Volatile Compound Profiles of Raw and Roasted Peanut Seeds of the Runner and Virginia Market-types

The unique flavor of peanuts that develops during roasting is the primary driving force for the consumption of peanut products. Although rarely consumed raw, the raw state of the peanut contains the precursors involved in the transformations that lead to the distinct flavor development in roasted peanuts. Volatile compounds extracted from the headspace above raw and roasted peanut samples of the runner and virginia market types by solid phase microextraction were characterized using two-dimensional gas chromatography coupled with time-of-flight mass spectrometry. The roasting treatment and peanut market-type each had a significant impact on the types and concentrations of small molecular weight compounds found. Among 361 sample components detected, 290 compounds were found to be significantly different between the raw and roasted treatments (p < 0.05). The roasted samples contained pyrazines, pyrroles, thiazoles


Introduction
The unique roasted flavor of peanuts is the basis of most consumer purchases of products containing them (Buckholz, 1981;Sanders et al., 1989). It is well established that the composition and concentration of volatile compounds produced during thermal processing are largely responsible for the characteristic flavor and aroma of roasted peanuts (Coleman et al., 1994;Sanders et al., 1997;Neta et al., 2010). During roasting, the precursor compounds present in the raw peanut participate in reactions that produce the volatiles known to impact the perception of roast peanut flavor (Hodge, 1953;Coleman et al., 1994;Ku et al., 1998). The predominant pathways for formation of volatile compounds in roasted peanuts are through the Maillard reaction, lipid oxidation, and caramelization (Coleman et al., 1994). Heterocyclic nitrogen-containing compounds produced by the Maillard reaction are recognized to be the prime contributors to peanut flavor, specifically pyrazines Newell et al., 1967;Walradt et al., 1971;Buckholz et al., 1980;Baker et al., 2003). However, the determination of the specific compounds responsible for roasted peanut flavor has been difficult. Analysis of volatile compounds from roasted peanuts to date typically results in several hundreds of compounds identified and complex chromatograms (Brown et al., 1968;Walradt et al., 1971;Ho et al., 1981;Oupadissakoon & Young, 1984;Braddock et al., 1995;Warner et al. 1996;Baker, 2003;Schirack et al., 2006;Chetschik et al., 2008;Greene et al., 2008;Neta et al., 2010). In the last 50 years, more than 300 volatile compounds have been associated with roasted peanuts (Chetschik et al., 2008;Neta et al., 2010;Zhang et al., 2023). However, no single class of volatile compounds, including pyrazines, independently forms the flavor of roasted peanuts (Schirack et al., 2006). Some other volatile compound classes have been found essential for roasted flavors, include aldehydes, ketones, alcohols, certain hydrocarbons, and phenolic and furan derivatives (Manzano et al., 2013).
were blanched with a convection oven (Despatch, Minneapolis, MN, USA) heated to 92 °C for one hour, followed by forced air cooling and physical removal of the skins using a model EX whole nut blancher (Ashton Food Machinery, Newark, NJ, USA). After blanching, the samples were stored as 0.45 kg aliquots in vacuum-sealed mylar bags at -80 °C until analysis (Klevorn & Dean, 2018).
For the roasted treatment, the 2.27 kg samples of runner (n=15) and virginia-type (n=15) peanuts were dry roasted to a Hunter L-value = 48 ± 1. Colors were verified using a Hunter Model Colorimeter (Hunter Labs, Reston, VA, USA). The samples were roasted as previously described by Poirier et al. (2014) using an Aerolab T-8 lab scale batch roaster (Buhler Aeroglide, Cary, NC, USA). Samples were roasted in removable square product trays that were uniformly perforated, with dimensions of 20 cm x 20 cm, and a depth of 7.62 cm. The temperature of the roaster was set to 177 °C and had an air flow rate of 1 m/s. To simulate industry roasting parameters, the airflow switched from up-flow to down-flow halfway through the roast. Immediately after roasting, peanuts were cooled to ambient temperature (~25 °C) using forced air. The skins were manually removed during the cooling. When the peanuts were cool, the samples were stored in vacuum-sealed mylar bags as 0.45 kg aliquots in a -80°C freezer until analysis (Klevorn & Dean, 2018).
