A Comprehensive Characterisation of Volatile and Fatty Acid Profiles of Legume Seeds

Legumes are rich in unsaturated fatty acids, which make them susceptible to (non) enzymatic oxidations leading to undesirable odour formation. This study aimed to characterise the volatile and fatty acid profiles of eleven types of legumes using headspace solid-phase microextraction gas chromatography–mass spectrometry (HS-SPME-GC-MS) and GC coupled with a flame ionisation detector (GC-FID), respectively. Volatile aldehydes, alcohols, ketones, esters, terpenes and hydrocarbons were the chemical groups identified across all the legumes. The lipids comprised palmitic, stearic, oleic, linoleic and α-linolenic acids, with unsaturated fatty acids comprising at least 66.1% to 85.3% of the total lipids for the legumes studied. Multivariate data analysis was used to compare volatile and fatty acid profiles between legumes, which allow discriminant compounds pertinent to specific legumes to be identified. Results showed that soybean, chickpea and lentil had distinct volatile and fatty acid profiles, with discriminating volatiles including lactone, ester and ketone, respectively. While all three Phaseolus cultivars shared similar volatile profiles, 3-methyl-1-butanol was found to be the only volatile differentiating them against the other eight legumes. Overall, this is the first time a multivariate data analysis has been used to characterise the volatile and fatty acid profiles across different legume seeds, while also identifying discriminating compounds specific for certain legume species. Such information can contribute to the creation of legume-based ingredients with specific volatile characteristics while reducing undesirable odours, or potentially inform relevant breeding programs.


Introduction
The seeds of legume plants (usually referred to as 'legumes') are a nutritious source of proteins, carbohydrates, lipids, vitamins and minerals [1]. Unprocessed legumes have a distinct odour due to their inherent plant metabolism [2]. Regrettably, legumes are not widely utilised to their full potential due to various factors, such as low protein digestibility, their hard-to-cook-effect and undesirable odours. The undesirable odours are largely influenced by lipoxygenase-catalysed unsaturated fatty acid oxidation that occurs during several postharvest processes [3,4]. Trained sensory panels have used negative descriptors such as beany, musty, haylike, grassy and green to describe the odours of legumes such as soybean and peas. It has been recognised that these odours are majorly derived from volatile compounds such as hexanal and 1-octen-3-ol [3][4][5]. Previous studies [3,[6][7][8][9] have reported volatile compounds and their formation mainly in soybean, and there is still a limited understanding on the volatile profile of other commercially relevant legumes such as cowpea, lentil, common bean and pea [10][11][12][13]. However, these studies used targeted analysis without providing a holistic picture of all to analysis, method parameters were optimised in order to capture a wide range of volatile compounds. The optimisation included sample weight, sample dilution, sample to salt ratio and type of GC column.
Upon sample analysis, legume flour was weighed (2.5 g) into a 20 mL glass vial, and 5 mL of saturated sodium chloride solution (360 g/L) was added to increase the solution's ionic strength and drive the legume volatiles into the headspace. The vial was then tightly sealed with PTFE-coated silicon septa screw cap (Supelco, Sigma-Aldrich, St. Louis, MO, USA). The sealed vials were then vortexed for 30 s.
For the GC-MS analysis (Agilent 6890N, Agilent Technologies, Santa Clara, CA, USA), extracted volatiles were desorbed in the injection port at 230 • C for 2 min, then injected in splitless mode onto a ZB-Wax capillary column (30 m × 0.25 mm × 0.25 µm; Agilent Technologies, Santa Clara, CA, USA) for separation with helium as the carrier gas at 1.5 mL/min. To facilitate separation and elution of the injected headspace volatile compounds, the GC oven was maintained at 50 • C for 5 min before the temperature was ramped up to 210 • C at 5 • C/min, after which it was again ramped to 240 • C at the rate of 10 • C/min, for a total GC-MS run time of 37 min. For the MS, the quadrupole was set at 70 eV, and the ion sources were 150 • C and 230 • C, respectively, with a mass-to-charge ratio scanning range of 30-300 m/z. Thereafter, the SPME fibre was regenerated according to the manufacturer's instruction. The same SPME fibre was used across all samples. The volatile profiling for each legume seed was conducted in five independent replicates.
With regard to preprocessing of GC-MS chromatograms, volatile fingerprinting chromatograms often contain co-eluting peaks, which can confound data analysis. Therefore, an automated mass deconvolution and identification system (AMDIS; version 2.72, build 140.24, Agilent Technologies, Santa Clara, CA, USA) was used to deconvolute potential co-eluting peaks. The spectra obtained were further processed by mass profiler professional (MPP; version 14.9.1, build 1316, Agilent Technologies, Santa Clara, CA, USA), a peak filtering and alignment software. This creates aligned peaks lacking nonreproducible and background peaks. Afterwards, a table of retention time and volatile amount expressed as peak area was obtained.
Tentative identification of volatile compounds was conducted manually. In the present work, three criteria were employed to increase the power of compound identification: (i) match and reverse match with the NIST library of no less than 90%; (ii) comparison of experimental retention index with RI according to literature; and (iii) matching retention time and spectra with authentic standards from chemical groups of detected volatiles (alcohol, aldehyde, terpene and acid) (See Table S1).

