Major Sensory Attributes and Volatile Compounds of Korean Rice Liquor (yakju) Affecting Overall Acceptance by Young Consumers

The sensory characteristics and volatile compounds that affect consumers’ acceptance of rice liquors were investigated. A total of 80 consumers evaluated 12 yakju samples and descriptive analysis by 11 trained panelists was conducted. Solvent-assisted flavor evaporation-gas chromatography-mass spectrometry analysis also was conducted revealing 120 volatile compounds in the yakju samples. Sensory attributes (n = 31) except appearance attributes were used for principal component analysis (PCA). As results, fruit odor (apple, hawthorn, omija, and pineapple odor) and flower odor (chrysanthemum, pine, and peppermint odor) were placed on the positive side of PC1 whereas persimmon vinegar odor, bitter taste, alcohol flavor, stinging and coating mouthfeel were located on the negative side of PC1. The yakju samples were mainly characterized by their alcohol content and supplementary ingredients. Sensory descriptors (n = 31; except appearance attributes and p > 0.05) and volatile compounds (n = 30; p > 0.5 correlation coefficient with overall acceptance) were chosen for multiple factor analysis (MFA). The MFA correlation map showed that ethyl propanoate, ethyl-2-hydroxy-2-methylbutanoate, methyl 2-furoate, γ-butyrolactone, 4-ethoxycarbonyl-γ-butyrolactone, hawthorn odor, apple flavor, grape flavor, and sweet taste were positively correlated with young consumers’ overall acceptance. Additionally, negative correlation with overall acceptance was found in 1,3-butanediol, 2,3-butanediol, and 1,1-diethoxy-3-methylbutane.


Introduction
Yakju is a representative traditional Korean rice wine along with makgeolli. Yakju is made with rice as a starch source, water, and nuruk. Nuruk is a starter made from grains and it plays an important role in the flavor of yakju during fermentation because it contains various microorganisms, including fungi and wild types of yeast [1,2]. Thus, during fermentation, sugars, various organic acids, and numerous volatile compounds are produced, affecting the flavors of the yakju [3]. The sensory properties of yakju are affected by various factors, and they are influenced not only by the yeast strain [2] but also by the fermentation process and the starch ingredients [4], and by the degree of milling of the rice used for the nuruk [5]. In addition, supplementary ingredients, such as mulberry [6], Codonopsis

Descriptive Analysis
Descriptive analysis and consumer test were approved by the institutional review board of Dankook University (DKU 2018-10-002). Eleven panelists (females, aged 37-49 years) were selected from 30 preliminary panelists based on their ability to discriminate and describe tastes and flavors in a screening test. These panelists participated in 15 training sessions (2 h per session, twice a week). During the panel training, 48 sensory descriptors (appearance = 3, odor = 17, taste/flavor = 13, mouthfeel = 6, and aftertaste/after-mouthfeel = 9), definitions, reference materials, and intensity of reference materials for the 12 yakju products were developed ( Table 2). The samples were evaluated in an individual sensory booth equipped with a computerized data collection system (Korea Food Research Institute, Wanju-gun, Korea) using a 15 cm line scale (0: none to 15: very strong). The panelists evaluated six samples for each session, and they were given 15 min to test one sample. After testing three samples, they were required to rest for 15 min to prevent sensory fatigue. The yakju samples (40 g) were monadically presented in a glass cup (55 mL, diameter of top of cup = 5 cm) coded with a three-digit randomized number and were presented in a randomized order to prevent bias. Each sample was covered with a watch glass (7 cm in diameter) to minimize changes in odor during the evaluation. Filtered water and crackers were provided as palate cleansers. A spit cup was also provided for panelists to use when they did not want to swallow the samples. Evaluation sessions were conducted in three replications.

Consumer Test
All of the consumers who participated in this study were users of yakju, and they were recruited based on them having no allergies to alcohol or yakju and based on their willingness to participate in the study. A total of 80 consumers (in their 20 s; male = 38, female = 42) participated in the evaluations. The samples were served following the Williams Latin Square design. The participants evaluated the overall acceptance using a 9-point hedonic scale. Six samples were provided for one evaluation session. Each sample (30 mL) was served in a paper cup (70 mL), and filtered water was provided as a palate cleanser. The consumers were asked not to consume any food for at least 1 h before the evaluation.