Peanuts were removed from the -80 °C freezer immediately before pasting. Each individual 453-gram sample of frozen peanuts was processed into a paste with a Blixer-3 food processor (Robot Coupe, Jackson, MS, USA). Pasted peanut samples were weighed (0.075 g) directly into 10 mL clear screw cap vials (Microliter Analytical Supplies Inc. Suwanee, GA, USA). Next, 0.6 grams of NaCl (Sigma Chemical Corp., St. Louis, MO, USA) was added to each vial to "salt out" volatile components from the samples. Internal standards, pyrazine-d4 and pyrrole-d5, were then added to each vial at a concentration of 100 µg/L. Lastly, 933 mL of HPLC grade water was added to each vial to give it a final sample volume equal to 1.5 mL. Capped sample vials were vortexed for 1 min to ensure adequate mixing. NIST samples were acquired already pasted; and otherwise prepared with the same procedure. Blank samples without addition of peanut paste were prepared the same as the samples. An n-alkane series (Sigma Chemical Corp.) of alkane standards, octane through icosane (C8-C20) and henicosane through tetracontane (C21-C40), was used for retention index calculations (Van den Dool & Kratz, 1963). The alkane standards were prepared per run by pipetting 3 µL of a mix of alkanes: henicosane through tretacontane (C21-C40) into one clear screw cap vial, and then 1 µL of a mix of alkanes: octane through icosane (C8-C20) into another vial.

Sampling Design
Samples were grouped by commercial lots (1 lot for each market type and warehouse per batch for analysis) and randomized for analysis order with an online randomizer (random.com). Samples were placed in the temperature-controlled sample tray at 2 °C. Five batches of 33 samples were utilized to complete this study. One raw and one roasted peanut seed paste from each warehouse was tested in every batch (n=6 raw, n=6 roasted) along with the NIST quality control samples (n=3), alkane standards (n=2), and blank samples between runs to minimize carryover (n=16). In each run, the alkane standards were sampled first and last, and the NIST QC were sampled 9th, 19th, and 27 th .

Solid-Phase Microextraction (SPME) of Volatile Components
Peanuts volatile compounds were captured using headspace solid-phase microextraction (HS-SPME) and a CTC Analytics combiPAL autosampler (CTC Analytics, Zwingen, Switzerland). The volatile compounds were sampled using a 1 cm SPME fiber with three-phases divinylbenzene/carbonex/polydimethylsiloxane (DVB/CAR/PDMS) (Supelco Corp., Bellefonte, PA, USA) and a coating thickness of 50/30 µm. The samples were equilibrated at 40 °C for 15 minutes with agitation at 500 rpm. The SPME fibers were incorporated through the vial septa at a depth of 12 mm and exposed to the headspace above the sample. The SPME fibers then equilibrated with the sample headspace for 40 minutes at 40 °C with agitation at 100 rpm to extract the sample volatiles. The fiber desorbed into the GCxGC-TOFMS instrument for 15 minutes (Neta, 2010).

Comprehensive Two-Dimensional Gas Chromatography-Time-of-Flight Mass Spectrometry (GCxGC-ToFMS) Analysis
The peanut volatile compounds were profiled using a LECO Pegasus III two-dimensional gas chromatograph (GCxGC) coupled with time-of-flight mass spectrometer (ToFMS) (Model #614-100-700, Leco Corporation, St. Joseph, MI, USA). The instrument was connected to an Agilent GC (Model# 6890 N, Agilent Technologies; Santa Clara, CA, USA) fitted with a secondary oven. The system was assisted by a thermal modulator that was cooled with liquid nitrogen at a modulation time set to 1.75 sec and a hot jet pulse time of 0.35 sec. The cool time between stages was 0.53 sec. Separation of components was conducted using a polyethylene glycol column (SolGel-Wax TM , 30 m x 0.25 mm ID x 0.25 µm df)(SGE, Austin, TX, USA) as the first dimension column, and a 14% cyanopropylphenyl -86% dimethyl polysiloxane column (RTX 17-01, 1 m x 0.1 mm ID x 0.1 µm df) (Restek, Bellefonte, PA, USA as the second dimension column. Helium was utilized as the carrier gas at a constant flow rate of 1.3 mL/ min. The transfer line was set as 250 °C and operated in pulsed splitless mode with a pulse pressure of 37 psi for 1 min, and the split vent was opened 2 min after injection. The primary oven temperature was set to 40 °C, and programmed to increase to 140 °C at a rate of 5 °C/min, then 10 °C/min to 250 °C with an initial hold of 2 min, and a final hold of 3 min. The secondary oven temperature increased from 55 °C to 155 °C at 5 °C/min then 10 °C/min to 250 °C with an initial hold of 2 min, and a final hold of 4 min. The ToFMS detector was operated at -70 eV and an ion source temperature of 200 °C. Masses within the range of 25-500 m/z were collected. The detector voltage was 1500 V, with a scan rate of 200 spectra/sec (Neta, 2010).