Determination of Fatty Acids in Legume Seeds Using FAME-GC-FID
For fatty acid analysis, legume lipid was extracted and converted to fatty acid methyl esters (FAME) and detected using gas chromatography flame ionisation detection (GC-FID) according to AOAC method 963.22 [17] with modifications.

Total Lipid Extraction Based on Soxhlet Method
Legume flour sample (2 g) was weighed and placed inside a cellulose extraction thimble (26 × 60 mm, Whatman, Buckinghamshire, UK). The filled extraction thimble was fitted onto a Soxtec distillation apparatus (Tecator, Hilleroed, Denmark). Meanwhile, aluminium cups were filled with five to ten antibumping granules and weighed before adding with 25 mL of organic solvent mix consisting of a 2:1 (v/v) ratio of chloroform (EMPARTA, Merck, Darmstadt, Germany) and methanol (Ajax Univar, North Shore, New Zealand). The cups were then fitted underneath the thimble and above the heating plate of the Soxtec distillation apparatus. Continuous reflux distillation of the samples was then carried out for 1 h with the heating plate set at 160 • C. After that, solvent was evaporated from the sample, and then the cups were released from the apparatus, and the residual solvent was allowed to evaporate inside a convection oven (Sanyo MOV-212F, New South Wales, Australia) set at 50 • C for 15 min.
The lipid yield of each sample was estimated by weight difference between the weight of the empty cups (filled with antibumping granules) and the weight of the cup filled with lipids upon completion of the solvent extraction. After weighing, 15 mL of hexane (Ajax Finechem, Auckland, New Zealand) was used to resuspend the lipids in each aluminium cup. The hexane containing lipid solution was then stored in a refrigerator at 4 • C in a tightly sealed glass tube before proceeding to the lipid purification step. The lipid extraction of each legume seed was conducted in four independent replicates.

Lipid Purification
Fatty acid methyl esters (FAMEs) were obtained by purifying lipids by saponification to remove nonsaponifiable materials, followed by esterification into FAMEs. A volume of lipid containing hexane solution equivalent to 5 mg lipid was pipetted into a sealable glass tube. Thereafter, 5 mL of a solution containing potassium hydroxide (0.5 M, AnalaR, Leuven, Belgium) dissolved in methanol was added, and the tube was sealed immediately. The fatty acid saponification was carried out for 20 min at 80 • C on a heating block. The tubes were removed from the heating block and allowed to cool in ambient air for 10 min.
To the cooled solution, 3 mL of diethyl ether (LabServ, Auckland, New Zealand) and 7 mL of milliQ water were added, and the test tube was inverted to mix. The mixture was allowed to stand for 2 min to allow separation between water and organic solvent layers. The top layer of diethyl ether was then discarded to remove any nonsaponifiable material. The fatty acids were liberated by neutralisation to~pH 7 with concentrated hydrochloric acid (37%; EMSURE, Merck, Darmstadt, Germany). Then, another 4 mL of diethyl ether was added and inverted to mix. The top diethyl ether layer formed was collected in a clean glass test tube for derivatisation.

