Inclusion of Dry-Aged Beef Trimmings as a Quality and Flavor Enhancer for Ground Beef

Sample collection and aging process

Details on the aging process can be found in our published parallel study (Setyabrata et al., 2022b). Briefly, paired beef loins (M. Longissimus lumborum) were collected at 5 d postmortem from a commercial processing facility and transported to Purdue University Meat Laboratory. The loins were then split into 2 equal sections (4 sections per paired loins) and randomly assigned to 4 different aging treatments (wet aging [WA], conventional dry aging [DA], dry aging in water-permeable bag [DWA] and UV-light dry aging [UDA]). All sections were aged for 28 d at 2°C, 65% relative humidity, and 0.8 m/s airflow. The UDA samples were given 2 UV-light treatments per day, each lasting for 5 min and totaling a dose of 5 J/m² (Philips TUV T8 UVC light, Eindhoven, the Netherlands). At the end of aging, dried surfaces were removed from all sections. The sections were then trimmed, and the lean trim (mainly consisting of M. Multifidus dorsi) and fat trim from each sample were collected, individually packed, and stored at −40°C until application.

Beef patties processing

A total of 3 independent batches of ground beef patties were made for the study utilizing the collected trimmings from the previously described beef loins. For each batch, trimmings from the beef loins from 4 carcasses were combined to produce ground lean and ground fat for inclusion in beef patties. The carcasses were selected to maximize the quantity of ground lean and fat trim quantity that can be utilized within each independent batch. Both lean and fat trimmings originating from the same carcasses were used within the same batch. The same carcass combinations were used for all the different aging treatments within the same batch. Trimmings were thawed at 2°C for 24 h prior to processing.

For each independent batch, fresh beef top round (M. Semimembranosus) and subcutaneous fat were collected at 3 d postmortem to be combined with the collected trimmings (aged lean and fat) from different aging treatments. All samples were ground through a meat grinder equipped with an 8 mm plate (M-12-FS, Torrey; Monterrey, NL, Mexico). The fresh ground meat, fresh ground fat, aged ground meat, and aged ground fat were then combined to make 80:20 (lean:fat) beef patties with 6 different formulations (Table 1). The DA fat only (DA-FAT) was added to identify the impact of only adding aged fat trimming, compared to the inclusion of both aged fat and lean trimmings. The mixture was manually mixed by hand for 5 min and re-ground for uniformity. At least 6 patties (125 g) were collected from each treatment for further analyses described below. Samples for simulated color display, cooking loss, and texture profile analysis were immediately used for the analyses. Samples not immediately used were vacuum packed individually and frozen at −80°C until analyses.

pH and cook loss measurements

The pH was measured following the method described by Nondorf and Kim (2022). A total of 3 g of samples was homogenized in 27 mL of double distilled water. The pH was then measured using a benchtop pH meter (Sartorius Basic Meter PB-11, Sartorius AG; Goettingen, Germany) calibrated following the manufacturer’s guidelines. The pH measurements were conducted in duplicate.

The cooking loss was used to measure the water-holding capacity of the samples. The cooking loss was determined by measuring the initial and final weights of the patties following cooking and was expressed as a percent loss. The patties were cooked using a double-sided clamshell griddle (Griddler GR-150, Cuisinart; Glendale, AZ, USA) to 71°C internal temperature. After cooking, the samples were allowed to rest for 10 min before being weighed. Before weighing, all samples were gently blotted using paper towels. The cook loss (%) was measured twice and calculated using the following formula: (Initial weight − Final weight)/Initial weight × 100.

Simulated color display

One patty from each treatment was collected, placed on a foam tray with a soaking pad, and overwrapped using Polyvinyl Chloride (PVC) films. The patties were then displayed under light (1800 lx, OCTRON^® T8 Lamps, Osram Sylvania LTD, Canada) for 5 d. The color of the patties was evaluated daily, using a Hunter MiniScan EZ-4500L Spectrophotometer (Hunter Associates Laboratory, Inc.; Reston, VA, USA) equipped with a 25 mm (diameter) opening. Illuminant A and a standard 10° observer were used. The equipment was calibrated to black glass and white tile prior to any color measurement. The CIE L*, a*, and b* color values were collected from 3 random locations on the surface of the patties. Hue angle (tan⁻¹(b*/a*)) and Chroma ((a*² + b*²)^½) value were calculated (King et al., 2023). All samples were vacuum packaged and frozen at −80°C until used for lipid oxidation analysis following the completion of the simulated display.