GCxGC-ToFMS Data Processing
ChromaTOF® software version 4.33 (Leco Corporation, St. Joseph, MI, USA) data processing methods were employed to detect and quantify peaks established on unique masses as set by the deconvolution algorithm. Prior to processing the data, the retention times of the alkane standard peaks were identified to create a retention index table. The NIST/EPA/NIH Mass Spectral Library (National Institute of Standards and Technology (NIST), Gaithersburg, MD, USA, 2005) was utilized for tentative identification of deconvoluted chromatographic peaks for the compounds. Chemical names were attached to peaks that had at least a mass spectral similarity ≥ 750, where 1000 is an exact match. The peak area calculations utilized the unique mass (U) for every peak, as assigned by the ChromaTOF® deconvolution algorithm. In StatCompare® in ChromaTOF®, samples were assigned to their corresponding class including blanks, samples, and quality control (QC) samples. Compounds across all samples were then aligned using the software algorithm and mass spectral similarity match ≥ 600. Tentative identification of volatile compounds in peanuts was performed by comparison of volatile compounds mass spectra and retention times to those reported in the literature, utilizing the National Institute of Standards and Technology (NIST) Chemistry Webbook (https://webbook.nist.gov/chemistry/) (last accessed April 20, 2023). Confirmed identifications were made for those components that matched an authenic standard run on the same instrument under the same conditions.

Statistical Analysis
The aligned chromatographic data was exported to Microsoft Excel 2016 (Microsoft Corporation, Redmond, WA, USA) for data compilation. A new column that combined peak number and peak name was incorporated to give each peak a unique indentifier for statistical analysis. Missing value replacement was performed in Excel with the "randbetween" function to provide substitution data that reflected possible responses below the detection limit of the method for undetected components (Johanningsmeier & McFeeters, 2011). Analysis of variance (ANOVA) and hierarchical cluster analysis (HCA) were performed in JMP Genomics version 9.4 (SAS Institute, Cary NC, USA). To normalize the peak area variances prior to statistical analysis, a log2 transformation was utilized. An ANOVA was run on the log2 peak areas to detect differences in the compounds among the treatments. To check for instrumental drift, batch variation, and sample variation, the responses in quality control samples were observed for similar resolution of chromatographic peaks. Additionally, to ensure the variation in sample responses was due to the roast level and market-type, the weighted average proportion of variance across principal components was observed among the sample variables, the run order, the batch, and the residual variation. Metabolite peaks with significant differences (p < 0.05) among treatment groups were incorporated into hierarchical clustering analysis (HCA) utilizing the Fast Ward method as the clustering process (Johanningsmeier & McFeeters, 2011). The results were displayed in heat maps and were examined for relevant trends. Clusters of metabolites that were present in select treatment groups were selected for further investigation (Johanningsmeier & McFeeters, 2011). Components that existed only in roasted peanut samples were presumed to have been formed due to chemical changes that occurred during the roasting process. Further statistical analysis was performed using XLStat version 9.4 for Windows TM (Addinsoft, Paris, France). Principal component analysis (PCA) was employed as a non-supervised data reduction technique to visualize the relationships in the overall VOC profiles among roast treatments and market-types.

Identification and Quantification of Selected Volatile Compounds in Peanuts
Approximately 515 peaks with a signal to noise ratio (S/N) ≥ 20 were detected among the peanut samples. Manual inspection of the chromatograms and peak table data for chromatograms for the peanut samples compared to blanks containing sodium chloride in water established that 361 of the volatile compounds were assigned to the peanut samples. The 154 artifact peaks included system contaminants such as formaldehyde, were attributed to column bleed at the high end of the temperature program. The polar-semipolar two-column combination enables isolation of these artifacts from sample volatile components. Unlike one-dimensional GC, this makes it possible to identify and quantify low-level volatile metabolites despite system contaminants (O"Hagan et al., 2007;Johanningsmeier & McFeeters., 2011).
