The Formation Mechanism and Regulation of Fermented Bean Paste Flavor: Microbial Community Succession, Flavor Compound Synthesis, and Applications of Intelligent Biomanufacturing Technology

ABSTRACT

As an important carrier of East Asian dietary culture, the unique flavor and quality of fermented bean paste directly affect its market competitiveness. However, previous studies have primarily focused on microbial communities or overall flavor components, and there is still a lack of comprehensive understanding of the systematic correlation and dynamic regulation mechanisms between the two. The scientific elucidation and precise regulation of the microbial–metabolic network during the fermentation process have become critical challenges facing the industry today. Based on current research on the flavor, microbial community, and flavor‐related pathways of fermented bean paste, this review provides a systematic framework integrating microbial communities, flavor formation, and intelligent regulation technologies. It systematically elaborates on the spatiotemporal heterogeneity of microbial communities during fermented bean paste fermentation, clarifying how core functional microbiota drive flavor synthesis through stage‐specific metabolic activities. Furthermore, it discusses the application potential of synthetic microbial communities and digital twin technology in the precise monitoring of fermentation and targeted regulation of flavor. This review aims to provide theoretical support for the standardization and intelligent upgrading of fermented bean paste production processes, thereby promoting the development of the fermented bean paste industry toward high quality, high efficiency, and intelligence.

1. Introduction

As a traditional fermented food with a long history in East Asia, fermented bean paste is deeply cherished by consumers for its unique flavor and rich nutritional value (Park et al. 2024). This bean paste, made from beans through microbial fermentation, is a profound symbol of East Asian culinary culture. It carries millennia of agricultural civilization and family heritage, while also serving as an indispensable daily condiment that shapes the eating habits and taste preferences of hundreds of millions of people in modern Asian diets. It is primarily made by cleaning, soaking, steaming, and cooling beans, which are then pressed into bricks. After inoculation with microorganisms, the bricks are dried in a ventilated area for 1–2 months to form koji (L. Liu, Chen, et al. 2022; X. Yue et al. 2021). This is followed by the fermentation stage of the paste mash, where the mold on the surface of the koji is washed off, and the koji is cut into pieces and mixed with 20% saline solution in a jar (X. Yue et al. 2021). Fermentation continues for another 1–2 months, ultimately yielding a reddish‐brown or brown finished fermented bean paste. The entire fermentation process typically lasts 4–6 months and can be roughly divided into three stages based on the biological transformation of compounds (X. Yue et al. 2021). The initial stage is the raw material decomposition period, during which various enzymes secreted by microorganisms gradually break down proteins and starch in the ingredients. The mid‐stage is the flavor formation phase, where a series of biochemical reactions create the characteristic aroma of fermented bean paste. The final stage is the flavor balancing period, during which flavor substances are further transformed and balanced under the influence of environmental microorganisms, harmonizing the overall flavor. However, due to differences in raw material selection and production techniques across countries and regions, fermented bean paste exhibits distinct flavor and texture characteristics. For example, Japanese miso blends the flavors of soybeans and rice, Korean doenjang highlights the natural taste of soybeans, while Chinese Northeast‐style fermented bean paste combines soybeans with wheat (J. Zhao et al. 2011). These differences enable fermented bean paste to meet the diverse tastes and dietary needs of global consumers.

The formation of fermented bean paste's flavor profile is a complex biochemical process involving the synergistic action of various microorganisms and the dynamic accumulation of metabolites, representing a core scientific issue in fermented bean paste research (X. Yue et al. 2021). The perception of fermented bean paste's flavor is determined by the synergistic interaction of nonvolatile taste substances and volatile flavor compounds, with the latter directly shaping consumers' first impressions through olfactory receptors (B. Wu et al. 2022). The aroma characteristics of fermented bean paste arise from the cooperative metabolism of microorganisms such as yeasts and molds on the bean matrix (X. Yue et al. 2021). These microorganisms secrete enzymes to break down macromolecules into small‐molecule precursors, which are then transformed into a rich array of flavor compounds, ultimately creating a multilayered sensory experience (X. Zhang, Wei, et al. 2021). Traditional natural fermentation relies on the diversity of environmental microorganisms to ensure flavor complexity, but seasonal changes and fluctuations in temperature and humidity can lead to microbial imbalances, potentially causing quality inconsistencies and safety risks (Mudoor Sooresh et al. 2023). In contrast, industrial production ensures stability through the inoculation of pure cultures and parameter control, but the limited strains artificially selected struggle to replicate the metabolic network of natural microbial communities, resulting in the loss of certain flavor compounds. Therefore, how to scientifically analyze and reconstruct the microbial–metabolic network of traditional fermentation has become a central challenge in current research.

In recent years, the rapid development of molecular biology and omics technologies has provided new research tools for elucidating the microbial mechanisms of fermented bean paste (Tong et al. 2025). By integrating multiomics technologies, researchers have systematically uncovered the correlation mechanisms between microbial communities and flavor compounds. In microbial community analysis, 16S rRNA gene sequencing and metagenomic sequencing technologies not only enable dynamic tracking of microbial population succession but also reveal the evolutionary patterns of functional gene (K. Zhang, Tang, Li, et al. 2025). For flavor compound detection, techniques such as gas chromatography–mass spectrometry (GC–MS) and gas chromatography–ion mobility spectrometry (GC–IMS) have achieved precise identification of volatile flavor compounds (B. Yang, Wang, et al. 2025; Yu et al. 2022). Building on this foundation, the introduction of smart technologies has further propelled the fermentation process of fermented bean paste toward precision and intelligence (Nikoloudaki et al. 2024). Given the dynamically complex and multifactor coupled nature of traditional fermentation, the stability and reproducibility of its flavor heavily rely on real‐time monitoring and regulation of microbial activity and metabolites. Intelligent biomanufacturing technologies such as synthetic biology and digital twins have emerged as the times require, providing new approaches for the automated regulation and intelligent optimization of fermentation processes, as well as laying a solid theoretical foundation for flavor optimization strategies based on microbial regulation (İnanlar and Altay 2026; Nikoloudaki et al. 2024; Y. Zhang, Zhu, et al. 2025).