Lipid oxidation analysis

The 2-thiobarbituric acid reactive substances (TBARS) assay was conducted using the method described by Setyabrata and Kim (2019). The lipid oxidation measurement was performed on samples before and after the simulated display. Briefly, 5 g of samples was homogenized in 15 mL of double distilled water with 50 μL of butylated hydroxyl anisole. The homogenate was then mixed with 20 mM 2-thiobarbituric acid solution in 15% trichloroacetic acid solution. The mixture was mixed, heated in 80°C water bath, cooled in ice water for 10 min, and centrifuged at 2000 × g for 10 min. The mixture was then filtered through filter paper and the absorbance was read at 531 nm using an Epoch^™ Microplate Spectrophotometer (BioTek Instrument Inc., Winooski, VT, USA). The TBARS value was expressed as mg malondialdehyde/kg meat.

Texture profile analysis

The texture profile analysis was performed on patties previously used for the cook loss measurement. Following the cooking, samples were stored at 2°C overnight prior to texture analysis. A total of 4 cores were collected from each patty (d = 2.5 cm) and subjected to texture analysis using TA-XT Plus Texture Analyzer (Stable Micro System Ltd., Godalming, UK). The parameter of the test was set following the description by Xue et al. (2021). The hardness (g), adhesiveness (g.sec), resilience (%), cohesiveness (%), springiness (%), and chewiness were obtained.

Descriptive sensory analysis

The sensory characteristics of the beef patties were evaluated by 6 trained panelists recruited from the Department of Animal Sciences at Purdue University (West Lafayette, IN, USA). The panelists were trained following the American Meat Science Association Research Guidelines for Cookery, Sensory Evaluation, and Instrumental Tenderness Measurements of Meat (AMSA, 2016). The panelists were trained to evaluate 10 different sensory attributes (i.e., beefy/brothy, brown roasted/grilled, bloody, metallic, rancid, umami, salty flavors notes, hardness, and cohesiveness as well as moisture sensory feelings) as described by Xue et al. (2021). The patty evaluation was conducted over 3 sessions, with the sample serving order randomized during each session. The panelists were seated in an individual booth in a room equipped with red lighting.

The patties were cooked in a similar manner as described previously in the cooking loss measurements. Following cooking, patties were cut into 6 wedge-shaped pieces, individually placed in a sampling cup with a lid, and stored in a warmer. All samples were served to the panelist within 10 min. The panelists were supplied with unsalted crackers and water to cleanse their palate between samples. The patties were evaluated using a 15-point anchor, with 0–5 points indicating an intensity level of slight, a 6–10 point scale indicating medium intensity, and 11–15 points equivalent to strong intensity. All scores were then pooled and subjected to statistical analysis. Cooked patties samples were also collected for volatile and metabolomics analysis.

Volatile compounds analysis

Profiling of volatile compounds was conducted following the methods described by Gardner and Legako (2018). Cooked patties were minced and placed into a gas chromatography (GC) vials with 10 μL of 1,2-dichlorobenzene as an internal standard. The vials were then sealed and loaded to a Gerstel agitator (Gerstel Inc., Linthicum Heights, MD, USA) for incubation (5 min, 65°C) prior to 20 min of extraction via headspace solid-phase microextraction. The extracted compounds were then injected into a capillary column (30 m × 0.25 mm × 1.0 μm, Agilent Technologies, Inc.; Santa Clara, CA, USA), and selective ion monitoring scan mode was utilized for data collection. The compounds were identified by comparing them to an external authentic standard (Sigma-Aldrich, St. Louis, MO, USA) for validation. The volatile compounds concentrations were reported in nanograms per gram of cooked sample.

Untargeted metabolomics analysis

Metabolite extraction. Metabolomics analysis was conducted on both raw and cooked samples. Prior to metabolite extraction, samples were powdered by submerging the patties into liquid nitrogen and immediately pulverized using a blender (Waring Products, CT, USA). Then, 100 mg of the powdered meat samples was homogenized in 300 μL chloroform and 300 μL methanol using Precellys 24 tissue homogenizer (Bertin Instruments, Bretonneux, France) for the extraction as described by (Setyabrata et al., 2021a). The samples were homogenized through 3 cycles of 30 s at 6500 rpm followed by 30 s rest between the cycles. Water was then added to the homogenized mixture and centrifuged at 16,000 × g for 8 min. The upper layer was collected and dried for chromatographic separation.

