibration equations, correlation coefficients, concentration ranges of reference standards, and odor descriptions pertinent to the quantitative determination of volatile compounds via SPME coupled with GC-O-MS are provided in Table 1. A total of 39 volatile compounds were detected across the three sample groups, comprising 10 aldehydes, 15 alcohols, 2 ketones, 4 esters, 4 pyrazines, 1 olefin, 1 furan, 1 thioether, and 1 other compound (Fig. 2A). The types and concentrations of volatile compounds varied significantly among the three sample groups (Fig. 2B). In the DR group, 29 compounds were detected, showing the highest total content, particularly of aldehydes and alcohols. The HR group contained 21 compounds, with a notably high concentration of alcohol. The YR group revealed 33 compounds, primarily aldehydes, but exhibited the lowest total content. Therefore, aldehydes and alcohols were identified as the key odor compounds across the three sample groups, primarily resulting from lipid oxidation and characterized by low odor thresholds, which impart fatty, grassy, painted, warmed-over, and fishy notes. These compounds are likely the principal contributors to off-flavor in ready-to-heat roasted catfish following reheating.

Fig. 2.

Gas chromatography - mass spectrometry determination of odor characteristics of ready-to-heat roasted catfish after reheating. (A) the stacking diagram of volatile compound species in samples. (B) the heat map of volatile compound content in samples. (C) the volatile compounds with OAV ≥ 1 and their scores.

3.4. Screening of key odor compounds

The presentation of off-flavor is influenced not only by the concentration of volatile compounds but also by their respective odor contributions. To further elucidate the characteristic compounds among the identified volatiles, a screening based on OAV was conducted. This analysis revealed 29 compounds with OAV > 1 across the three sample groups. Specifically, 19, 11, and 21 characteristic substances were identified in the DR, HR, and YR samples, respectively (Fig. 2C). Among these, six common characteristic compounds - hexanal, 1-penten-3-ol, heptanal, pentanol, 1-hexanol, and 1-heptanol—were identified across all three sample sets, all of which are products of lipid oxidation. Hexanal is characterized by a grassy and fatty odor, while 1-penten-3-ol has a fishy and irritating scent. Heptanal presents a metallic and fatty aroma. Both 1-pentanol and 1-hexanol exhibit floral and fruity notes, although they can emit a fishy odor at elevated concentrations (Iglesias et al., 2009). Heptanol is noted for its fatty and rubbery odor.

The 13 characterized substances in the DR included octanal (OAV = 28.30), (E)-2-hexenal (OAV = 65.22), (E,E)-2,4-heptadienal (OAV = 34.18), 3-methylbutanal (OAV = 3536.84), ethanol (OAV = 15.63), benzyl alcohol (OAV = 33.88), 3-methyl-1-butanol (OAV = 3.07), 1-octen-3-ol (OAV = 62.25), linalool (OAV = 9.15), 2,3-octanedione (OAV = 1379.24), hexanoic acid, ethyl ester (OAV = 6.92), Sulfide, allyl methyl (OAV = 210.84), and d-limonene (OAV = 13.93). In the HR, The 5 characterized substances were decanal (OAV = 52.20), benzaldehyde (OAV = 3.42), 1-nonanol (OAV = 1.00), benzyl alcohol (OAV = 100.30), and ethanol (OAV = 6.29). The 15 characterized substances in the YR comprised nonanal (OAV = 41.28), pentanal (OAV = 11.85), octanal (OAV = 34.12), decanal (OAV = 590.52), (E,E)-2,4-heptadienal (OAV = 2.88), benzaldehyde (OAV = 15.94), 3-methylbutanal (OAV = 21,710.36), 1-nonanol (OAV = 1.01), 2,3-octanedione (OAV = 715.42), 2-methylbutanal (OAV = 397.15), 2-methylpyrazine (OAV = 4.15), 2-ethyl-5-methylpyrazine (OAV = 2.45), 2,6-dimethylpyrazine (OAV = 2.58), 2,3-dimethylpyrazine (OAV = 2.20), and d-limonene (OAV = 20.82). The odors associated with these characteristic substances, in conjunction with GC-O analysis identified the typical odors of the samples as earthy, fishy, grassy, warmed-over, metallic, fatty, and nutty.