Identification of Volatile Compounds by GC-MS
Prior to the GC-MS analysis, the volatile compounds of the samples were extracted using the liquid-liquid continuous extraction (LLCE)/solvent-assisted flavor evaporation (SAFE) method. Each sample (1150 mL) and 3-heptanol (328 µg; internal standard) were placed in LLCE apparatus and extracted for 8 h at room temperature using 300 mL of redistilled dichloromethane as an extraction solvent. The LLCE extracts were frozen for 12 h at −20 • C to remove water and then concentrated to 120 mL using a gentle N 2 gas stream. SAFE was applied to remove non-volatile compounds and impurities, such as pigments. The extract (120 mL) was distilled for 30 min at 40 • C under 8.0 × 10 −3 Pa, and the SAFE extract was frozen for 12 h at −20 • C and then dried over anhydrous sodium sulfate (3 g). This extract was concentrated to 500 µL using a gentle N 2 gas stream and then analyzed by GC-MS.
Analyses were performed using an Agilent 7890B GC/Agilent 5977A mass selective detector (Agilent, Santa Clara, CA, USA) with a DB-wax column (60 m length × 0.25 mm i.d. × 0.25 µm film thickness; J&W Scientific, Folsom, CA, USA). Helium was used as a carrier gas with a flow rate of 1.0 mL/min. The oven temperature was initially set at 40 • C for 5 min and then increased at a rate of 5 • C/min to 200 • C, which was held for 20 min. The samples (1 µL) were injected into GC-MS apparatus at split mode (50:1). The injector and detector temperatures were 200 • C and 250 • C, respectively. The ionization voltage was 70 eV, and the mass range was 33-350 m/z. Analyses were conducted in triplicate.
Volatile compounds were identified based on their retention indices (RI), Wiley Registry of Mass Spectral Data (9th ed.) and by using a NIST08 database (Agilent). The concentration of the volatile compounds was semi-quantified using Formula (1), wherein the correlation coefficient of the peak area ratio and the amount ratio was assumed to be 1.

Statistical Analysis
The descriptive analysis results and the consumers' acceptance data were analyzed by analysis of variance (ANOVA) to determine the differences among the samples. Student-Newman-Keuls (SNK) multiple comparison was used when a significant difference was found among the samples (p < 0.05). Principal component analysis (PCA) was conducted to summarize the results of the sensory characteristics of the yakju samples. Pearson's correlation analysis was performed to investigate the relationship between consumers' acceptance and volatile compounds. Moreover, multiple factor analysis (MFA) was performed to investigate the relationship among sensory attributes, volatile compounds, and consumers' overall acceptance of the 12 yakju samples. All statistical analyses were conducted using XLSTAT (Ver. 2017.1, Addinsoft, Paris, France).