Most existing reviews on fermented bean paste fermentation focus on single or partial characteristics, and systematic and integrated analyses of the microbial community structure, flavor characteristics, and intelligent regulation of fermented bean paste remain insufficient. Therefore, this paper systematically elaborates on the spatiotemporal heterogeneity of microbial communities in fermented bean paste and the application of detection technologies. Second, it summarizes the composition of key flavor compounds and discusses the core role of metabolomics in flavor analysis. On this basis, the biosynthetic pathways of key flavor substances and their association mechanisms with microbial metabolism are further integrated. Finally, effective strategies for replicating the traditional flavor characteristics of fermented bean paste using intelligent biomanufacturing technologies are proposed. This review provides a relatively comprehensive knowledge framework covering microbial community succession, flavor biosynthesis, and intelligent regulation (Figure 1). It not only further enriches and deepens the systematic understanding of the entire fermentation process of fermented bean paste, but also lays a scientific foundation for promoting the high‐quality and high‐efficiency development of the fermented bean paste industry.

FIGURE 1.

Discussion on microbial community, flavor compounds, correlation analysis and application of intelligent biomanufacturing technology in fermented bean paste.

2. Spatiotemporal Heterogeneity of Microbial Communities in Fermented Bean Paste and Detection Techniques

Fermented bean paste harbors abundant microbial resources, such as Bacillus, Lactobacillus, Mold, and Yeast, which play a crucial role in the fermentation flavor and quality of the paste (Table 1; X. Yue et al. 2021). These microorganisms may originate from raw materials, air, fermentation containers, water, and so forth, and their growth and metabolism in fermented bean paste contribute to its unique flavor (Tong et al. 2025). Therefore, studying the composition of microbial communities in the internal environment of fermented bean paste using various detection techniques is of great significance.

TABLE 1.

Types, functions, and related flavor types of microorganisms in fermented bean paste.

Microbial Genus		Description	Function	References
Molds	Aspergillus oryzae	Widely used in fermented bean paste production	Decomposes starch and protein, contributing to basic flavor and aroma precursors	Hu et al. (2022)
	Aspergillus niger	Used in fermented bean paste fermentation	Produces acidic substances and enzymes, contributing to acidification and flavor development	F. Tian, Woo, et al. (2022)
	Aspergillus sojae	Used for fermented bean paste production	Decomposes protein and starch to produce unique flavor compounds	X. Yue et al. (2021)
	Mucor mucedo	Present in fermented bean paste fermentation	Decomposes protein, fat, and cellulose, affecting flavor and taste	Ashu et al. (2016)
	Monascus purpureus	Used in some fermented bean paste varieties	Produces red pigment and enzymes, influencing color and flavor	Y.‐D. Li, Deng, et al. (2022)
Yeasts	Torulopsis versatilis	Used in fermented bean paste production	Decomposes soybean components, producing significant flavor compounds	Teng et al. (2004)
	Saccharomyces cerevisiae	Present in fermented bean paste fermentation	Ferments sugars, produces alcohol and CO2, contributing to flavor and taste	Yao et al. (2022)
	Zygosaccharomyces rouxii	Commonly used in high‐salt fermented foods like fermented bean paste	Produces alcohols and esters in high‐salt environments	Yao et al. (2022)
	Candida	Present in fermented bean paste microbiota	Involved in sugar fermentation and production of various flavor substances	Yao et al. (2022)
	Hansenula anomala	Found in fermented bean paste fermentation	Contributes to flavor substance generation and improves product quality	Cui et al. (2020)
LAB	Lactobacillus plantarum	Common in fermented bean paste	Produces lactic acid and other organic acids, lowering pH and inhibiting spoilage microbes	S. Li et al. (2021)
	Lactococcus lactis	Present in fermented bean paste	Ferments sugars to produce lactic acid, enhancing flavor and texture	Ilango and Antony (2021)
	Lactobacillus pentosus	Found in fermented bean paste	Participates in fermentation, improving flavor and taste	J. Lee et al. (2023)
	Lactobacillus fermentum	Used in fermented bean paste production	Ferments sugars to produce lactic acid and other flavor substances	Nacchio et al. (2022)
	Leuconostoc mesenteroides	Common in fermented bean paste	Produces lactic acid and CO2 from sugars, contributing to flavor and taste	P. Zhang, Zhang, et al. (2019)
	Tetragenococcus halophilus	Found in high‐salt fermented foods like fermented bean paste	High‐salt tolerance; produces lactic acid and other flavor compounds	Pan et al. (2022)
	Lactobacillus acidophilus	Present in some fermented bean pastes	Improves flavor and increases nutritional value	J.‐S. Kim et al. (2016)