The odor compounds identified include aldehydes (8), alcohols (6), ketones (1), esters (2), pyrazines (4), thioether (1), and olefins (1). Among these, aldehydes are known to produce nutty, fatty, and grassy odors; for instance, octanal and nonanal contribute to grassy and fatty notes. Additionally, decanal, pentanal, (E)-2-hexenal, and (E,E)-2,4-heptadienal have been reported to be associated with off-flavors. Benzaldehyde is characterized by its nutty aroma, while 3-methylbutanal imparts nutty, cheesy, and salty notes. Most of these aldehydes arise from fat oxidation, possess low odor thresholds, and interact with other compounds, thereby intensifying the odor profile of roasted fish. Benzyl alcohol contributes a floral aroma, whereas 3-methyl-1-butanol offers musty and nutty flavors. Notably, 1-octen-3-ol (OAV = 313.61) is recognized for its distinct “mushroomy” odor. The majority of these alcohols are derived from linoleic and oleic acids as precursors. Upon reheating, the “fruity” flavor of roasted catfish is predominantly generated by ketones and esters, such as hexanoic acid, ethyl ester, and 2-methyl-butanoic acid, methyl ester. Pyrazines are known for their roasted, nutty aroma and are significant contributors to the overall aroma of roasted fish. Apart from lipid oxidation, protein degradation also results in the loss of desirable aromas, contributing to off-flavors (Pegg, Kerrihard, & Shahidi, 2014; Ruenger, Reineccius, & Thompson, 1978; Zhang et al., 2022a). Additionally, linalool (floral), sulfides, allyl methyl (garlicky, pungent odor), and d-limonene (citrusy, lemony) are characteristic odors of spices, primarily utilized to mask undesirable flavors.

3.5. Odor recombination analysis

To assess the accuracy of identification and quantification of key odor compounds after reheating three groups of roasted catfish, compounds with OAVs ≥1 were recombined by integrating them into odorless matrices for each group. This process simulated the temperatures and times of actual measurements, and recombination experiments were conducted through sensory evaluation. Seven odor attributes were selected for descriptive analysis in our odor model, including earthy, fishy, grassy, warmed-over, metallic taste, fatty, and nutty. The odors of the recombined models were then compared to the characteristic odors of the original samples by a trained sensory panel. The similarity indices for DR, HR, and YR were found to be 84.98 %, 78.35 %, and 78.83 %, respectively, indicating a relatively accurate characterization of key odor compounds in each group (Fig. 3A). Recombinant model A exhibited the highest similarity to the original model, where grassy, warmed-over, fatty, and nutty odors were found to be less pronounced than in the original model, while earthy and fishy odors were more closely aligned, and metallic odors were stronger. This suggests that the key compounds screened in the DR may better characterize the odor profile of the original samples. However, other compounds' lack of odor modification led to the perception of undesirable odors. Recombinant model B demonstrated the lowest similarity to the original model, particularly regarding the fatty and nutty odors, indicating that these compounds may have been overlooked during the identification and screening of HR. The difference in similarity between recombinant model C and the original model was primarily attributed to the warmed-over flavor, which may have mitigated the perception of undesirable odors due to the identification of various compounds presenting nutty aromas in the YR. In summary, recombinant model A was relatively well characterized among the three sets of samples.

Fig. 3.

A recombination model was constructed to screen for key off-flavor substances in the samples. (A) DR, HR and YR recombination models and odor profiles of the original samples. (B) Structural diagram of key off-flavor substances in ready-to-heat roasted catfish after reheating.