Of the 361 compounds detected in the peanut samples, 274 (75.9%) were tentatively identified by ChromaTOF® data processing with the use of the best spectral match to the NIST library with similarity ≥ 750. The 87 other compounds were unidentified. Comparisons to retention indices reported in the literature resulted in the presumptive identification of 155 volatile compounds in roasted and raw virginia and runner-type peanuts ( Table  1). Published retention indices were not found or did not match for the other 119 compounds. Authentic standards of 57 compounds were individually chromatographed to confirm the quality of the tentative identifications, and 54 were found to be positively identified by comparisons of the volatile compound"s retention index and mass spectra. Metabolites that did not match literature values were further investigated, and either concluded to be left as a mass spectral match or an unknown. Compounds matched to multiple peaks were visually investigated using the chromatographs and mass spectra to further determine the identity. The volatile compounds identified in the raw and roasted virginia and runner-type peanuts included 74 nitrogen containing compounds of which there were 28 pyrazines, 23 furans, six pyridines, and 17 pyrroles. In addition, 19 ketones, six diketones, seven esters, three ethers, 17 sulfur containing compounds, 27 alcohols, 25 aldehydes, and 40 hydrocarbons were identified. The roasted peanut samples contained a higher number of peaks and peaks in greater abundances than raw peanuts in both the runner-type and virginia-type peanut samples, as visualized in the chromatograms in Figures  1A and 1B. Between the raw and roasted peanut samples, a total of 252 volatile compounds were found to be significantly different (p < 0.05). Of these, 96 compounds were significantly different (p < 0.05) between the two market-types. For further investigation of trends in the data, hierarchical cluster analysis and principal component analysis were utilized.
This study reported 119 volatile compounds that have not previously been reported in peanuts, including 11 furans, seven pyrroles, five pyridines, and 12 pyrazines. Additionally, more volatile compound differences were identified between runner and virginia market types than previously reported where no differences (p < 0.05) between virginia and runner-type peanuts in their concentrations of total alcohols, aldehydes, alkanes, pyrroles, ketones, pyrazines, and furan derivatives were found (Wang et al., 2017). Figure 1A. GCxGC-ToF-MS chromatogram of roasted (orange) and raw (blue) runner-type peanuts Figure 1B. GCxGC-ToF-MS chromatogram of roasted (orange) and raw (blue) virginia-type peanuts

Principal Component Analysis
Principal component analysis (PCA) was performed to evaluate the grouping of the peanut samples based on the overall volatile compound composition (Figure 2). Principal components (PC) 1 and 2 accounted for a total of 71.6% of the variance in the VOC profiles, with PC1 covering 63.7% and PC2 accounting for 7.2%. The roasted peanut samples loaded positively on PC1, and raw peanut samples loaded negatively on PC1. Several alcohols, such as 1-nonanol, 1-propanol, 2-methyl-1-propanol, (S)-2-heptanol, 3-methyl-4-penten-1-ol, ethanol, (Z)-2-penten-1-ol, and 1-hexanol loaded negatively on PC1 and were strongly associated with the raw samples (data not shown). Numerous aldehydes, such as hexanal, were also correlated with the raw peanuts. Alcohols have been reported to serve as precursors for lipoxygenase-mediated reactions in raw peanuts (Singleton et al., 1976).

Hierarchical Cluster Analysis
Hierarchical cluster analysis (HCA) was employed to further visualize the 361 VOCs that were significantly different among treatment groups (p < 0.05) The resulting heat map was evaluated for trends, and six distinct clusters were evident (Figure 3). The groups represented compounds present in greater concentrations due to the raw or roasted state and/or market-type. From top to bottom, the six groups represent components present in significantly greater concentration in 1(dark blue): roasted virginia and runner, 2(orange): roasted virginia, 3(teal): roasted runner, 4(tan): raw and roasted virginia, 5(green): raw and roasted runner, and 6(light blue): raw virginia and runner. The majority of differentiating volatile compounds clustered together in Group 1 (Figure 3) and represented the volatile compounds generated in roasted peanuts from both market types. These 252 compounds included a wide variety of compound classes, comprised of 21 aldehydes, 28 pyrazines, 18 furans, 18 pyrroles, 20 sulfur-containing compounds, along with 61 unknown analytes. Group 2 shows 14 compounds that are predominantly present in roasted virginia peanuts. These compounds included hydrocarbons, aldehydes, a furan, and an oxazole. The 19 compounds in Group 3 were differentiated based on their more prevalent concentration in roasted runner peanuts. These compounds included hydrocarbons, aldehydes, two furans, a ketone, and a sulfonyl compound. The compounds in Group 2 and Group 3 were also the compounds that loaded in PC2 as discussed in the previous section. In Group 4, nine compounds differentiated the raw and roasted runner peanut samples, including six hydrocarbons, one alcohol, and two unknowns., where as Group 5 is comprised of 29 compounds that were dominant in the raw and roasted virginia-type peanut samples. Several nitrogen-containing compounds are present in this group, along with hydrocarbons, alcohols, aldehydes, and 13 unknowns. Finally, Group 6 contains the volatile compounds that were in higher abundance in the raw peanuts from both market types, represented by 21 compounds, including 12 alcohols, hexanal and 2-hexenal, along with n-caproic acid vinyl ester, pentanoic acid, 1-nitrohexane, and nitromethane.