Ultra-performance liquid chromatography–mass spectrometry. Following extraction, untargeted metabolomics analysis was conducted using the methods previously detailed by Setyabrata et al. (2022a). The samples were separated using the Agilent 1290 Infinity II UPLC system (Agilent Technologies, Palo Alto, CA, USA) equipped with a Waters Acquity HSS T3 (2.1 × 100 mm × 1.8 μm) separation column (Waters, Milford, MA, USA). The column was maintained at 40°C with the binary mobile phase flow set at 0.45 mL/min. The binary mobile phase consisted of solvent A (0.1% formic acid (v/v) in ddH2O) and solvent B (0.1% formic acid (v/v) in acetonitrile). Initial conditions of 100:0 A:B were held for 1 min, followed by a linear gradient to 70:30 over 15 min, changed to a linear gradient of 5:95 over 5 min, and 5:95 hold for 1.5 min. The sample injection volume was 5 μL.

Separated compounds were identified using Agilent MassHunter B.06 software (Agilent Technologies, Santa Clara, CA, USA), and the mass accuracy was improved by infusing Agilent Reference Mass Correction Solution (G1969-85001, Agilent Technologies; Santa Clara, CA, USA). The Agilent ProFinder (v B.08) was utilized for peak deconvolution. Peak identification was improved by applying data-dependent tandem mass spectrometer (MS/MS) collection on composite samples with 10 eV, 20 eV, and 40 eV collision energy. The metabolites were annotated using MS-DIAL software and database (http://prime.psc.riken.jp/).

Statistical analysis

This study was a randomized complete block design with the treatments (CON, DA-FAT, DA, DWA, UDA, and WA) serving as the main fixed effects and the batch serving as a both block and an experimental unit. Within each batch, the treatments were randomly assigned to patties, which served as pseudo-replicates. For the analyses subjecting the patties to the display, the storage period was also included as a fixed effect in the model. The data were analyzed using PROC GLIMMIX procedure from SAS 9.4 software (SAS Institute Inc., Cary, NC). Least-squares means for all traits were separated, and the significance level was defined at the level of P < 0.05. The trend was defined at the level of 0.05 ≤ P ≤ 0.10.

Metabolomics analysis was conducted using Metaboanalyst 5.0 (Pang et al., 2021). The metabolites were also analyzed using ANOVA to identify features significantly affected by the aging treatment. The Student’s t-test was utilized to identify significant features between cooked and raw samples. Significance was defined at P < 0.01 and adjusted using the false discovery rate (FDR) method (adj P < 0.01). Unsupervised principal component analysis (PCA) and hierarchical clustering analysis (HCA) were performed to visualize the data. Unsupervised principal component analysis was conducted as an explorative analysis on the metabolomics data.

pH, water-holding capacity, and texture profile analyses

The pH of the beef patties was found to be similar across the treatments (Table 2). However, there was a trend (P = 0.0876) of increasing pH with the addition of lean and fat trim, regardless of the aging treatment applied to the trims. The increase in the pH could potentially be attributed to the higher pH of the trims. The dry-aged meat products were reported to have a pH between 5.6 and 5.7 following the aging treatment in our parallel study (Setyabrata et al., 2022b), explaining the currently observed increased pH. Park et al. (2020) and Lee et al. (2022) found that the addition of dry-aged beef crust trimmings influenced the final product pH. Park et al. (2020) reported that the pH increased with the addition of beef crust in brown sauce, while Lee et al. (2022) observed that the pH decreased with the addition of beef crust to pork patties. This highlights the potential quality variation of dry-aged beef trimmings. Hence, consideration should be given when applying dry-aged beef trimmings as it could alter final product pH and differently influence final product functionality and quality.

No significant effect of aged beef trim inclusion was observed on cook loss measurement (Table 2). Different from the current study, Park et al. (2018), Xue et al. (2021), and Lee et al. (2022) reported that the addition of dry-aged beef trims reduced the cook loss compared to the control. The different observation could potentially be attributed to the fact that those authors utilized the dehydrated crust trim portion in a lyophilized powder form compared to the ground lean and fat trim portion utilized in the current study. It was suggested by those authors that the lyophilized meat powder could be rehydrated when exposed to moisture, thus allowing greater moisture retention following the cooking process.