Descriptive Analysis of Yakju Samples
The results of the ANOVA showed significant differences in 33 attributes of the 48 attributes (Table 3). Among the samples, Y10 and Y8 showed the highest scores for redness and yellowness, respectively. The redness of Y10 might be due to the red color of the raw materials, such as the Schizandra chinensis fruit (omija), Cornus officinalis, and the Lycium chinense fruit. Similarly, some of the samples were characterized by attributes induced by their raw materials. For example, Y5, which contained chrysanthemum, had a strong chrysanthemum, peppermint, and pine odor. Y1 was highest in ginseng and hawthorn odor and Y7 had the highest score for bitterness and bitter aftertaste. This result might be due to saponin, which has a bitter taste [17]. Other ingredients such as fructose, also seemed to affect sweet taste in the Y2 and Y10 samples. In terms of alcohol flavor, samples with high alcohol contents such as Y7 (16.0%) and Y8 (18.0%) tended to be the highest in alcohol flavor, whereas Y10 (12.0%) and Y12 (11.0%), which contained relatively low alcohol content, were the lowest in alcohol flavor. Especially, Y8, which had the highest alcohol content, was the highest in body, coating and residue mouthfeel.
Y4 showed the highest scores for persimmon vinegar odor, sourness, stinging mouthfeel, and stinging aftertaste. This result is possibly due to the formation of acetic acid during fermentation, considering that acetic acid is associated with vinegar scent and a pungent odor [18].
The PCA result for the 31 sensory descriptive attributes of the 12 yakju samples is shown in Figure 1. Appearance-related descriptors (degree of clearness/redness/yellowness) were excluded to focus on the odor and flavor of the yakju samples, and attributes that had significant difference among the samples (n = 31) were used for the PCA. A total of 67.6% of variance could be explained by PC1 (40.4%) and PC2 (27.3%). Characteristics related to sweetness, flower, and fruit ingredients were located on the positive side of PC1, whereas bitter taste, burning mouthfeel, and alcohol odor and flavor were located on the negative side of PC1. Specifically, samples such as Y1, Y2, Y3, Y5, Y10, and Y12, which contained fruit or floral ingredients, were located on the positive side of PC1; these samples had hawthorn odor, sweet taste, sour taste, apple flavor, grape flavor, or tangerine peel odor/flavor. Y4 showed the highest scores for persimmon vinegar odor, sourness, stinging mouthfeel, and stinging aftertaste. This result is possibly due to the formation of acetic acid during fermentation, considering that acetic acid is associated with vinegar scent and a pungent odor [18].
The PCA result for the 31 sensory descriptive attributes of the 12 yakju samples is shown in Figure 1. Appearance-related descriptors (degree of clearness/redness/yellowness) were excluded to focus on the odor and flavor of the yakju samples, and attributes that had significant difference among the samples (n = 31) were used for the PCA. A total of 67.6% of variance could be explained by PC1 (40.4%) and PC2 (27.3%). Characteristics related to sweetness, flower, and fruit ingredients were located on the positive side of PC1, whereas bitter taste, burning mouthfeel, and alcohol odor and flavor were located on the negative side of PC1. Specifically, samples such as Y1, Y2, Y3, Y5, Y10, and Y12, which contained fruit or floral ingredients, were located on the positive side of PC1; these samples had hawthorn odor, sweet taste, sour taste, apple flavor, grape flavor, or tangerine peel odor/flavor. On the contrary, samples containing Supplementary Materials other than fruit or floral ingredients, such as sweet pumpkin (Y4), balloon flower root (Y6 and Y7), or soy bean (Y8), were located in quadrant 2. These samples were mainly characterized by yeast odor, roasted grain odor, alcohol flavor, and body and stinging attributes. Characteristics related to mouthfeel and taste were      On the contrary, samples containing Supplementary Materials other than fruit or floral ingredients, such as sweet pumpkin (Y4), balloon flower root (Y6 and Y7), or soy bean (Y8), were located in quadrant 2. These samples were mainly characterized by yeast odor, roasted grain odor, alcohol flavor, and body and stinging attributes. Characteristics related to mouthfeel and taste were located on the positive side of PC2, whereas most of the attributes related to odor were located on the negative side of PC2. Particularly, salty taste, sour taste/aftertaste, and body mouthfeel were located on the positive side of PC2, whereas pine odor, peppermint odor, and chrysanthemum odor were located on the negative side of PC2. Samples with high alcohol levels were located on the positive side of PC2, whereas those with relatively low alcohol content, such as Y5 and Y12, were located on the negative side of PC2. Among the samples, Y9 was the only sample located in quadrant 3, and it had a low score for sour, sweet, umami, and fruit-related notes compared with other samples.
Overall, the present results indicated that the yakju samples were characterized mainly by their supplementary raw materials and alcohol content, which affect the overall odor and flavor of the samples.