3.6. Identification of key off-flavor compounds by omission experiments

To identify the key odor compounds contributing to off-flavor, omission experiments were conducted on the 19 characteristic volatiles in the recombinant model A based on DR, with the omitted compounds categorized according to Table 4. The exclusion of aldehydes exhibiting grassy, fishy, fatty, and metallic odors—such as hexanal, heptanal, (E)-2-hexenal, octanal, (E,E)-2,4-heptadienal, and 3-methylbutanal significantly reduced the off-flavor. Similarly, the omission of substances with fatty, earthy, rancid, and plastic odors, including 2,3-octanedione, 1-octen-3-ol, heptanol, 3-methyl-1-butanol, benzyl alcohol, and hexanoic acid ethyl ester, also led to a significant reduction in off-flavor. In contrast, the exclusion of 1-penten-3-ol, 1-pentanol, 1-hexanol, and ethanol, which possess floral, sweet, and pungent odors, significantly enhanced the off-flavor in the omission model, indicating that these compounds are negatively correlated with off-flavor. Furthermore, sulfide, allyl methyl, d-limonene, and linalool are volatile compounds unique to spices, and their garlicky, pungent, citrusy, lemony, and floral aromas can mask off-flavor, which is exacerbated by the absence of these three substances. In summary, hexanal, heptanal, (E)-2-hexenal, octanal, (E,E)-2,4-heptadienal, 3-methylbutanal, 2,3-octanedione, 1-octen-3-ol, heptanol, 3-methyl-1-butanol, benzyl alcohol, and hexanoic acid ethyl ester are identified as key contributors to off-flavor in reheated ready-to-heat roasted catfish. Meanwhile, sulfide, allyl methyl, d-limonene, and linalool may serve as masking agents to modulate the perception of off-flavor.

Table 4.

Omission Tests from the recombination model A.

Number	Compounds	Odor description	n/18a	Significanceb	Effects
1	3-Methylbutanal	Malty	11	*	+
2	Sulfide, allyl methyl	Garlicky	15	***	−
3	Hexanal	Grassy	14	***	+
4	1-Penten-3-ol	Rubbery	10	*	−
5	Heptanal	Fatty, warmed-over, creamy	12	**	+
6	d-Limonene	Citrusy, lemony	17	***	−
7	3-Methyl-1-butanol	Fruity, floral	10	*	+
8	(E)-2-Hexenal	Nutty, grassy	13	***	+
9	Hexanoic acid, ethyl ester	Fruity, floral	10	*	+
10	1-Pentanol	Floral	10	*	−
11	Octanal	Fatty, warmed-over	14	***	+
12	2,3-Octanedione	Warmed-over	15	***	+
13	1-Hexanol	Floral	12	**	−
14	1-Octen-3-ol	Mushroom-like, earthy, fishy,	16	***	+
15	1-Heptanol	Rubbery	11	*	+
16	(E,E)-2,4-Heptadienal	Fatty, warmed-over, fishy	13	***	+
17	Linalool	Floral	15	***	−
18	Benzyl Alcohol	Green, floral	11	*	+
19	Ethanol	Pungent odor	10	*	−

ns, not significant; “+,” the omission of these compounds attenuates the off-flavor, which means that these compounds are a source of the off-flavor; ‘-,’ the omission of these compounds enhances the off-flavor, which means that these compounds can mask the off-flavor.

^aNumber indicates a significant odor difference among the samples the triangle test based on 18 panelists.

^bSignificance: ***p < 0.001; **p < 0.01; *p < 0.05.