2.1. Spatiotemporal Heterogeneity of Microbial Community Structure in Fermented Bean Paste

The fermentation process of fermented bean paste is a dynamic and continuous succession of microbial communities, exhibiting distinct stage‐specific characteristics (Figure 2). Through complex microbial interaction networks, this process drives material transformation and flavor formation. In the initial fermentation stage, molds such as Aspergillus oryzae, Aspergillus sojae, Mucor mucedo, and Aspergillus niger dominate the microbial community (J. H. Lee et al. 2010). These molds secrete hydrolytic enzymes such as proteases and amylases, which break down macromolecular proteins and starch in beans into smaller molecules such as peptides, amino acids, and sugars (Tong et al. 2025). This not only significantly enhances the umami and sour characteristics of the paste but also provides a nutrient substrate for subsequent microbial growth. During the midfermentation stage, lactic acid bacteria (LAB; e.g., Tetragenococcus halophilus, Lactobacillus plantarum, Pediococcus pentosaceus, and Pediococcus acidilactici) and yeast (e.g., Pichia, Rhodotorula, and Candida) gradually become the dominant microbial groups (D.‐H. Kim et al. 2002). LAB lower the pH of the system through acid production, which effectively inhibits the growth of harmful microorganisms while creating favorable conditions for other functional microbes, thereby promoting symbiotic and synergistic interactions within the fermentation system (Elhalis et al. 2024). Yeast converts sugars such as maltose and glucose into metabolic products like ethanol, glycerol, arabitol, and succinic acid through glycolysis (Tong et al. 2025). These compounds serve as important precursors for the synthesis of esters and sugar alcohols, directly influencing the formation and enhancement of fermented bean paste flavor and contributing to its foundational aromatic profile. In the late fermentation stage, as salt concentration increases, oxygen levels decrease, and metabolic products accumulate, salt‐tolerant microorganisms such as Bacillus, Weissella, Leuconostoc, and certain yeasts begin to dominate the fermentation process (X. Liu et al. 2025). Among these, Bacillus secretes various enzymes that are significantly positively correlated with aromatic flavor indicators (S.‐H. Yi and Hong 2021). Weissella increases the content of organic acids and esters in the fermented bean paste, promoting the synthesis of flavor compounds (Gong et al. 2025). Leuconostoc can act as a fermentation starter, flavor enhancer, and leavening agent to improve product safety and sensory characteristics (Shin and Han 2015).

FIGURE 2.

Spatiotemporal heterogeneity of microbial communities in fermented bean paste.

In the spatial dimension, the microbial community distribution within the fermented bean paste fermentation system exhibits clear nonuniform characteristics, manifested as differences in microbial population structure, quantity distribution, and metabolic activity along the vertical stratification of the fermenting mash (Figure 2; S. Zhao, Niu, et al. 2023). Specifically, the upper layer of the mash, directly exposed to sunlight, has higher temperature and water activity, which favors the growth of aerobic or facultative anaerobic microorganisms such as Staphylococcus and Zygosaccharomyces (J. Y. Jung et al. 2014). These microorganisms possess enhanced amino acid synthesis capabilities, and the amino acids they produce can serve as key precursors that are further converted into volatile flavor compounds through processes such as the Strecker degradation, thereby promoting the accumulation of flavor substances (S. Liu, Sun, et al. 2022; S. Zhao, Niu, et al. 2023). The microbial diversity in the middle layer of the mash lies between that of the upper and lower layers, but microbial interactions are more tightly knit. Co‐occurrence network analysis reveals that bacteria and fungi in this layer have the lowest average network distance and clustering coefficient, indicating stronger synergistic interactions. However, their metabolic activity is relatively weaker, suggesting a primary role in intermediate product transformation. The lower layer of the mash, with limited sunlight and low oxygen levels, is primarily enriched with anaerobic microorganisms such as Bacillus. These microorganisms tend to utilize nitrate or nitrite for anaerobic respiration, consuming amino acids and organic acids to maintain the stability of the fermentation system. Notably, Bacillus still carries genes related to arginine, proline, and phenylalanine metabolism, enabling it to participate in protein degradation and generate flavor precursors (S. Zhao, Niu, et al. 2023). Gradient variations in environmental factors such as light intensity and water activity further exacerbate microbial spatial differentiation. Sunlight not only directly affects the growth of light‐sensitive microorganisms but also inhibits the activity of Bacillus (J. Y. Jung et al. 2014). Differences in water activity shape distinct ecological niches by regulating microbial intracellular osmotic balance. This spatial heterogeneity is directly reflected in the distribution of microbial abundance. Additionally, microbial interactions are closely related to spatial heterogeneity. For example, Staphylococcus and Zygosaccharomyces in the upper layer may synergistically promote flavor precursor synthesis through metabolic cross‐feeding, while their competitive relationship with Bacillus in the lower layer further reinforces community spatial differentiation (S. Zhao, Niu, et al. 2023).

During the fermentation of fermented bean paste, microbial communities undergo significant spatiotemporal heterogeneous succession. This succession is not a random change but rather a continuous adaptation and turnover over time and space, ultimately forming a stable microbial community structure. Within this stable structure, various microorganisms do not exist in isolation but are closely interconnected through synergistic interactions. They cooperate and mutually promote one another, collectively exerting a positive influence on material transformation and flavor compound generation during fermentation. This plays a crucial role in enhancing the final product's flavor quality, working together to complete the fermented bean paste fermentation process and endowing it with unique flavor characteristics.