Consumers' Acceptance
The mean overall acceptance scores by 80 young consumers are shown in Table 4. Significant differences among the 12 samples were found in the overall acceptance scores. Overall acceptance was highest for Y3 (6.71), followed by Y10 (6.41), and Y2 (6.34) and was lowest for Y8 (3.25). Generally, overall acceptance was higher for samples with fruit-related supplementary ingredients (Y2, Y3, and Y10) than that for samples with root-related bitter ingredients (Y7 and Y8). Apparently, consumers preferred those samples with fruit-related characteristics over the bitter and yeasty samples. Similarly, Lee and Lee [19] studied the sensory attributes and acceptance of 10 yakju samples with supplementary ingredients, and they reported that the acceptance of various clusters of consumers were positively associated with fruit flavor, sweet aroma, and medicinal herb aroma but astringent mouthfeel, bitter taste, and yeast flavor were negatively associated with consumers' acceptance. Moreover, Kwak et al. [20] reported that the key liking factors of rice wine by American panelists were sweet, sour, and apricot flavors, whereas the key disliking factors were yeasty and nutty characteristics. They reported positive correlations between overall acceptance and fruit-related characteristics (apple, peach, and pear), confirming the key liking factors for the yakju samples [20].
Alcohols, the largest group of volatile compounds in alcoholic beverages such as red wine [24,25], are produced during yeast metabolism, and alcohol content varies depending on the yeast starters [26]. Fusel alcohols, such as n-propyl alcohol, iso-butyl alcohol, and iso-amyl alcohol, are produced during the fermentation of amino acid in yeast [12,27]. Among them, isoamyl alcohol has the highest amount of the volatile compounds as reported by Kim et al. [28]. Isoamyl alcohol is known to boost aroma and flavor when it exists in low amounts [27]. Additionally, 2-methyl-1-propanol, which is one of the aliphatic alcohols related to alcohol odor [1], was high in Y1 (10,085 µg/L), Y2 (12,313 µg/L), and Y5 (10,222 µg/L).
Esters are produced during alcoholic fermentation by yeast; they have a fruit-like aroma and thus they positively influence the aroma quality of liquors [29][30][31]. The large number of volatile ester compounds (n = 32) could be explained by the various supplementary ingredients used to produce the samples. Isoamyl acetate is related to sweet and fruity aromas, and Mamede et al. [29] reported that low concentration of isoamyl acetate results in low consumers' acceptance of sparkling wine samples. Isoamyl acetate content was highest in Y10 and lowest in Y8. Considering that overall acceptance of Y10 (6.41) was much higher than that of Y4 (3.25) as in Table 4, the results of this study confirmed that isoamyl acetate might be one of the key compounds affecting consumer acceptance. Among the samples, Y4 contained high amounts of ethyl lactate (45,468 µg/L), isoamyl lactate (446 µg/L), diethyl succinate (8015 µg/L), and ethyl hydrogen succinate (6056 µg/L). Apostolopoulou et al. [32] showed that the ethyl lactate content of samples of bottled Greek distillates (tsipouro) differs from that of homemade samples, suggesting that production methods might influence the amount of ethyl lactate. Argyri et al. [33] also analyzed the volatile compounds of meat samples under different temperatures. As the temperature increased (from 0 • C to 15 • C), the amount of ethyl lactate also increased. These results suggested that the ethyl lactate content of Y4 was affected by the manufacturing environment, such as production methods and temperature.
Ethyl octanoate contents in all of the samples was less than the threshold, which was reported to be 170 µg/L [34]. Y7 had the highest level of 2-phenylethyl acetate (759 µg/L), followed by Y6 (600 µg/L). This might be due to the specific yeast used in Y6 and Y7, considering that this compound is known to be formed during fermentation by yeast [34]. The intensity of 2-phenylethyl acetate in Y7 and Y6 was much higher than threshold (180 µg/L), which was reported in [34].