The majority of differentiating volatile compounds clustered together in Group 1 (Figure 3) and represented the volatile compounds generated in roasted peanuts from both market types in the HCA. These 252 compounds included a wide variety of compound classes, comprised of 21 aldehydes, 28 pyrazines, 18 furans, 18 pyrroles, 20 sulfur-containing compounds, along with 61 unknown analytes. Group 2 was comprised of 14 compounds

Pathways to the Formation of Volatile Compounds
Among the hundreds of VOCs detected in this study, a variety of components from various compound classes were identified. In peanuts, aroma active compounds are generated through chemical reactions induced by heat treatment, including the Maillard reaction, Strecker degradation, thermal degradation of sugars, and lipid oxidation Lykomitros et al., 2016b).
Caramelization, also known as the thermal degradation of sugars, yields low molecular weight open-chain oxygen containing products and heterocyclic oxygen-containing compounds including furan derivatives (Coleman et al., 1994). As expected, the roasted peanut samples contained significantly more (p < 0.05) furans than the raw samples.
Lipids are also an important source of flavor compounds. While they do contribute flavors of their own, their primary importance is as precursors to volatile compounds that produce flavors in foods (Forss, 1969;Pattee et al., 1983). Lipids with more than ten carbons are insoluble in water, have low volatility, and do not participate in basic taste (Pattee et al., 1983). Foods with polyunsaturated fatty acids are known to be highly susceptible to lipid oxidation, which leads to the formation of oxygen containing compounds including aliphatic aldehydes, acids, ketones, and alcohols (Coleman et al., 1994;St. Angelo et al., 1996;Warner et al., 1996). Each of these compound classes was detected in the virginia and runner type peanut samples. Oil composes up to 50% of a peanut seed, of which about 50% is oleic (18:1) and 30% is linoleic (18:2) in normal oleic peanuts (St. Angelo et al., 1996;Davis & Dean, 2016). Heat damage to the cell structure can augment the transfer of oxygen to peanut tissues by the release of substances from cell compartments. Due to this, lipid oxidation can potentially be accelerated by roasting (Perren & Escher, 2013). Long chain unsaturated fatty acids in peanuts have little participation in basic tastes but are easily oxidized. The presence of the double bonds enables free radicals to stabilize through the delocalizing of unpaired electrons. This leads to hydroperoxide formation, which is unstable and quickly decomposes into secondary reaction products (St. Angelo et al., 1996;Warner et al., 1996). Products of oxidation, including some alcohols, aldehydes, and furans were detected in greater abundance (p < 0.05) in the roasted peanut samples. Several of these pathways are discussed in section 4.2 below.
The conversion of alcohols, which were significantly more abundant in the raw peanut samples (p < 0.05), to corresponding aldehydes homologs, which were more abundant in the roasted peanut samples is likely related to enzymatic reactions (Singleton et al., 1976). Previously, the volatile flavor profiles of raw peanuts have been correlated to enzyme activity across stages of peanut seed maturation (Pattee et al., 1970). Singleton et al. (1976) observed n-propanol and n-hexanol were converted to their respective aldehydes when raw peanut extracts were treated with lipoxygenase. Pattee et al. (1970) found that the predominating volatile compounds in raw peanuts were: acetaldehyde, methanol, pentane, ethanol, and hexanal, which were also detected in this study with the exception of ethanol. Those authors speculated that these compounds, aside from hexanal, were produced via lipoxidase and alcohol dehydrogenase (ADH) in peanut seeds. Lovegren et al. (1982) found that methanol, acetaldehyde, ethanol, and an acetone group totaled approximately 80% of the total volatile peaks in raw virginia-type peanuts. The volatile compound data collected in this study includes those previously found and shows a much wider array of compounds that contribute to the volatile profiles of the raw peanut samples.