2.2. Key Technologies for Detecting Microbial Communities in Fermented Bean Paste

2.2.1. Traditional Microbial Community Detection Techniques

In the field of food microbiology, traditional microbial culture techniques, polymerase chain reaction‐denaturing gradient gel electrophoresis (PCR‐DGGE), and terminal restriction fragment length polymorphism (T‐RFLP) have all demonstrated unique value in studying microbial communities in fermented foods such as fermented bean paste (El Sheikha and Hu 2020). Traditional culture techniques, regarded as the “gold standard” for microbial analysis, utilize selective culture media to simulate the natural growth environment of microorganisms, enabling the isolation, purification, and identification of microbes in samples (Dabban et al. 2024). These techniques play an irreplaceable role in fermented bean paste research. For example, researchers have successfully isolated and identified key functional strains such as Monascus, Candida, and Aspergillus from fermented bean paste through gradient dilution and plate streaking (Ren et al. 2021). These strains are directly involved in flavor formation and quality regulation during fermentation (Ren et al. 2021; Li et al. 2021). However, this method heavily relies on the cultivability of microorganisms, being time‐consuming, labor‐intensive, and exhibiting low sensitivity (Srinivasan et al. 2015). It also fails to detect unculturable or slow‐growing microorganisms, limiting the completeness of microbial community information (Pillay 2020). Additionally, contamination risks and interference from nontarget microorganisms further reduce the reliability of results. To address the shortcomings of traditional culture methods, molecular fingerprinting techniques such as PCR‐DGGE and T‐RFLP emerged. These techniques, based on microbial DNA characteristics, enable rapid assessment of microbial community composition and diversity without cultivation. For instance, Ge et al. (2020) used PCR‐DGGE to reveal the succession patterns of fungal communities at the family level during the natural fermentation of fermented bean paste, providing a theoretical basis for optimizing inoculation strategies. K. Zhang, Tang, Liu, et al. (2025) employed T‐RFLP to analyze the composition and dynamic changes of complex microbial communities in fermented bean paste, offering scientific support for fermentation process optimization and product quality improvement. These two techniques are high‐throughput, capable of analyzing large numbers of samples simultaneously, and exhibit significantly higher sensitivity than traditional culture methods. As a result, they were widely applied in fermented bean paste microbial ecology research over the past decade (Ashaolu and Reale 2020; El Sheikha and Hu 2020). However, their limitations are also evident: PCR‐DGGE and T‐RFLP have limited detection capability for low‐abundance microorganisms and can only reflect relative abundance, failing to provide absolute quantitative data.

2.2.2. High‐Throughput Sequencing and Metagenomic Technologies

In recent years, the rapid advancement of high‐throughput sequencing and metagenomic sequencing has provided revolutionary tools for analyzing complex fermentation systems such as fermented bean paste (Jin et al. 2024). Targeted amplicon sequencing technologies represented by the Illumina platform (e.g., 16S rRNA and ITS sequencing) have become the primary methods for studying microbial communities in fermented bean paste due to their simplicity and relatively low cost (Yao et al. 2022). By specifically amplifying the V3–V4 region of bacterial 16S rRNA genes or fungal ITS regions, this technology rapidly obtains microbial community composition information, revealing clear temporal succession patterns during fermentation (P. Huang et al. 2023). Such high‐resolution dynamic monitoring provides direct evidence for optimizing process parameters such as salinity and temperature. For example, studies have found that the abundance of Lactobacillus increases with fermentation time, with the 40–70 days period identified as the optimal stage for flavor and nutritional value formation (J. Wu et al. 2018). Under high‐salinity conditions, Bacillus, Staphylococcus, and Clostridium emerge as dominant bacterial genera (Ghosh et al. 2019). However, this technology has inherent limitations, such as primer bias‐induced detection deviations and the limited taxonomic resolution of short read lengths, particularly its inability to distinguish metabolically active microorganisms (Stefanini and Cavalieri 2018). In contrast, metagenomic sequencing fundamentally transforms the research paradigm by performing unbiased sequencing of all microbial DNA in a sample. This whole‐genome‐level analysis method not only overcomes the bottleneck of traditional microbiology, which relies on isolation and cultivation‐enabling the identification of over 99% of unculturable microorganisms—but also directly provides functional gene and metabolic pathway information (Ko et al. 2022). For instance, in fermented bean paste fermentation studies, this technology has revealed the functional associations between Rhizopus and Penicillium with carbohydrate metabolism, Aspergillus with protein hydrolysis, and LAB with γ‐aminobutyric acid (GABA) synthesis, confirming the critical roles of these microbial groups in fermentation characteristics (Xie et al. 2019). Additionally, the technology can precisely analyze the functional features of low‐abundance microbial communities and reveal complex microbial interaction networks (Xie et al. 2020). W. Y. Jung et al. (2016) found that during fermented bean paste fermentation, the abundance of fungal communities decreased while bacterial communities proliferated, with significant enrichment of functional genes related to carbohydrate and amino acid metabolism. Their study precisely quantified the dynamic succession patterns of functional microbial communities, thereby establishing a molecular biological foundation for elucidating the relationship between microbial metabolic pathways and the synthesis of flavor compounds.

In summary, the technological framework for analyzing microbial communities in fermented bean paste has undergone several generations of development and refinement (Figure 3). Traditional molecular fingerprinting methods, such as PCR‐DGGE and T‐RFLP, are gradually being phased out of mainstream research due to their limited resolution and inability to comprehensively characterize complex microbial ecosystems. Although conventional culture‐dependent techniques are characterized by low throughput and limited coverage, they remain indispensable as foundational approaches for microbial resource isolation, functional verification, and subsequent industrial application. At present, high‐throughput amplicon sequencing has become a routine method for investigating overall microbial community structure because of its efficiency, convenience, and relatively low cost, whereas metagenomic sequencing further enables in‐depth exploration of microbial functional potential, metabolic pathways, and ecological mechanisms. The complementary integration of these two approaches has therefore become the dominant research paradigm in this field. However, the current technical system still presents several limitations. Metagenomic analysis is associated with high sequencing costs, complex bioinformatics processing, and substantial challenges in data interpretation. In addition, most existing studies mainly focus on taxonomic composition and relative abundance, while insufficient attention has been paid to in situ microbial interactions, spatial distribution patterns, and dynamic regulatory mechanisms during the fermentation process. With the rapid development of emerging technologies such as machine learning, spatial transcriptomics, and multiomics integration, research in this field is gradually overcoming the limitations of traditional empirical approaches and shifting toward data‐driven precise regulation and intelligent process control. In the future, it will be necessary to further integrate multidimensional omics data with intelligent modeling strategies to establish a comprehensive research framework encompassing microbial community analysis, functional validation, process optimization, and safety assurance, thereby promoting the modernization and high‐quality development of the traditional fermented bean paste industry.