Lipid Derivatisation to Fatty Acid Methyl Esters (FAMEs)
One millilitre of boron trifluoride (14%) in methanol (Sigma-Aldrich, St. Louis, MO, USA) was promptly added as derivatisation agent. Fatty acid esterification was carried out for 20 min at 80 • C on a heating block and then cooled. After cooling, 7 mL of saturated sodium chloride solution (360 g/L) was added and vortexed for 15 s. The top diethyl ether formed was then collected for FAME GC-FID analysis.
Samples (1 µL) were injected in split mode (20:1 ratio) at 240 • C for separation with hydrogen gas at 2.2 mL/min. To ensure good separation of fatty acid methyl esters, the GC oven temperature was increased from its initial temperature of 120 • C to 225 • C at the rate of 3 • C/min, then ramped to 245 • C at 10 • C/min. Once the GC oven temperature reached 245 • C, the column was held at this temperature for another 2 min. For the FID, the detector temperature was set at 250 • C.

Identification and Data Preprocessing of FAME
Chromatograms obtained from GC-FID were analysed with GC ChemStation (Build 4.01, Agilent Technologies, Santa Clara, CA, USA) and individual peaks manually identified by matching retention time with commercial standards (FAMQ-005, AccuStandards, New Haven, CT, USA). Following manual peak alignment and removal of interfering background compounds, the proportion of signal abundance of each fatty acid was calculated in % abundance of total signal abundance. Thereafter, a table of fatty acid profile for each legume was obtained.

Multivariate Data Analysis and Identification of Compounds Relevant to Specific Legume Type
Multivariate data analysis, marker selection and marker identification were performed on the combined data sets comprising legume fatty acid and headspace volatile data sets.
Using both volatile and fatty acid data, multivariate data analysis was conducted using principle component analysis (PCA), followed by partial least square discriminant analysis (PLS-DA), utilising Solo software (Version 8.2.1, Eigenvector Research, Manson, WA, USA). Firstly, PCA was used as an unsupervised, exploratory technique to determine grouping/separation in the data, as well as to detect outlier. Secondly, PLS-DA was used as a supervised technique to detect similarities and differences between different legume seeds, as well as correlation between volatile compounds and legumes. Thereafter, a bi-plot was generated as a visual representation of the information obtained (OriginPro, OriginLab, Northampton, MA, USA).
Volatile compounds that showed a clear discriminant correlation with each legume were selected through determination of variable identification (VID) coefficients [18]. VID values are the corresponding correlation coefficients between X-variables (volatile compounds and fatty acids) and predicted Y-variables (Legume type). An absolute threshold value of |0.800| was selected. Therefore, volatiles with an absolute VID coefficient higher than 0.800 were plotted as bar graphs, and statistical significance between the means was determined using analysis of variance, conducted through SPSS Statistics (IBM, Version 26), followed by Tukey's post-hoc test (p < 0.05). Those compounds were considered important discriminant compounds associated with each legume seed.
Discriminant volatile compounds were identified by comparing the deconvoluted mass spectra with an established mass spectra library using National Institute of Standards and Technology (NIST) Mass Spectral Search Program (Version 2.2, build June 10, 2014). The identities were also rechecked with a minimal 90% match and reverse match on NIST, as well as comparison of retention index with literature.