Relationship among Sensory Attributes, Volatile Compounds and Consumers' Acceptance of Yakju Samples by MFA
For the MFA, 30 volatile compounds (acids = 3; alcohol = 8; aldehyde = 1; ester = 8; ketone = 2; lactone = 3; phenol = 1; miscellaneous = 4) were selected from the 120 volatile compounds based on a correlation coefficient of >0.5 with consumers' overall acceptance. A correlation map of descriptive attributes, volatile compounds, and consumer acceptance is in Figure 2a and loading of 12 yakju samples in the first two dimensions by MFA is shown in Figure 2b. A total of 59.3% of variance was explained by F1 (47.7%) and F2 (11.7%). Similar to the results in Figure 1, fruit-related sensory attributes were placed on the negative side of F1, while bitterness, mouthfeel, and alcohol flavor were placed on the positive side of F1 (Figure 2a).
One of the abundant and important volatile lactones found in wine is γ-butyrolactone [25], which has a fruity aroma [31]. Not only γ-butyrolactone but also 4-ethoxycarbonyl-γ-butyrolactone was positively related with consumers' acceptance. Considering that the content of volatile lactones varies depending on aging time and type of yeast strain in sherry wine [41], the type of yeast strain in the nuruk used for the yakju samples might have influenced the production of those volatile compounds. In addition, Lee et al. [40] reported that nuruk generally contains various kinds of microorganisms compared with to ipguk (koji), which only contains Aspergilus oryzae, leading to the production of more volatile compounds in nuruk than in ipguk. Lee et al. [42] reported an absence of 2,3-butanediol in ipguk samples, and a relatively low amount of 2,3-butanediol in Y2 (198.9 µg/L) and Y3 (205.5 µg/L) in this study implies that Y2 and Y3 samples might be made of ipguk.
Volatile alcohol compounds have a pungent mouthfeel and a pungent and "herbaceous" odor [43]. Described as having a grassy, medicinal, fusel, and spirituous odor, 1-butanol has an odor threshold of 150,000 µg/L [22,43,44]. Although 1-butanol correlated with consumers' acceptance, the odor of 1-butanol may be imperceptible considering that the 1-butanol contents in the samples ranged from 44.8 µg/L to 1,159.9 µg/L, which was considerably lower than the threshold value.
Volatile phenolic compounds are also formed primarily through alcoholic fermentation [45] and are known to be important in the overall aroma of wine [46] and flavors of dark beer [47]. Generally, volatile phenolic compounds such as 4-vinylphenol have a "nutty" odor similar to "almond shell" [45], spicy, and medicinal-like aromas [48]. Butkhup et al. [46] also reported that these compounds had "heavy pharmaceutical" odor. The existence of 4-vinylphenol was only found in Y1 (111 µg/L), Y2 (73.1 µg/L), Y3 (85.4 µg/L), and Y12 (36.8 µg/L). Although this compound had a positive relationship with young consumers' acceptance (r = 0.64), the effect of 4-vinylphenol on young consumers' overall acceptance for yakju might be negligible considering its threshold (610 µg/L) [46]. Some volatile alcohols (3-butanediol, 2,3-butanediol, 1-octanol, and propylene glycol), acetic acid, γ-hexalactone, and 1,1-diethoxy-3-methylbutane were associated with persimmon vinegar odor, roasted grain odor, yeast odor, kudzu flavor, and coating mouthfeel. Y4 and Y8, which had lower consumer acceptance than other samples, were associated with volatile alcohols such as 1,3-butanediol, 2,3-butanediol, and propylene glycol. This result implied that lower consumer acceptance of these samples might be caused by those volatile alcohol compounds and bitter taste, bitter aftertaste, alcohol flavor, and kudzu flavor. Butanediols were produced from carbohydrates during the alcoholic fermentation primarily by S. cerevisiae, a major yeast in yakju [49][50][51]. In particular, 2,3-butanediol is the dominant volatile compound in wine, and it has a bitter taste [50]. The highest contents of these compounds in Y4 and Y8 might be due to yeast, such as S. cerevisiae, which is involved in the fermentation of rice or supplementary starch ingredients, such as sweet pumpkin and soybean. This result suggested pungent and sour odors like persimmon vinegar odor, related to volatile acetic acid, and some volatile alcohols such as 1,3-butanediol, 2,3-butanediol, and propylene glycol, might negatively affect young consumers' acceptance.
starch in addition to rice. Kim et al. [4] conducted a sensory evaluation of traditional liquor samples made from various starch sources. They showed that the acceptance of liquors made with corn starch and brown rice was higher than that of liquors made with glutinous rice or non-glutinous rice or potato starch. Liquors made with corn starch or brown rice tended to contain a relatively lower amount of acetic acid and a higher amount of fructose than the liquors made with non-glutinous rice, suggesting that the type of starch could affect the sensory and chemical properties of liquors. Apart from starch source, rice protein in yakju samples increases the pH of yakju samples, causing the formation of off-flavor [54]. Moreover, León-Rodríguez et al. [55] suggested that the presence of minor compounds (e.g., some volatile ethyl esters, terpenes, acids, and furans) at low concentrations could cause a synergic effect with other volatile compounds, leading to the production different odor characteristics. Therefore, further investigation on interactions among volatile compounds is needed to understand the odor characteristics and volatile compounds that affect consumers' overall acceptance.