FIGURE 3.

Technological evolution of microbial community analysis in fermented bean paste.

3. Multisensory Flavor Characteristics of Fermented Bean Paste and Their Detection Techniques

The formation of fermented bean paste flavor is a complex biochemical process, the core of which lies in the production of nonvolatile and volatile flavor compounds through various metabolic pathways under the action of microorganisms (An et al. 2021). Nonvolatile flavor substances primarily contribute to basic tastes such as umami, sweetness, bitterness, and sourness, while volatile flavor substances determine the aroma characteristics of the food (An et al. 2023). Through synergistic interactions, combined with sensory stimuli from color and texture, these two types of compounds ultimately create a multilayered flavor experience in fermented bean paste. Therefore, a combination of modern analytical techniques and traditional detection methods is essential to systematically investigate the flavor compounds in fermented bean paste.

3.1. Sources and Production of Flavor

In terms of flavor formation, nonvolatile components play a decisive role. These water‐soluble substances diffuse through saliva, bind to receptor cells on taste buds, and ultimately form taste perception through neural signal transduction. As a high‐protein raw material, the degradation of proteins during bean fermentation is particularly critical for flavor development. Proteases secreted by microorganisms first cleave large protein molecules into peptides, which are further hydrolyzed by peptidases into amino acids, forming the foundational flavor of fermented bean paste (Souza et al. 2023). This process not only produces flavor‐active amino acids, but also produces small peptides with unique flavor characteristics (Figure 4). Among them, umami peptides containing glutamic acid or aspartic acid residues can interact with taste receptors on human taste buds, presenting specific taste characteristics. Meanwhile, γ‐glutamyl peptides are recognized by calcium‐sensing receptors, inducing a rich and long‐lasting “kokumi” sensation‐a robust and enduring taste experience (H. J. Kim et al. 2021). More importantly, these flavor peptides can also exhibit synergistic effects with other compounds, such as significantly enhancing umami intensity when combined with nucleotides. γ‐Glutamyl peptides can be produced by microorganisms such as Aspergillus oryzae, which possess γ‐glutamyl transpeptidase activity, through fermentation. Glutathione is another typical bioactive compound with a thick, rich taste, which can be oxidized by Lactobacillus via glutathione reductase. Additionally, these flavor‐active peptides can further complement or enhance the overall taste of fermented bean paste through three synergistic mechanisms: peptide–peptide, peptide–nucleotide, and peptide–cation interactions, making the flavor and mouthfeel more harmonious, mellow, and multilayered (An et al. 2023). Besides protein degradation, microbial enzymes produced through various metabolic pathways during fermentation may also be an important source of umami peptides in fermented foods (An et al. 2023; J. Zhang, Zhao, et al. 2019; Zhuang et al. 2016). Apart from flavor peptides, free amino acids, as end products of protein degradation, exhibit diverse taste contributions. Glutamic acid and aspartic acid impart typical umami, glycine and serine contribute sweetness, while arginine and phenylalanine produce bitterness (An et al. 2023). These amino acids undergo dynamic changes during different fermentation stages, ultimately constituting the base flavor of the product. Multiple studies have shown that the amino acid content in bean products increases during fermentation, particularly in the early stages (Cao et al. 2024). Notably, certain functional amino acids also serve as precursors for volatile flavor compounds. For example, cysteine can be converted into sulfur‐containing compounds, and phenylalanine can give rise to aromatic substances, creating a tightly interconnected network between taste and aroma formation.

FIGURE 4.

The production mechanism of flavor compounds in fermented bean paste. Different microorganisms utilize various nutrients in soybeans and produce diverse flavor compounds through multiple fermentation mechanisms.

Carbohydrate metabolism is another important source of flavor (Vagadia et al. 2017). Soybeans contain approximately 30% carbohydrates, which are gradually degraded into glucose and fructose under the action of amylases and cellulases (Vagadia et al. 2017). Research indicates that strains such as Bacillus subtilis are the primary functional microorganisms in this process. The resulting sugars can directly contribute sweetness or be further metabolized by microorganisms into organic acids (Y. Wang, Wu, et al. 2021). Lactic acid is a representative product that not only provides a refreshing sour taste, but also plays an important role in coordinating the overall flavor (Shukla et al. 2010). In LAB‐dominated fermentation, three metabolic pathways—homofermentative, heterofermentative, and bifid—jointly regulate lactic acid production. More complex organic acids, such as citric acid and malic acid, are linked to the tricarboxylic acid cycle and other metabolic networks, providing precursors for the synthesis of flavor compounds (Sun et al. 2019).

