Predicting Sake Flavor and Aroma Profiles Using UHPLC-QTOF-MS Metabolomics

Researchers at the National Research Institute of Brewing (Hiroshima, Japan) investigated the correlation between the components of sake and its sensory characteristics, with the sake metabolome analysis method developed using ultra-high-performance liquid chromatography-quadrupole time-of-flight mass spectrometry (UHPLC-QTOF-MS). In addition, orthogonal projections to latent structure models were constructed to predict sensory evaluation data obtained through the quantitative descriptive analysis method from the sake metabolome data. This study provides a new option for explaining the sensory characteristics of sake from its components, contributing to a deeper understanding of these characteristics. A paper based on this research was published in Metabolites (1).

A complex alcoholic beverage composed of numerous components due to it being prepared through many processes and parameters, including the raw ingredient rice and the involvement of microorganisms, such as koji mold and yeast (2), investigation of sake flavor has long been a subject of interest, with the correlations between the components and sensory characteristics studied for many years (3–6). Sensory evaluation of sake is a basic and essential analytical method for the beverage, conducted in sake breweries as well as in laboratories; some evaluations, including those conducted at sake awards, such as the Annual Japan Sake Awards, have assisted in the understanding of sensory characteristics as well as the improvement of sake quality (7).

Previous research has evaluated the components contributing to sake’s flavor using component analysis (8). These findings have informed the organization of flavor terminology and the development of flavor wheels, thereby establishing systematic sensory evaluation methods for sake (9). Other studies have employed statistical approaches to construct models predicting sake flavor from its chemical composition (3). For example, prediction models have been developed for sweetness and fullness using analytical values such as reducing sugar (glucose) and acidity (10,11). However, sake contains hundreds of compounds, and its sensory characteristics arise from complex interactions among them. The correlations between individual components and sensory properties remain incompletely understood, suggesting that some compounds in sake may not yet have been identified. This possibility motivated the study conducted by researchers at the National Research Institute of Brewing (1).

For two years of study, 8 sensory evaluation models of the 2016 brewing year and 11 sensory evaluation models of the 2017 brewing year, including color, ethyl hexanoate, Hine-ka, Nama hine-ka, ethyl acetate, grainy/sweet aroma, sweetness, sourness, body, astringency, harsh taste/acrid taste, aftertaste, and overall quality, demonstrated a predictive performance with Q² > 0.5. Liquid chromatography-based analytical data indicated that it is possible to predict not only taste but also aroma. Additionally, the generalization performance of the prediction models for sensory evaluation attributes common to both years was verified.

The researchers suggest that their work demonstrates the potential to construct predictive models even in the absence of primary aroma components, highlighting the utility of this approach for understanding sake’s sensory characteristics. Key variables identified through model construction provided insights into both known and unknown components influencing sensory evaluation attributes, as well as information on previously unidentified peaks. Comparing prediction models across different years allowed verification of their accuracy in predicting sensory evaluation scores, offering a novel means of linking sake’s chemical composition to its sensory profile. By elucidating correlations between sake components and sensory characteristics, this approach also clarifies the relationship between the sake-making process and sensory outcomes, thereby supporting more efficient product development and quality control (1).

References

1. Kobayashi, T.; Komatsu-Hata, Y.; Saito, R. et al. Modeling the Sensory Characteristics of Japanese Sake Using the Sake Metabolome Analysis Method. Metabolites 2025, 15 (8), 559. DOI: 10.3390/metabo15080559
2. Yazawa, H.; Tokuoka, M.; Kozato, H. et al. Investigation of Relationship between Sake-Making Parameters and Sake Metabolites Using a Newly Developed Sake Metabolome Analysis Method. J. Biosci. Bioeng. 2019, 128, 183–190. DOI: 10.1016/j.jbiosc.2019.01.013
3. Iwano, K.; Ito, T.; Nakazawa, N. Correlation Analysis of a Sensory Evaluation and the Chemical Components of Ginjyo-Shu. J. Brew. Soc. Jpn. 2005, 100, 639–649. DOI: 10.6013/jbrewsocjapan1988.100.639
4. Yoshizawa, K.; Suzuki, D.; Shindo, H.; et al. Effects of Addition of Flavor Components in Sake on Sake Flavor and Taste. J. Brew. Soc. Jpn. 1997, 92, 217–223. DOI: 10.6013/jbrewsocjapan1988.92.217
5. Ito, T.; Komatsu, Y.; Takato, A. et al. Tastes of the Aromatic Alcohols in Ginjyo-Syu. J. Brew. Soc. Jpn. 2008, 103, 562–569. DOI: 10.6013/jbrewsocjapan1988.103.562
6. Ohba, T. Seisyu No Aji. J. Soc. Brew. Jpn. 1980, 75, 623–627.
7. Awards & Contests. National Research Institute of Brewing website. https://www.nrib.go.jp/English/kan/kaninfo.html (accessed 2024-05-01)
8. The Brewing Society of Japan. Jozo-Butsu-No-Seibun; The Brewing Society of Japan, 1999; pp. 2–108.
9. Utsunomiya, H.; Isogai, A.; Iwata, H. et al. Flavor Terminology and Reference Standards for Sensory Analysis of Sake. Rep. Res. Inst. Brew. 2006, 178, 45–52.
10. Sato, S.; Kawashima, H.; Maruyama, Y. Studies on the Taste of Sake Part Iii. Application of Regression Models Relating Sweetness, Fullness and Chemical Date. J. Soc. Brew. Jpn. 1974, 69, 774–777. DOI: 10.6013/jbrewsocjapan1915.69.774
11. Utsunomiya, H.; Isogai, A.; Iwata, H. Amakara Categories for Type Designation. J. Brew. Soc. Jpn. 2004, 99, 882–889. DOI: 10.6013/jbrewsocjapan1988.99.882