The runner peanut samples were most differentiated by compounds: 2,2,4-trimethylpentane, 1,3-dimethylbenzene, ethylbenzene, 3-methylpentanal, 3-methylbutanal, benzaldehyde, tetrahydrofuran, 2,3,6-trimethylpyridine, and 2,3-pentadedione in the HCA. Brown et al. (1972) found 3-methylbutanal to be a predominant compound distinguishing between raw and roasted runner-type peanuts. Those authors found that this branched chain aldehyde in large concentrations resulted in the harsh aroma of roasted peanuts. In other studies, 3-methylbutanal has been found to have a malty/chocolate aroma (Matsui et al., 1998;Greene et al., 2008). Low molecular weight aldehydes in general have also been reported to be responsible for a harsh aroma note in peanuts . They are formed during roasting as a product of Strecker degradations Smit et al. 2008). For example, leucine can form 3-methylbutanal through deamination followed by decarboxylation. Both leucine and 3-methylbutanal were more abundant in virginia-type peanuts (Klevorn et al., 2019). Aldehydes were the most common functional class of the compounds in Group 5, followed by alcohols. These compounds were detected in greater abundance in the raw and roasted virginia-type peanut samples, which may be more susceptible to lipid oxidation.

Lipid Oxidation Products
Volatile products of autoxidation of fatty acids are significant for the aroma of foods due to their low threshold concentrations for aroma and flavor (Schieberle & Grosch, 1981). The mechanism of autoxidation of linoleic acid (C 18:2) involves removal of a hydrogen atom from the methylene group adjacent to the double bond on carbon-11 producing a pentadienyl radical. Oxygen molecules then attack from both end positions to create an equal combination of conjugated 9-and 13-hydroperoxide isomers (Frankel, 1984). These conjugated isomers are typical, however, 10-and 12-hydroperoxide isomers also exist. These compounds can undergo additional reactions that lead to a number of compounds (Kolchar, 1996), several of which were found in the study reported here.
(E)-2-octenal was more abundant in raw (four-fold) and roasted (two-fold) virginia peanuts than raw and roasted runner-type peanuts. Roasted virginia-type peanuts also contained two-fold more 2,4-decadienal than runner-type peanuts.
While 2,4-decadienal can only be formed from 9-hydroperoxide isomers, hexanal can arise from 9-or 13-hydroperoxide isomers. Thus, hexanal is the most abundant aldehyde product of linoleic acid oxidation (Schieberle & Grosch 1981). This was observed in the volatile compound data, as hexanal had the highest relative abundance of the oxidation products for roasted virginia-type peanuts and the highest relative abundance of all compounds for raw virginia-type peanuts. Hexanal was four times more abundant in the raw treatments than the roasted samples, two times more abundant in raw virginia peanut samples than the raw runner-type samples, and three times more abundant in roasted virginia samples than roasted runner-type (p < 0.05).
Heptanal can be formed from 2-octenal by autoxidation into a radical acid intermediate. This peroxyacid then decomposes with carbon dioxide as a byproduct, which allows the enol to rearrange into heptanal (Schieberle and Grosch, 1981). Heptanal was found in greater abundance in the roasted peanut samples of both market-types (Group 1) than of the raw samples. The virginia-type peanut samples experienced a two-fold increase, and runner-type samples saw a five-fold increase after roasting.
The 12-hydroperoxide isomer has been detected in vegetable oils and can lead to the formation of aldehydes (Kochlar, 1996). The decomposition mechanism of linoleate 12-hydroperoxide into 2-heptenal involves an alpha scission. A vinyl radical reacts with oxygen to produce vinyl hydroperoxide, which then interacts with other lipid molecules to form 2-heptenal. Additional beta scission pathways modify the stuctures (Kochlar, 1996), which could have produced several of the unsaturated aldehydes found in this study, including, 2-methyl-2-hexenal, 2-ethyl-trans-2-butenal, and 2-butenal. 2-heptenal was significantly more abundant in the roasted virginia peanut samples (Group 2) than the other groups.