Fatty Acid Analysis of Legume Seeds
GC-FID analysis of the fatty acid methyl esters was able to detect five clearly separated peaks in eleven legume samples. The legume lipid fractions consisted of palmitic (C16:0), stearic (C18:0), oleic (C18:1), linoleic (C18:2) and α-linolenic (C18:3) acids (Table 1). A commonality of the lipid profiles is that the level of saturated fatty acids such as palmitic and stearic acids is low, with the ratio of saturated to unsaturated fatty acids at 1:2 in cowpea and mung bean, and up to 1:5.4 and 1:6 in soybean and chickpea. On the other hand, all legumes had a high level of essential polyunsaturated fatty acids, namely, linoleic and α-linolenic, ranging from 47.3% in orange lentil to 71.0% in black bean. Out of the eleven types of legumes, four of them, i.e., chickpea, orange lentil, pea and fava bean, had a high (>20%) proportion of oleic acid, a monounsaturated fatty acid. This affirms that legumes are a good source of unsaturated fatty acids [19], with their fatty acid profile favourable from a cardioprotective perspective [20]. Table 1. Relative fatty acid abundance of 11 types of legume seeds, as analysed by fatty acid methyl ester gas chromatography coupled with a flame ionisation detector (FAME-GC-FID). Values expressed as mean ± standard deviation (n = 4). C16:0 = palmitic acid; C18:0 = stearic acid; C18:1 = oleic acid; C18:2 = linoleic acid; C18:3 = α-linolenic acid; SFAs = saturated fatty acids; MUFAs = monounsaturated fatty acids; PUFAs = polyunsaturated fatty acid; n-6/n-3 = ratio of omega-6 to omega-3 fatty acids. Means with different superscripts in the same column indicate significant difference (p < 0.05).

Volatile Analysis of Legume Seeds
The HS-SPME-GC-MS fingerprinting method was able to detect an increased number of volatile compounds, totalling 97 different volatiles across all 11 legumes. Visually, the total ion chromatograms appear to be different based on the number and intensity of the peaks present amongst the samples. Some representative total ion chromatograms of the samples are shown in Figure 1.
The chemical classes of detected volatile compounds consisted of alcohols, aldehydes, ketones, esters, lactones, terpenes, hydrocarbons, furans, pyrroles and sulphur-containing compounds. Note that the percentage of specific compounds mentioned in this section refers to their relative abundance, not absolute concentration. It is also important to note that soybean and chickpea had volatiles with the highest total peak area compared to other legumes.

Comparison of the Volatile and Fatty Compositions Among the Eleven Legumes and Identifying Discriminant Compounds
Multivariate data analysis (MVDA), which is an advanced chemometrics technique, was used to compare the volatile and fatty acid profiles among the 11 legume samples and identify discriminating compounds. In order to investigate the interdependence and relation among the measured attributes, the volatile and fatty acid data were merged into a single data matrix and analysed with MVDA. A principle component analysis (PCA) was first used as an unsupervised exploratory technique to detect groupings, separations or outliers within the volatile and sample data. From the PCA modelling (results not shown), it was able to be determined that there is indeed some distinct grouping and separation within the samples and that there were no outliers. Thereafter, a partial least squares discriminant analysis (PLS-DA) model was constructed using the volatile and fatty acid profiles as X-variables and the 11 types of legumes as categorical Y-variables. A bi-plot constructed using the first two latent variables (LVs) is shown in Figure 2. On the bi-plot, samples that are close to each other are considered similar, whereas samples that are further apart are considered different [18]. Figure 2 clearly shows that soybean and chickpea are projected further away from other samples in their own quadrant, indicating a large difference compared to the other legumes. The third quadrant is shared by cowpea and lentil, again indicating differences from other legumes. The fourth quadrant is occupied by pea, mung bean, fava bean, adzuki bean and all three Phaseolus beans, indicating similarity between the samples, especially between kidney, navy and black beans. This similarity may be attributed to them belonging to the same species. In addition to the legume samples, unfilled circles on the bi-plot represent volatile and fatty acid compounds (X-variables). The location of each circle represents its relation to other measured attributes (X-variables) or samples (Y-variables). Hence, a PLS-DA bi-plot provides a graphical representation of the relation between measured attributes and legume types. To gain further understanding into the specific volatiles or fatty acids which are clearly different between legume samples, variable selection was performed using a VID technique. The selected discriminant compounds are listed in Table 2. To illustrate the differences amongst the seeds, some representative discriminant volatiles and fatty acids are also visually presented in Figure 3, with significant difference (p < 0.05) determined using analysis of variance and Tukey's post-hoc test. Key points are discussed in Section 4.