The addition of aged beef trimmings to the beef patties significantly influenced the chewiness of the product (P < 0.05, Table 2). The current result showed that DWA had lower chewiness compared to CON (P < 0.05), while both DWA and CON samples were not different from DA-FAT, WA, DA, and UDA samples (P > 0.05). The hardness, adhesiveness, resilience, cohesion, and springiness of the beef patties samples were not impacted by the addition of differently aged beef trimmings (P > 0.05). The current results indicated that the addition of aged beef trimmings minimally impacts the textural properties of the final beef patties, regardless of the aging methods. The current reported results, however, were not in line with previous studies adding dry-aged beef trimmings to meat patties. While different, the results of the current study were not surprising as the aged beef trimmings were only minimally dehydrated from the aging process and were not further processed (lyophilized) prior to inclusion. Xue et al. (2021) reported that the addition of lyophilized dry-aged crust trimmings increased the hardness, adhesiveness, gumminess, and chewiness of beef patties. Similarly, Lee et al. (2022) found that the addition of lyophilized dry-aged beef crust trimmings increased the shear force of pork patties, indicating increased hardness. Both authors attributed the changes in texture profile to the higher protein content of the lyophilized dry-aged trimmings, allowing for stronger protein binding. As such, it could be assumed that the aged beef lean and fat trimmings had a more similar physiochemical quality and protein content to those of regular beef compared to the dehydrated dry-aged beef crust.

Color and lipid oxidation stability

Among the color measurements, hue angle (instrumental discoloration) was the only color trait significantly impacted by treatment and storage interaction (P < 0.05, Figure 1). The inclusion of DWA trimmings in beef patties caused greater discoloration compared to other treatments. An overall increase in hue angle value was observed across the different samples throughout the display. However, the DWA samples had a higher hue angle value on day 5 of the display compared to other treatments (P < 0.05). No significant interaction effect was observed for L*, a*, b*, and chroma.

Figure 1.

Impact of treated trimmings inclusion on hue angle changes of beef patties during 5 d of the display period. Different formulation treatments: CON (80% fresh beef + 20% fresh fat), DA-FAT (80% fresh beef + 20% DA fat), WA (50% fresh beef + 30% WA lean + 20% WA fat), DA (50% fresh beef + 30% DA lean + 20% DA fat), DWA (50% fresh beef + 30% DWA lean + 20% DWA fat), and UDA (50% fresh beef + 30% UDA lean + 20% UDA fat). ^a,bDifferent letters indicate significant differences between treatments within the same display day (P < 0.05).

A significant treatment effect was observed for a*, b*, and chroma (Table 3). The instrumental redness (a*) and yellowness (b*) were observed to be higher in CON compared to UDA and DWA (P < 0.05), while DA-FAT, DA, and WA samples were not different from those treatments (P > 0.05). CON had higher chroma values compared to DWA (P < 0.05), which were not different compared to all other treatments (P > 0.05). The L*, a*, b*, and chroma were impacted by storage effect, showing a significant decrease at the end of the simulated color display compared to the beginning of the display regardless of treatment (P < 0.05, Table 3).

The color of dry-aged beef has often been reported to be darker compared to its wet-aged counterpart due to the dehydration process (Dikeman et al., 2013; Ribeiro et al., 2021). Based on the current simulated color display results, it was found that the inclusion of aged beef trimming had a minimal impact on the initial color of the beef patties. During the display, however, the addition of DWA trimmings negatively affected the color of the beef patties, exhibiting rapid color degradation and accelerated discoloration during the display. Similar results were previously reported by Setyabrata et al. (2022b), in which the authors found that DWA samples had greater hue angle and visual discoloration by day 5 of the display compared to all other aging methods. The authors speculated that the utilization of dry-aged bags hindered the dehydration process, inducing more oxidation and thus leading to lower color stability during display. Thus, it could be postulated that the inclusion of DWA beef trimmings impacted the overall observed color of the beef patties and potentially exacerbated the discoloration rate of the patties during retail display. The current results differ when compared to other studies utilizing dry-aged beef crust trimmings, in which the additional trimmings significantly reduce the lightness of the final product (Lee et al., 2022; Xue et al., 2021). It was reported that the addition of lyophilized dry-aged beef crust trimmings caused significant darkening in the final products, mainly because the crust powder had darker color and reduced surface moisture as the dehydrated powder retained more moisture (Lee et al., 2022; Xue et al., 2021). Similar to the current study, Xue et al. (2021) reported a decrease in a* following dry-aged crust inclusion in beef patties; however, Lee et al. (2022) reported a conflicting impact on the a*, showing an increase with dry-aged crust inclusion in pork patties. This result suggests the need to further evaluate the impact of aged trim inclusion on final product color quality.