Compared to proteins and carbohydrates, the lipid content in soybeans (approximately 19.9%) is relatively low, but their degradation products should not be overlooked in flavor formation (Vagadia et al. 2017). Lipases hydrolyze triglycerides into free fatty acids, among which unsaturated fatty acids can be further converted into aromatic substances such as aldehydes, ketones, and alcohols (Gamayurova et al. 2021; L. Liu, Chen, et al. 2022). Additionally, functional components like isoflavones and saponins in soybeans may initially contribute bitterness, but they are transformed by microorganisms during fermentation, and this reduction in bitterness is also an important aspect of flavor improvement (An et al. 2023). Overall, the primary metabolites generated by microorganisms through the breakdown of proteins, carbohydrates, and lipids serve as both direct flavor components and precursors for the synthesis of more volatile compounds (L. Liu, Chen, et al. 2022). This cascade of metabolic reactions forms the material basis for the development of fermented flavors.

3.2. Aroma‐Active Compounds and Related Biosynthetic Pathways

Volatile odorants in fermented bean paste are released from the food and reach the nasal cavity through the anterior nostrils or the oral–nasopharyngeal pathway, stimulating olfactory mucosal neurons to produce the sense of smell (Goldberg et al. 2018). The key aromatic compounds in fermented bean paste are typically determined by whether their odor activity value (OAV) exceeds 1 (K. Zhang, Li, et al. 2025). Compounds with an OAV greater than 1 are considered key flavor components that directly influence the overall aroma. In contrast, compounds with an OAV less than 1 indicate that their concentration does not exceed the olfactory detection threshold of the substance, contributing relatively little to the overall aroma and are generally considered background aroma in analysis. With advancements in flavor analysis technology, over 100 aroma‐active compounds have been identified in fermented bean paste, primarily including esters, aldehydes, ketones, alcohols, acids, pyrazines, furans, phenols, terpenes, hydrocarbons, and sulfur‐containing compounds (X. Zhang, Wei, et al. 2021). Among these, esters, aldehydes, ketones, and alcohols are the most abundant (Table 2). The formation of these flavor compounds is closely associated with specific functional microorganisms, key enzymes, and their metabolic pathways.

TABLE 2.

Aroma‐active compounds in traditional fermented bean paste.