Conclusions
Twelve yakju samples were characterized, based on sensory descriptors by PCA. The result of the PCA showed that yakju samples were characterized mainly by their supplementary raw materials, which affect their overall odor and flavor. As shown by the result of the MFA correlation map, the Furthermore, one of the volatile acetals, 1,1-diethoxy-3-methylbutane, was also associated with Y4 (49.4 µg/L) and Y8 (142 µg/L), which had lower consumers' acceptance than other samples. This compound was found in liquor samples such as Chinese liquor [23] and Italian grape marc spirit [52]. Volatile acetals are known to be produced by aldehydes in the presence of excessive content of ethanol [23,53]. This suggested that 1,1-diethoxy-3-methylbutane might be produced by relatively high contents of ethanol or aldehyde-related compounds, as in Y4 and Y8. Significant differences in fruit-related attributes were found among the samples, even though the intensities of fruit-related attributes were weak. Thus, overall, the correlation map by MFA showed that fruit-related sensory attributes and volatile compounds were closely related with consumer acceptance. The fruit-related sensory attributes and volatile esters were placed in quadrant 2 and 3, while roasted grain odor, yeast odor, root-related flavor (kudzu flavor), and all mouthfeel attributes were placed in quadrant 1 and 4. The result of this study confirmed the results by Jung et al. [1], who reported that rice wine samples were distinguished by their volatile alcohol and volatile ester. Furthermore, descriptors such as yeast odor and cereal flavor were located opposite to the fruit and sweet aroma in the PCA map.
In addition to volatile compounds that might originate from supplementary ingredients, major ingredients such as different sources of starch might affect the flavors and therefore consumers' acceptance. While most of the samples were produced mainly from rice, Y1, Y2, Y3, and Y5 used corn starch in addition to rice. Kim et al. [4] conducted a sensory evaluation of traditional liquor samples made from various starch sources. They showed that the acceptance of liquors made with corn starch and brown rice was higher than that of liquors made with glutinous rice or non-glutinous rice or potato starch. Liquors made with corn starch or brown rice tended to contain a relatively lower amount of acetic acid and a higher amount of fructose than the liquors made with non-glutinous rice, suggesting that the type of starch could affect the sensory and chemical properties of liquors. Apart from starch source, rice protein in yakju samples increases the pH of yakju samples, causing the formation of off-flavor [54]. Moreover, León-Rodríguez et al. [55] suggested that the presence of minor compounds (e.g., some volatile ethyl esters, terpenes, acids, and furans) at low concentrations could cause a synergic effect with other volatile compounds, leading to the production different odor characteristics. Therefore, further investigation on interactions among volatile compounds is needed to understand the odor characteristics and volatile compounds that affect consumers' overall acceptance.

Conclusions
Twelve yakju samples were characterized, based on sensory descriptors by PCA. The result of the PCA showed that yakju samples were characterized mainly by their supplementary raw materials, which affect their overall odor and flavor. As shown by the result of the MFA correlation map, the yakju samples tended to be classified by sensory attributes and volatile compounds. The results of this study showed that acceptance of yakju samples was largely influenced not only by their alcohol content but also by their supplementary ingredients, especially those related to fruit-related aroma. On the contrary, volatile acetic acid, and some volatile alcohols (1,3-butanediol, 2,3-butanediol, and propylene glycol), and 1,1-diethoxy-3-methylbutane were related with persimmon vinegar odor, roasted grain odor, and yeast odor, and negatively correlated with young consumers' acceptance of yakju. This is a first report on how the major sensory attributes and volatile compounds of Korean rice liquor (yakju) affect overall acceptance by young consumers, even though the number of consumers who participated in this study (n = 80) was not sufficient. Overall, the results of this study suggested that acceptance of yakju products could be improved by controlling some volatile esters that resulted from supplementary ingredients, or specific volatile alcohols and acids produced during fermentation. Considering that the aroma and flavor of yakju could vary depending on the starch source, supplementary ingredients, yeast strains, and fermentation process, further investigation is needed on the specific yeast strains and fermentation conditions that affect the formation of volatile compounds and consumers' acceptance of yakju. Along with this, further research on interactions among volatile compounds is also needed to understand the odor characteristics that affect consumers' overall acceptance.

Conflicts of Interest:
The authors declare no conflict of interest.