Discussion
The information gathered in this study has shown some key trends emerging from data analysis. These subsections discuss them in relation to the existing literature and known reaction pathways.

Ratio of Unsaturated Fatty Acids in Studied Legumes
The Phaseolus group (navy, kidney and black bean) has a low ratio (<1) of omega-6 (n-6; linoleic acid) to omega-3 (n-3; α-linoleic acid) fatty acids. Adzuki bean, mung bean and cowpea also have relatively low n-6/n-3 ratio. This is in contrast to chickpea and fava bean, which have a very high n-6/n-3 ratio of more than 18 and 14, respectively (Table 1). Diets where the ratio of fat consumed has a high n-6/n-3 ratio (≥10:1) have been linked to risk of developing chronic noncommunicable diseases, such as autoimmune and inflammatory disease, cardiovascular disease and cancer, whereas a ratio of less than 4:1 generally indicates better health outcomes [21]. Therefore, legumes, excluding chickpea and fava bean, are a good source of essential fatty acid with a desirable n-6/n-3 ratio.
However, despite their health benefits, unsaturated fatty acids are known to be more susceptible to oxidation than saturated fatty acids, due to the presence of one or more double bonds. The extent of lipid oxidation in foods can be reliably measured by determining the peroxide value so that preventive measures can be taken to delay oxidation in legume seeds after harvest while preventing the production of undesirable rancid flavour in legumes [22]. Endogenous enzymes present in legumes such as lipoxygenase utilise unsaturated fatty acids as substrate, producing volatile compounds, some of which possess undesirable odours. This is especially a problem for legumes with a high proportion of unsaturated fatty acids (>80%), such as soybean and chickpea ( Table 1). The evolution and presence of volatile compounds are discussed further in the next section.

Aldehydes, Alcohols, Ketones and Terpenes in Studied Legumes
Aldehydes were the most abundant chemical class detected in all samples, except for navy, kidney and black beans; these beans seem to be rich in alcohols. Hexanal was the most abundant compound detected in all samples, except for navy bean. Nonanal and 2-hexenal were the second and third most abundant aldehyde in the legumes studied in this work. This is consistent with previous studies in soybean, winged bean [23] and three cultivars of common beans (black, pinto and dark red kidney) [10]. The difference in hexanal abundance may be attributed to the difference in the amount of linoleic and linolenic acid (Table 1), since they are the precursor for lipoxygenase-catalysed evolution of hexanal [3].
Alcohols comprised the second-most abundant chemical class of volatile compounds in the headspace of legume seeds, except for navy bean, where hexanol is the most abundant volatile compound. This is not surprising, as legumes contain alcohol dehydrogenases that act on products of the lipoxygenase pathway described above. For example, three isozymes of alcohol dehydrogenase have been described in chickpea, which catalyse the interconversion of aldehydes, alcohol and acid [30], possibly explaining the abundance of 1-hexanol and 1-pentanol in chickpea. Other alcohols prominent in legumes possibly arising from enzymatic actions include 1-penten-3-ol and 1-octen-3-ol, two compounds which had been described as having undesirable odour.
Ketones and hydrocarbons are also detected in the headspace volatile fractions of legumes, though less so than alcohol and aldehydes. Ketones and hydrocarbons are also derived from lipid oxidation, from both (non) enzymatic oxidative degradations [2]. For example, 2,3-pentanedione is found in oxidised soybean oil and has a buttery flavour (Seals and Hammond 1970). Acetone, 6-methylhept-5-en-2-one (methyl heptenone) and (E,E)-3,5-octadien-2-one detected in this study have been found in large quantities in dry beans [10].
Terpenes were also detected in the headspace volatile of legume seeds; the largest number is detected in lentils. Unlike aldehydes, alcohols, ketones and hydrocarbons, which are products of fatty acid oxidation, terpenes are naturally present/synthesised by the plant [2]. α-Pinene and β-pinene are the two most common terpenes in legume samples investigated.