Aroma components	CAS	Aroma characteristics	Source	Techniques/instruments	References
(E)‐2‐Nonenal	18829‐56‐6	Soap, fatty	Tempeh	GC–O + GC × GC–ToF–MS	Jeleń et al. (2013)
(E, E)‐2, 4‐Decadienal	65909‐91‐3	Fatty, green, fried	Tempeh	GC–O + GC × GC–ToF–MS	Jeleń et al. (2013)
(E, Z)‐2, 6‐Nonandienal	557‐48‐2	Cucumber	Tempeh	GC–O + GC × GC–ToF–MS	Jeleń et al. (2013)
(Z)‐4‐Heptenal	6728‐31‐0	Rancid	Tempeh	GC–O + GC × GC–ToF–MS	Jeleń et al. (2013)
1‐Methoxy‐4‐propenylbenzene	104‐45‐0	Pungent, green	Douchi	GC–O + GC‐GC	Han et al. (2024)
1‐Octen‐3‐ol	3391‐86‐4	Mushroom‐like, moldy, earthy, lavender, green, fruity, floral	Douchi, Soybean paste	HS–GC–IMS + HS–SPME–GC–MS	Y. Chen et al. (2021), Kumazawa et al. (2013)
1‐Octen‐3‐one	4312‐99‐6	Mushroom‐like	Tempeh	GC–O + GC × GC–ToF–MS	Jeleń et al. (2013)
2‐Decanone	693‐54‐9	Sweet, fruity	Douchi	GC–MS	Lan et al. (2023)
2(5H)‐Furanone	497‐23‐4	Caramel‐like, buttery	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
2,3,5,6‐Tetramethylpyrazine	1124‐11‐4	Musty, burnt, cocoa, sesame oil, fermented soya bean	Douchi, Soybean paste	HS–SPME–GC–MS	Y. Chen, Qin, et al. (2022), J. Li, Sun, et al. (2024), Y. Yang, Yang, et al. (2025)
2,3,5‐Trimethylpyrazine	14667‐55‐1	Roasty, earthy, burnt, roast bean, sesame oil, nutty, cocoa, coffee, popcorn, roasted peanuts, hazelnuts	Douchi, Soybean paste	GC–O–MS/GC–MS	Qiao et al. (2024), Tan et al. (2024), Y. Yang et al. (2026)
2,3‐Butanediol	6982‐25‐8	Creamy, buttery, fruit‐like, flowers, sweet	Soybean paste	GC–MS	K. Zhang, Tang, Li, et al. (2025)
2,3‐Butanedione	431‐03‐8	Cheese‐like, sweet cream, butter, yogurt	Douchi, Miso, Tempeh	GC–O + GC × GC–ToF–MS/GC × GC–TOFMS + UPLC–Q–TOFMS/HS–GC–MS	Jeleń et al. (2013), Sui et al. (2024), S. Wang, Chang, et al. (2021)
2,3‐Diethyl‐5‐methyl pyrazine	18138‐04‐0	Earthy, roasted	Tempeh	GC–O + GC × GC–ToF–MS	Jeleń et al. (2013)
2,3‐Pentanedione	600‐14‐6	Sweet cream, caramel, nuts, buttery	Soybean paste	GC–IMS	B. Yang, Wang, et al. (2025)
2,5‐Dimethylpyrazine	123‐32‐0	Nutty, roasted, earthy, flowers, butter, barley, fried potatoes, cocoa, roasted, coffee	Soybean paste	HS–SPME–GC–MS	J. Li, Sun, et al. (2024)
2,6‐Dimethylpyrazine	108‐50‐9	Roasted nut, cocoa, roast beef, nutty, cooked rice‐like, toast, nut‐like, peanuts, potato, coffee	Douchi, Soybean paste	HS–SPME–GC–MS/GC–O–MS	J. Li, Sun, et al. (2024), Qiao, et al. (2024)
2,6‐Dimethoxyphenol	91‐10‐1	Seasoning‐like, smoky	Soybean paste,	SAFE + SPME	Jo et al. (2011)
2‐Phenyl‐2‐butenal	4411‐89‐6	Fruit, cooked bean, fermented soya bean	Douchi	SPME + SDE + GC–O + GC–MS	Q.‐C. Chen et al. (2011)
2‐Phenylacetic acid	51146‐16‐8	Floral, honey‐like	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
2‐Propenyl hexanoate	123‐68‐2	Fruity	Soybean paste	SPME–GC–MS	X. Zhang, Wei, et al. (2021)
Endo‐borneol	24393‐70‐2	Burnt	Soybean paste	SPME–GC–MS	X. Zhang, Wei, et al. (2021)
2‐Butanol	78‐92‐2	Spoilage	Tempeh	GC–O + GC × GC–ToF–MS	Jeleń et al. (2013)
2‐Heptanone	110‐43‐0	Green, fruity, soap, mild, medicinal, fragrance	Douchi	GC–O + GC–MS	Han et al. (2024)
2‐Methyl‐3 (2‐furyl) acrolein	874‐66‐8	Bake, smoke	Douchi	GC–MS	Lan et al. (2023)
2‐Methyl‐1‐propanol	110597‐10‐9	Wine, solvent‐like, ethereal, cortex, artificial musk, bitter	Soybean paste	HS–GC–IMS	Ren et al. (2025)
2‐Methylpropanal	78‐84‐2	Nutty, malty, chocolate‐like, rancid, pungent	Soybean paste, Tempeh	GC–O + GC × GC–ToF–MS/HS–GC–MS	Jeleń et al. (2013), S. Wang, Chang, et al. (2021)
2‐Methylpropanoic acid	6228‐78‐0	Sweaty, cheesy, pungent, rancid oil	Soybean paste	HS–GC–MS	S. Wang, Chang, et al. (2021)
Ethyl 2‐methylpropanoate	97‐62‐1	Fruity, sweet, rubber, vanilla	Douchi, Soybean paste	HS–GC–IMS + HS–SPME–GC–MS/SAFE + SPME	Y. Chen, et al. (2021), Jo et al. (2011)
2‐Methylbutanol	137‐32‐6	Ripe onion, rancid, nail polish, wine, alcoholic, malty	Douchi, Soybean paste, Tempeh	GC–O + GC × GC–ToF–MS/HS‐GC–MS	Jeleń et al. (2013), S. Wang, Chang, et al. (2021)
2‐Eethylbutanal	96‐17‐3	Cocoa, almond, cooked bean, malty, nuts, flowers	Douchi	GC–O + GC × GC–ToF–MS	Jeleń et al. (2013)
Methyl 2‐methylbutanoate	868‐57‐5	Fruity, sweet	Soybean paste	SAFE + SPME	Jo et al. (2011)
Ethyl 2‐methylbutyrate	7452‐79‐1	Apple, fruity, fresh, sweet, pineapple skin, plum skin	Douchi, Soybean paste	MDGC−MS/O/GC × GC–TQMS	Y. Liu et al. (2024), Lu, et al. (2021)
2‐Formylpyrrole	1003‐29‐8	Coffee‐like	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
2‐Carene	554‐61‐0	Rice	Douchi	HS–GC–IMS + HS–SPME–GC–MS	Y.‐Z. Zhang, Lin, et al. (2022)
2‐Furancarboxylicacid, ethyl ester	614‐99‐3	Nuts, sweet	Douchi	GC–MS	Lan et al. (2023)
2‐Pentyl furan	3777‐69‐3	Green bean, butter, sweet, spicy, fruity, vegetables	Douchi, Soybean paste	GC–MS	Q.‐C. Chen et al. (2011), X. Li et al. (2017)