Another compound that significantly increased (five-fold) after roasting was 1-octen-3-one is (Group 1). This ketone has been found to be a primary contributor to metallic tastes in fatty foods (Forss, 1969;Pattee et al., 1983;Greene et al., 2008) and to contribute to a mushroom-like flavor in peanuts (Kaneko et al., 2013;Erten & Cadwallader, 2017;Greene et al., 2008). The metal flavor may be attributed to the presence of inorganic salts of copper and iron, which are responsible for the lipid oxidation due to catalyzing the break-down and formation of 1-octen-3-one (Forss, 1969). Roasted peanuts are a good dietary source of each of these microminerals (Davis & Dean 2016). The alcohol homolog, 1-octen-3-ol, was in significantly higher concentrations (p < 0.05) in the virginia-type peanuts. This unsaturated alcohol is a major product of autoxidation of linoleic acid, and has been commonly found in meat volatiles, with a mushroom-like odor (Bleicher et al., 2022). Autoxidation of arachidonic acid is another mechanism to produce 1-octen-3-one and 1-octen-3-ol. Linoleic acid has been noted as a precursor of arachidonic acid in peanut oil (Truswell et al., 1994).
Furans can also be derived from oxidized linoleic acid. The proposed mechanism for formation involves linoleate 9-hydroperoxide decomposing to form 2-pentylfuran. This compound was found most abundant in the roasted virginia peanut samples (Group 2), which was two-fold more abundant than in both the raw virginias and roasted runner-type peanut samples. Along with several furan derivatives, 2-pentylfuran is responsible for flavor defects in reverted soybean oil, including metallic and grassy flavors (Kochlar, 1996). At concentrations between 1-10 ppm in refined, bleached, and deodorized cottonseed oil, 2-pentylfuran emits a beany odor, although the effluent of 2-pentylfuran from gas chromatograph has a licorice, and not beany odor (Krishnamurthy et al., 1967). Furan compounds can also be derived from the thermal degradation of glucose (Zhang & Ho, 1991), and through the Maillard browning reaction (Mottram, 1993). Certain furans have been found to have caramel-like, sweet, roasty, burnt, fruity, and pungent aromas characteristic of thermally processed foods (van Boekel 2006;Liu et al., 2011;Kaneko et al., 2013). Food flavor development has been associated with increases in furan levels (Tai & Ho, 1998). Of the 23 furans reported in this study, 18 were more abundant in the roasted peanut samples (Group 1), indicating that the development of most of the compounds in this group was related to thermal processing. However, 2-ethylfuran and 2-vinylfuran were most abundant in the raw virginia-type samples.

Products of the Maillard Reaction
Groups 1, 2, and 3 contained compounds that were present in roasted peanut samples at levels significantly higher than the raw counterparts. Many compounds associated with thermal processing and browning flavors are heterocylic (Fennema, 1996). These compounds commonly include nitrogen, sulfur, and oxygen substituents, and have been known to contribute general nutty, roasted, toasted, caramel, meaty, burnt, floral, and plant odors (Fennema 1996). Figure 4 displays the chemical structure of several of such compounds that were detected in this study in roasted peanut samples including pyrazines, pyridines, thiophenes, furans, thiazoles, pyrroles, and oxazoles. Many of these heterocyclic compounds are formed via the Maillard browning reaction, which utilizes amino groups and reducing sugars (Koehler et al., 1969;Koehler et al., 1970), or lipid-derived carbonyls that can also react with amino acids (Zamora & Hidalgo, 2011). Pyrazines have been found to be important contributors to flavors in many foods (Mottram, 1994). The 28 pyrazines detected in this study may have formed through several mechanisms. One pathway for pyrazine formation utilizes a reaction between amino acids in the peanut and α-dicarbonyl compounds, which are intermediates in Maillard reactions. This reaction is carried out via the Strecker degradation, where α-amino carbonyls are produced and then condensed into alkylpyrazines (Fennema, 1996). Different α-amino carbonyls can participate in the Strecker degradation and produce a variety of pyrazines including 2,6-dimethylpyrazine, trimethylpyrazine, 2-ethyl-5-methylpyrazine, and methylpyrazine, which were all identified in the roasted peanuts in this study (Group 1). The alkyl group on the pyrazine is often obtained from α-amino carbonyl group of the reactant derived from sugar (Shibamoto & Bernhard, 1977). However, the distribution of reaction products in pyrazine formation can be influenced by reaction temperature, reactant ratio, presence of anitoxidants or prooxidants, and oxygen (Shibamoto & Bernhard, 1977). Each of these compounds had significant (p < 0.05) increases in relative abundance after roasting, reported as follows: trimethylpyrazine (three-fold), methylpyrazine (two-fold), 2-ethyl-5-methylpyrazine (two-fold), and 2,6-dimethylpyrazine (was not detected in the raw samples). While many α-amino carbonyl fragments have been proposed, no intermediates or fragments in pyrazine formation have been isolated or identified to date (Shibamoto & Bernhard, 1977).