Only storage effect was found to be affecting the lipid oxidation (P < 0.05, Figure 2). No significant treatment and treatment × storage interaction effects were observed. A greater increase in lipid oxidation was observed in all samples following the display (P < 0.05), regardless of treatments applied. Our findings are in agreement with previous studies, observing minimal differences in the extent of lipid oxidation from different aging treatments based on TBARS measurement (Setyabrata et al., 2022b; Xue et al., 2021).

Figure 2.

Impact of treated trimmings inclusion on lipid oxidation (TBARS) of beef patties before and after 5 d display. Different formulation treatments: CON (80% fresh beef + 20% fresh fat), DA-FAT (80% fresh beef + 20% DA fat), WA (50% fresh beef + 30% WA lean + 20% WA fat), DA (50% fresh beef + 30% DA lean + 20% DA fat), DWA (50% fresh beef + 30% DWA lean + 20% DWA fat), and UDA (50% fresh beef + 30% UDA lean + 20% UDA fat). ^a,bDifferent letters indicate significant differences within the different display periods (P < 0.05).

Trained sensory panel and volatile compound analysis

Trained panel sensory evaluation revealed that the addition of DA and UDA trims decreased the perceived bloody flavor in the final beef patties compared to CON, DA-FAT, and WA patties (P < 0.05, Table 4). The bloody flavor score for DWA patties was not different compared to all treatments (P > 0.05). Similarly, there was a tendency (P = 0.087) that the addition of DA and UDA trims reduced the metal flavor of the patties compared to other treatments. No significant impact was observed on beefy, brown roasted, rancid, umami, salty, hardness, cohesiveness, and moisture sensory attributes from the different beef patties.

It is to our surprise that the beef patties’ flavor was not impacted by the addition of the aged beef trimmings. Previous reports by Park et al. (2018), Xue et al. (2021), and Lee et al. (2022) found that the addition of dry-aged beef crust trimmings significantly improved the flavor scores by the trained panel. Xue et al. (2021) further reported that the addition of lyophilized dry-aged beef crust trimming significantly increased the brown roasted flavor and exhibited a strong trend of increasing umami flavor. Those authors suggested that the increased flavor was due to the increased concentration of the flavor precursor in the crust portion as a result of the dehydration process during dry aging. In the current study, the crust trim portion was not utilized in patty manufacturing. It is possible that the aged beef lean and fat trimmings minimally concentrated the flavor precursor compared to the crust portion as they were not immediately exposed to the environment, thus limiting the dehydration and flavor-enhancing ability.

The addition of DA trimmings decreased the bloody and tended to reduce the metal flavor of the patties (P = 0.087); however, only the addition of lean trim decreased the off flavors as the sole addition of aged fat trim in DA-FAT samples did not lead to the same result. The presence of metallic and bloody notes is often attributed to the increase in myoglobin and heme iron concentration in the meat product (Yancey et al., 2006), indicating that the metallic and bloody flavor is mainly governed by the lean portion of the meat. Interestingly, the current findings showed that DA and UDA trimming significantly reduced the metallic flavor, followed by DWA, while the addition of WA trimming did not impact the flavor. Previous reports by Setyabrata et al. (2021a) suggested that the dry-aging process might cause modification to myoglobin or heme iron, reducing the metallic flavor despite increasing the concentration of those compounds. It is possible that the modification through dry aging mitigated and reduced the development of these flavors, although further studies are still required to confirm this speculation.