3.7. Construction of a multivariate recombination model for key off-flavor substances

Recombination and omission tests can be employed to evaluate the contribution of individual odor-active compounds to the overall odor profile of a sample, as well as to identify key odors within the sample (Grimm & Steinhaus, 2019; Sabbatini et al., 2019). However, interactions among multiple odor compounds can influence odor perception. Consequently, in this experiment, a multivariate recombination model was developed based on the omission test for the 12 substances previously screened. The off-flavor compounds resulting from the reheating of roasted fish were further confirmed through sequential omission. The recombination model was constructed with reference to Table 3, and the 12 screened substances were recombined into the original model. As illustrated in Fig. 4, the grassy flavor of recombinant model 1 was lower than that of the original model, exhibiting 96.22 % similarity when substances 1–11 were included and benzyl alcohol was omitted. This indicates that benzyl alcohol may not significantly contribute to the off-flavor. In recombinant model 2, the addition of substances 1–10, along with the omission of benzyl alcohol and (E,E)-2,4-heptadienal, resulted in significantly reduced fishy and overripe flavors compared to the original model, while the grassy flavor was notably higher, achieving a similarity of 84.44 %. This suggests that the absence of (E,E)-2,4-heptadienal plays a significant role in the overall off-flavor profile of the model. Additions and omissions on a case-by-case basis revealed that recombinant models 3–12 exhibited similarities of 95.84 %, 87.14 %, 95.17 %, 84.68 %, 95.43 %, 86.11 %, 81.47 %, 82.85 %, 85.47 %, and 95.08 % to the original model, respectively. Among these, restructuring models 2, 4, 6, 8, 9, 10, and 11 demonstrated less than 90 % similarity to the original model. Consequently, the absent substances in these seven models specifically hexanal, heptanal, (E)-2-hexenal, octanal, 3-methyl-1-butanol, 1-octen-3-ol, and (E,E)-2,4-heptadienal—were identified as key off-flavor compounds in the reheating of the ready-to-heat roasted catfish (Fig. 3B). Previous studies have indicated that fat oxidation products such as hexanal, heptanal, octanal, and 1-octen-3-ol serve as potential indicators for the evaluation of WOF (Kim et al., 2016; O'Sullivan, Byrne, Jensen, Andersen, & Vestergaard, 2003; Zang et al., 2020; Zhang et al., 2021). Additionally, heptanal, octanal, and (E,E)-2,4-heptadienal have been recognized as contributors to fishy and WOF characteristics in surimi gels, as reported by An et al. (2022).

Fig. 4.

Multivariate reorganization models and their similarity analysis. Original:add 3-methylbutanal, hexanal, heptanal, 3-methyl-1-butanol, (E)-2-hexenal, hexanoic acid, ethyl ester, octanal, 2,3-octanedione, 1-octen-3-ol, 1-heptanol, (E,E)-2,4-heptadienal, benzyl alcohol. Reorganized model 1–12 is added according to Table 3. The numbers represent the degree of similarity.

4. Conclusions

After reheating, the ready-to-heat roasted catfish exhibited a pronounced off-flavor, characterized primarily by fatty, grassy, hard-boiled egg, metallic, and fishy notes. Utilizing a sensomics approach, key off-flavor compounds identified include hexanal, heptanal, (E)-2-hexenal, octanal, 3-methyl-1-butanol, 1-octen-3-ol, and (E,E)-2,4-heptadienal, which may serve as potential markers for off-flavor development following the reheating of ready-to-heat roasted catfish. These markers predominantly arise from lipid oxidation. Additionally, compounds such as sulfide, allyl methyl, d-limonene, and linalool may function as masking agents, modulating the perception of off-flavor. Furthermore, the loss of aroma compounds exacerbates off-flavor. Consequently, the increase in lipid oxidation products, protein degradation, and the introduction of exogenous aroma spices, along with their interactions, may significantly influence the emergence of off-flavor. Further research is warranted to elucidate the mechanisms underlying off-flavor production and to explore the integration of masking agents to enhance the flavor quality of prepared products.

CRediT authorship contribution statement

Mingzhu Zhou: Writing – review & editing, Writing – original draft, Visualization, Methodology, Data curation. Yiting Lu: Writing – original draft, Methodology, Formal analysis, Data curation. Jinyu Yu: Validation, Software, Data curation. Chao Wang: Writing – review & editing, Supervision. Wei Yu: Supervision, Resources. Liu Shi: Supervision, Conceptualization. Wenjin Wu: Writing – review & editing, Supervision. Lan Wang: Resources, Funding acquisition. Yu Qiao: Writing – review & editing, Supervision.

Declaration of competing interest

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

Data availability

The data that has been used is confidential.