Another mechanism of pyrazine formation involves ammonia as the nitrogen source, which is released from pyrolysis of amino acids (Mottram, 1994). Glutamine has been shown to yield considerable amounts of ammonia with moderate heating (110°C). At high temperatures (180°C), asparagine, which was more abundant (two-fold) in the virginia-type peanut samples, and aspartic acid released high amounts of ammonia. Free ammonia can react with α-hydroxycarbonyl compounds and form α-aminoketones in an Amadori rearrangement (Adams et al., 2008). Methional was found more abundant in the roasted virginia and runner-type peanut samples (Group 1). This compound is likely to have formed as a product of methionine through the Strecker degradation (Balance, 1961). Methionine was two-fold more abundant in the virginia-type peanut samples than in the runners (Klevorn et al., 2019). Methional has been associated with a baked potato and brothy odor and has been reported in roasted peanuts previously (Greene et al., 2008;Chetschik et al., 2008) and roasted almonds (Erten & Cadwallader, 2017).
Dimethyl trisulfide was also found to be significantly more abundant in the roasted virgina and runner-type peanut samples (Group 1). This compound may have been produced by reactions with hydrogen sulfide, or by oxidation (Yu & Ho, 1995). The reported aroma of dimethyl trisulfide is onion-y/garlic-y and sulfur/cabbage-y and has been found in other studies to be a key component in roasted peanuts (Chetschik et al., 2008;Neta et al., 2010), cauliflower, broccoli, and cabbage (Buttery et al., 1976), wine (Guth, 1997), almonds (Erten & Cadwallader, 2016), and in boiled meat aroma (Golovnja & Rothe, 1980).
Five thiazole compounds and one oxazole were found to have significantly greater concentrations in roasted peanuts than in the raw: 4-methylthiazole, thiazole, 2-methylthiazole, 4,5-dimethylthiazole, 4-methyl-2-(1-methylethyl)-thiazole, and trimethyloxazole. Thiazole is closely related to the structures of oxazoles, but is more abundant in food volatiles, especially fried or roasted foods (Mottram, 1994). These compounds can arise from the degradation of thiamine, which reacts with 1,2-carbonyls and aliphatic aldehydes derived from amino acids in the Strecker degradation (Mottram, 1994). Both market-type samples contained 2-methoxy-4-vinylphenol in their roasted forms. This compound was previously found to be unique in peanut oil, compared to other common vegetable oils (Hu et al., 2014) and has been associated with sweet/licorice (Greene et al., 2008) and spicy/phenolic (Matsui et al., 2008) aromas in peanuts. It can be formed from ferulic acid, which is an intermediate in the degradation of lignin polymers in plants (Steinke et al., 1964;Fiddler et al., 1967;Walradt et al., 1971). Lignins are important to the peanut plant for structural support of the tissues and for protection from pathogens and herbivores (Bennett et al., 2017).

Conclusion
Investigation of the volatile compound profiles between runner and virginia market-types of raw and roasted peanuts revealed a number of differences, including 119 compounds previously unreported in peanuts. Comprehensive two-dimensional gas chromatography with time-of-flight mass spectrometry (GCxGC-ToFMS) enabled detection of more components than previously possible, including the identification of 96 VOCs that differentiated the raw and roasted peanut samples. The roasted samples contained a greater abundance of volatile compounds than the raw samples, indicating the numerous changes in chemical composition formed during roasting. The roasted samples were abundant in nitrogen-containing compounds including pyrazines, furans, pyrroles, and pyridines. These compounds are products of the Maillard browning reaction, which are important for the development of the roasted peanut flavor, along with other characteristic attributes including aroma, color, and texture. The specific mechanism to achieve roasted peanut flavor is not yet known, however, the newly detected compounds provide a broader knowledge of small molecular weight volatile compounds that may contribute to aroma activity. The raw peanut samples contained numerous alcohols and products of lipid oxidation. Although there were fewer differences between the market-types in the raw form, oxidation products were detected more abundantly in virginia-type peanut samples. This can be attributed to the higher levels of polyunsaturated fatty acids in virginia-type peanuts compared to runners. Further investigation into the aroma activity of these compounds could serve to find which compounds influence the flavor of the roasted peanuts. This could enhance the ability to understand the compounds most important to achieving characteristic peanut flavors.