Lipoxygenase is the Most Substantial Contributor for Volatile Evolution
Out of the 11 samples, soybean contains the highest number (18) of discriminating compounds (including volatiles and fatty acids), and this can also be seen on the bi-plot ( Figure 2) as soybean samples are projected the furthest away from other legumes. Volatiles with high VID coefficients in soybean comprise aldehydes (e.g., heptanal), ketones (e.g., 1-octen-3-one) and alcohols (e.g., 1-octen-3-ol and 1-pentanol). In literature, these compounds are associated with undesirable odours in soybean. For example, 1-pentanol has a pungent fusel or solvent-like odour, while 1-octen-3-ol has a mushroom and earthy odour with a low detection threshold at 1 ppb in water and 1 ppm in soymilk [31]. Individual plots presented in Figure 3 illustrate their intensity/abundance in comparison to other legume samples. These compounds are products of lipoxygenase-catalysed fatty acid oxidation. Linoleic acid is a precursor of many aldehydes, ketones and alcohols, as it is susceptible to autoxidation and lipoxygenase-catalysed oxidation [3]. Accordingly, linoleic, stearic, palmitic and oleic acids are also detected in a higher amount in soybean, in descending order (Table 2 and Figure 3). This shows the potential of the multiplatform approach followed by chemometrics, such as multivariate data analysis, to demonstrate relationships amongst different measured attributes-in this case, fatty acids and volatile profile.
However, other legume seeds also contain fatty acids and their own isozymes of lipoxygenases [24]. Even a low amount of fatty acid may generate volatiles contributing to off-odours [4,32,33]. Therefore, while it holds that lipoxygenase-catalysed fatty acid oxidation products are detected in much higher abundance in soybean because of its high linoleic acid (and other fatty acid) content, these volatiles can also be present in other legume seeds.

Soybean and Cowpea Contain Distinctive Butyrolactones
Three butyrolactone compounds were selected as discriminant compounds in soybean (α-methyl-γ-butyrolactone and β-methyl-γ-butyro-lactone) and cowpea (γ-ethyl-γ-butyrolactone). β-scission of fatty acid hydroperoxides can yield carbonyl compounds, which can participate in Maillard and Strecker degradation reactions to yield the aforementioned cyclic volatile compound [34]. Perhaps it is due to the combination of lipoxygenase isozyme specific to soybean, combined with plentiful precursor (Table 1), that α-methyl-γ-butyrolactone was able to be detected exclusively in soybean ( Figure 3). As for the other three lactones, their presence in Phaseolus (β-methyl-γ-butyro-lactone) and all samples except for fava bean (γ-ethyl-γ-butyrolactone) indicates the presence of lipoxygenase with similar specificity. This is supported by Chigwedere, Tadele, Yi, Wibowo, Kebede, Van Loey, Grauwet and Hendrickx [15], with β-methyl-γ-butyro-lactone also detected in Phaseolus samples, whereby it was identified as a marker distinguishing aged and fresh beans cooked for 270 min.

Orange Lentil Contains Discriminant Terpene and Carotenoid Degradation Products
Based on Figure 2, orange lentil has a distinct volatile and fatty acid profile compared to other legumes, with 15 discriminant volatile compounds. Specifically, orange lentil is distinct for the presence of a high number of discriminating terpenes, terpene derivatives and cyclic hydrocarbons. This result is not surprising, as the orange colour of the lentil suggests the presence of carotenoids. A previous study conducted by Zhang, et al., [35] showed that the total carotenoid content of 20 cultivars of red lentil grown in Canada by dry weight ranged between 5.32 and 28.1 µg/g. In addition to being secondary metabolites, volatile terpenes may arise from degradation of carotenes by either legume lipoxygenases or hydroperoxides generated from autolytic and enzyme-catalysed lipid oxidation [2]. Linalool ( Figure 3) and D-limonene can also be considered as potential discriminant compounds for orange lentil, though they were also present in other samples. These volatiles are associated with a citrus and fresh odour [36].