The different addition of aged beef lean and fat trim only minimally influenced the volatile profile of the beef patties (Table 5). A total of 69 volatile compounds were detected. Of the detected volatile compounds, only 2-acetyl pyrrole was found to be affected by the addition of aged beef trimmings, exhibiting greater concentration with the addition of DA lean and fat trims compared to all other treatments (P < 0.05), except UDA where they were not different (P > 0.05). Three compounds tend to increase with the addition of the DA and UDA-aged beef trimmings (2-pentanone [P = 0.0925], 2-heptanone [P = 0.0627], and furfural [P = 0.0919]). Of the 4 affected compounds, 3 were found to be from the ketone group and one from the furan group.

The observed volatile results corroborate with the trained sensory panel observation in this study, exhibiting that the addition of aged beef trims minimally influenced the flavor profile of the beef patties. However, the results indicated that the DA and UDA process might significantly alter the lean and fat trims, generating a greater potential to alter and improve the final product flavor when added to beef patties. The identified compounds, 2-acetyl pyrrole, 2-pentanone, 2-heptanone, and furfural, were previously found to produce a positive flavor attribute in meat products, producing nutty, sweet, citrus, and almond odor, respectively (Sohail et al., 2022). While those compounds could positively impact the flavor of the final products, the concentration of the compounds might not be enough to induce the expected aroma. It is possible that the detection threshold of those compounds was higher and therefore did not meaningfully impact the flavor perceived by the sensory panelist. Further study to identify the flavor generation following different cooking techniques might be beneficial to understand and maximize the flavor generation process.

Metabolomics profile analysis

The addition of different aged beef lean and fat trims induced a distinct impact on the metabolomics composition of the samples before and after that cooking process. In the raw samples, PCA and HCA results (Figure 3) exhibited that the CON, DA-FAT, DA, and UDA patties had a more similar metabolite profile compared to the WA and DWA patties (significant metabolites presented in Table S1). The PCA and HCA of the cooked patties (Figure 4), however, revealed a different pattern. The analyses showed a clear clustering of WA, DA, DWA, and UDA treatments from the CON and DA-FAT treatments (significant metabolites presented in Table S2). The observation indicated that the aging process potentially alters the initial flavor precursor presence within the raw products. While different, however, the aged trims potentially went through a similar reaction as the final cooked products showed a similar metabolomics and flavor profile regardless of the aging method applied. It was previously reported that different dry-heat cookery strongly influenced the flavor production, thus affecting the perceived final meat flavor (Vierck et al., 2021). As such, it could be speculated that while the different aging methods alter the presence of the precursors, the cooking process applied might not promote the desired reaction to maximize flavor production, resulting in a product with similar flavor quality. Furthermore, while different, it is possible that the abundance of the flavor precursor might be too low to influence the flavor generation process, limiting the volatile compounds and subsequently flavor development. This postulation, however, needs to be confirmed.

Figure 3.

Principal component analysis (PCA) and hierarchical clustering analysis (HCA) of metabolites from raw beef patties made with different formulations. Different formulation treatments: CON (80% fresh beef + 20% fresh fat), DA-FAT (80% fresh beef + 20% DA fat), WA (50% fresh beef + 30% WA lean + 20% WA fat), DA (50% fresh beef + 30% DA lean + 20% DA fat), DWA (50% fresh beef + 30% DWA lean + 20% DWA fat), and UDA (50% fresh beef + 30% UDA lean + 20% UDA fat).

Figure 4.

Principal component analysis (PCA) and hierarchical clustering analysis (HCA) of metabolites from cooked beef patties made with different formulations. Different formulation treatments: CON (80% fresh beef + 20% fresh fat), DA-FAT (80% fresh beef + 20% DA fat), WA (50% fresh beef + 30% WA lean + 20% WA fat), DA (50% fresh beef + 30% DA lean + 20% DA fat), DWA (50% fresh beef + 30% DWA lean + 20% DWA fat), and UDA (50% fresh beef + 30% UDA lean + 20% UDA fat).