Presence of 2-Butanone and Methylated Compounds in Orange Lentil
In orange lentils, 2-butanone and pyrrole were selected as discriminant compounds with high VID coefficients ( Table 2). The ketone 2-butanone has a "moderately sharp, sweet, pungent, and acetone-like" odour and has been reported as a product of lipoxygenase-catalysed oxidation of unsaturated fatty acid hydroperoxides [37], as well as temperature-accelerated oxidation of saturated fatty acids [3]. According to individual plot (Figure 3), a high relative amount of 2-butanone was detected in orange lentil in the current study and has previously been reported in lentils, navy bean, red kidney bean and peas [2,12,37]. With orange lentil only having 3.9% lipid content (Table 1), the high relative abundance of 2-butanone suggests that the sample may have undergone heat treatment in its distribution chain, or that some components of orange lentil are especially vulnerable to heat treatment. This hypothesis is supported by the presence of pyrrole. Ma, Boye, Azarnia and Simpson [37] similarly reported pyrrole in navy bean, red kidney bean, green lentil and yellow pea that had undergone heat treatment.

Presence of Acid in Chickpea and Fava Bean Suggests Alcohol Dehydrogenase Activity
Hexanoic acid is selected as a discriminant compound in chickpea (Table 2). This compound has been linked with the alcohol dehydrogenase pathway, which converts alcohol (in this case, 1-hexanol) into the corresponding acid [38]. From a visual inspection of the individual plots of 1-hexanol and hexanoic acid (Figure 3), chickpea only had a moderate amount of 1-hexanol, but it had the highest amount of hexanoic acid detected. This seems to suggest that alcohol dehydrogenase was present in a high concentration or had relatively high activity compared to other isozymes. A similar observation can be seen with fava bean, where it had a distinctly higher relative amount of 3-methyl-butanoic acid (VID coefficient 0.863), corresponding to 3-methyl-1-butanol contributing to 19.1% of its headspace volatile. Similar to lipoxygenase, alcohol dehydrogenase also has isozymes. Gomes, Jadrić, Winterhalter and Brkić [30] have reported three isozymes in chickpea cotyledon. This could suggest that chickpea and fava bean contain alcohol dehydrogenase which converts aldehydes to alcohols.

Members of the Phaseolus Group Appears to Contain Similar Dominant Volatile Compounds
In this study, 3-methyl-1-butanol was identified as a discriminant compound common to kidney, black and navy beans. Figure 3 visually illustrates its abundance. As previously stated in Section 3.3, hexanal and 1-penten-3-ol are volatile common to the Phaseolus legumes and have previously been identified as marker compounds in three cultivars of common beans, having high scores in the principle components [10]. This indicates that despite being of different cultivars and colours, kidney, navy and black beans have similar characterising headspace volatiles.

Conclusions
The approach used in this paper, integrating fingerprinting and profiling, is effective to characterise the volatile and fatty acid profiles of the eleven legumes seeds selected. The detected volatiles can be grouped into aldehyde, alcohol, ketone, terpene, ester and lactone and hydrocarbon chemical classes. The lipid profiles comprised palmitic, stearic, oleic, linoleic and α-linolenic acids. Advanced chemometrics utilising multivariate data analysis were used to determine distinctive volatile compounds for different legume species. The occurrence of specific discriminant compounds is hypothesised to be majorly derived from the action of species-specific isozymes, especially lipoxygenases. While findings from this result majorly emphasised how certain volatile compounds discriminate each legume and how they are linked to fatty acids, this information may aid in choosing legume-based ingredients with the desired volatile profile. This insight can also be valuable to legume breeders in selecting legumes with certain fatty acid profile, aiming for higher n-6/n-3 ratios and/or oxidative stability. Lastly, since legumes are typically processed (such as soaking and cooking) prior to consumption, the effects of processing on these legume-specific volatiles is worth further investigation.