A clear clustering between raw and cooked samples was observed through PCA and HCA (Figure 5), showing a clear impact of cooking on the metabolite profile of the beef patties, regardless of the formulations. It could be observed from the PCA results that cooking decreased the variation of metabolites between treatments, generating products with more similar metabolites when compared to the raw samples. A total of 21 metabolites were identified through MS/MS match and found to be significantly influenced by the cooking process (FDR < 0.01, Table S3), regardless of the formulation. Of those, 10 metabolites were found to be more abundant in the raw patties and can be loosely grouped into nucleotides (adenosine diphosphate and hypoxanthine), amino acids/peptides (leucine and leucylleucine), and lipids (acetamidomethylpropyl acetate, glycyrrhetic acid, harringtonine, hydroxyquinazolinyl acetamide, gelomulide N, and phosphocholine). Eleven metabolites were observed to be greater in cooked patties and predominantly belonged to nucleotides (5'-S-methylthioadenosine, adenine, adenosine monophosphate, anabasamine, hypoxanthine, and NAD), followed by amino acids/peptides (cysteinylglycine, glutathione, and methionine) and vitamin (niacinamide). Interestingly, more umami-related compounds were found to be greater in abundance in the cooked samples compared to the raw samples. Greater abundance of 5'-S-methylthioadenosine, adenosine monophosphate, and glutathione were measured in cooked samples, indicating a potentially greater umami taste in the product (Toldrá and Flores, 2010). It is possible that the cooking further degraded ADP into the adenosine compounds and promoted the flavor through direct production of the adenosine compounds. Additionally, the degradation of ATP also leads to the liberation of ribose sugar that could further react with Maillard reaction products to generate a desirable meat aroma (Moloney and McGee, 2023). The greater abundance of meat flavor-related compounds such as cysteinylglycine and methionine were also observed in the cooked samples. Both cysteine and methionine have been known to participate in the Maillard reaction with ribose sugar to produce volatile compounds often related to meat flavor (Ramalingam et al., 2019).

Figure 5.

Principal component analysis (PCA) and hierarchical clustering analysis (HCA) of metabolites from raw and cooked beef patties made with different formulations. Different formulation treatments: CON (80% fresh beef + 20% fresh fat), DA-FAT (80% fresh beef + 20% DA fat), WA (50% fresh beef + 30% WA lean + 20% WA fat), DA (50% fresh beef + 30% DA lean + 20% DA fat), DWA (50% fresh beef + 30% DWA lean + 20% DWA fat), and UDA (50% fresh beef + 30% UDA lean + 20% UDA fat).

Metabolomic profiling also indicated that the inclusion of lean trim resulted in different metabolomics profiles in the patties and potentially greater flavor alteration (Figure 6). The PCA revealed a clear separation between CON, DA-FAT, and DA in raw samples, exhibiting a different metabolite composition across the samples. However, CON samples and DA-FAT samples clustered together following the cooking process compared to DA samples. These showed that while initially different, the CON and DA-FAT samples were more likely to undergo similar reactions during the cooking and generated a more identical product compared to DA samples. Similarly, HCA results also indicated that CON had more similarities to DA-FAT when compared to DA samples. A total of 8 metabolites were identified through MS/MS match and found to be significantly influenced by the lean and fat formulation (FDR < 0.01, Table S4). Seven of the metabolites were predominantly greater in DA samples and could be grouped into amino acids (L-5-oxoproline and proline), nucleotides (isonicotinic acid and naphthalenyl pyrimidinylamine), and lipids (isopropenyl napthalenone, hydroxy quinazolinone, and propionyl carnitine). The DA-FAT samples were found to have a greater abundance of tyrosine, and none was observed to be present in greater abundance in CON samples.

Figure 6.

Principal component analysis (PCA) and hierarchical clustering analysis (HCA) of metabolites from CON, DA, and DA-FAT raw and cooked beef patties. Different formulation treatments: CON (80% fresh beef + 20% fresh fat), DA-FAT (80% fresh beef + 20% DA fat), and DA (50% fresh beef + 30% DA lean + 20% DA fat).

This observation indicates that the lean trim portions might have a greater responsibility in driving the generation of flavor compared to the fat trim portion alone. It is possible that the greater abundance of compounds with flavor potentials in DA samples, such as isonicotinic acid, L-5-oxoproline, and proline, promoted more flavor compound production through the Maillard reaction in the beef patties, leading to greater flavor improvement in the samples. This finding is in line with the observed trained sensory panel results, showing lower bloody and metal flavor in the DA patties compared to CON and DA-FAT patties. It is possible that while limited, the flavor generated from these compounds is enough to mask the bloody and metal flavor perceived from the samples, thus potentially improving the product acceptability. Similar findings were previously reported by Jiang and Bratcher (2016) and Hicks et al. (2023) in which those authors found that different lean sources impacted the flavor precursor profile of the final ground beef patties, thus influencing the final product flavor.