2‐Ethyl hexenal	645‐62‐5	Rice	Douchi	HS–GC–IMS + HS–SPME–GC–MS	Y.‐Z. Zhang, Lin, et al. (2022)
2‐Ethyl‐3, 5‐dimethylpyrazine	55031‐15‐7	Roasty, earthy, potato‐like, toast, nut‐like	Tempeh	GC–O + GC × GC–ToF–MS	Jeleń et al. (2013)
2‐Ethyl‐6‐methylpyrazine	13925‐03‐6	Roasted, toast‐like, cooked potato	Miso	GC–MS	Mori et al. (1983)
2‐Acetyl‐1‐pyrrolidine	85213‐22‐5	Roasted, popcorn	Douchi	GC–O + GC × GC–ToF–MS	Jeleń et al. (2013)
2‐Acetylpyrrole	1072‐83‐9	Herbal, medicine, nutty, walnut, bread, bakery, Caramel, syrup	Douchi	GC–O + GC–MS	Han et al. (2025)
Acetylpyrazine	22047‐25‐2	Popcorn	Douchi	GC–O + GC–MS	S. Wang, Chang, et al. (2021)
2‐Acetylfuran	1192‐62‐7	Cocoa, coffee, Caramel, sweet, Almonds, nuts, milk	Douchi	GC–O + GC–MS	Han et al. (2025)
3,4‐Dimethylbenzaldehyde	5973‐71‐7	Almond	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
3‐Methyl‐1, 2‐cyclopentanedione	765‐70‐8	Caramel	Douchi	GC–O + GC–MS	S. Wang, Liu, et al. (2021)
2‐Hydroxy‐3‐methyl‐2‐cyclopenten‐1‐one	3400‐78‐0	Caramel‐like	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
3‐Methylphenol	155174‐22‐4	Phenolic, animal	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
3‐Methylbutanol	123‐51‐3	Malty, Rancid, nail polish, malty, burnt, chocolate‐like, Wines, ether, Fusel, fruit, banana, alcoholic, almond	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
3‐Methylbutanal	590‐86‐3	Fresh, cocoa, malty	Douchi, Soybean paste, Tempeh	GC–O + GC × GC–ToF–MS/HS–SPME–GC—MS	Jeleń et al. (2013), J. Li, Sun, et al. (2024), Y. Yang et al. (2026)
3‐Methylbutanoic acid	503‐74‐2	Old sock, rancid, fruit, pungent, cheese, sweaty, sour	Douchi, Soybean paste	GC–O + GC × GC–ToF–MS/HS–SPME–GC–MS	Jeleń et al. (2013), J. Li, Sun, et al. (2024), Y. Yang et al. (2026)
Ethyl 3‐methylbutanoate	108‐64‐5	Fruity, apple, bananas, sweet	Douchi, Soybean paste	SAFE + SPME/GC–IMS + GC–O–QTOF/MS	Jo et al. (2011), Miao et al. (2024)
3‐Methylvaleric acid	105‐43‐1	Sweaty, sour, herbal, grass	Douchi	GC–O + GC–MS	S. Wang, Liu, et al. (2021)
Methional	3268‐49‐3	Cooked potato‐like, meaty, sauce	Douchi, Soybean paste, Tempeh	GC–O + GC × GC–ToF–MS/GC–MS/HS–SPME–GC–MS/GC–MS	Jeleń et al. (2013), Kumazawa et al. (2013), A. Li, Liu, et al. (2024), A. Li, Yang, et al. (2024)
Acetoin	513‐86‐0	Creamy, milky, buttery	Soybean paste	GC–MS	Nacchio et al. (2022)
3‐Penten‐2‐one	625‐33‐2	Fruit, spicy	Soybean paste	GC–IMS	B. Yang, Wang, et al. (2025)
3‐Octanol	589‐98‐0	Mushroom‐ like, herbaceous, melon, citrus‐like odor, milk	Douchi	GC–MS	Lan et al. (2023)
3‐Octanone	106‐68‐3	Herb, legumina, resin, mushroom, earthy, metal, butter, keto, cheese	Douchi	HS–GC–IMS + HS–SPME–GC–MS	Y Chen, et al. (2021)
Pyrazine, 3‐ethyl‐2, 5‐dimethyl‐	13360‐65‐1	Potato, roast, barley, coffee, peanut, hazelnut	Soybean paste	GC–MS	이다나, 이경민, 이성은, and 김태오 (2024)
4‐Methyl‐2‐phenyl‐2‐pentenal	26643‐91‐4	Cocoa, cooked bean	Douchi	MDGC−MS/O	Y. Liu, et al. (2024)
4‐Methylphenol	2876‐02‐0	Phenolic, animal	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
4‐Methylvaleric acid	646‐07‐1	Sweaty, sour, Spicy, cheese	Douchi	GC–O + GC–MS	S. Wang, Liu, et al. (2021)
Ethyl 4‐methylpentanoate	25415‐67‐2	Fruity	Soybean paste	SAFE + SPME	Jo et al. (2011)
4‐Methylguaiacol	93‐51‐6	Smoky, clove	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
4‐Hydroxy‐2, 5‐dimethyl‐3(2H)‐furanone	3658‐77‐3	Caramel, sweet	Soybean paste	GC–MS + GC–O	J. Zhao et al. (2011)
4‐Ethylphenol	123‐07‐9	Smoky, Phenolic, spice, wood	Soybean paste	SAFE + SPME	Jo et al. (2011)
4‐Ethylguaiacol	2785‐89‐9	Caramel‐like, smoky, spicy, bacon	Soybean paste	GC–MS	M. Wang, et al. (2025)
4‐Vinylguaiacol	7786‐61‐0	Soy sauce‐like, clove‐like, smoky, woody, sweet	Soybean paste	GC–MS	M. Wang, et al. (2025)
5‐Methyl‐2‐phenyl‐2‐hexenal	21834‐92‐4	Nutty, cocoa	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
5‐Methylfurfural	620‐02‐0	Almond, sweet, spicy, roasted, green, thick	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
D‐Limonene	5989‐27‐5	Citrus‐like	Soybean paste	GC × GC–TQMS	Lu, et al. (2021)
α‐Pinene	80‐56‐8	Fresh, flavor	Douchi	HS–GC–IMS + HS–SPME–GC–MS	Y.‐Z. Zhang, Lin, et al. (2022)
α‐Terpinene	99‐86‐5	Oily	Douchi	HS–GC–IMS + HS–SPME–GC–MS	Y.‐Z. Zhang, Lin, et al. (2022)
β‐Damascenone	23726‐93‐4	Flora, honey	Douchi	GC–O + GC–MS	S. Wang, Liu, et al. (2021)
β‐Pinene	127‐91‐3	Spicy, mushroom	Douchi	HS–GC–IMS + HS–SPME–GC–MS	Y.‐Z. Zhang, Lin, et al. (2022)
β‐Phellandrene	555‐10‐2	Ginger	Douchi	HS–GC–IMS + HS–SPME–GC–MS	Y.‐Z. Zhang, Lin, et al. (2022)
γ‐Butyrolactone	96‐48‐0	Creamy, caramel‐like	Soybean paste	HS–GC–MS	J. Cai et al. (2022)
γ‐Nonanolactone	104‐61‐0	Coconut‐like, sweet, peach, fatty, cheese	Douchi, Soybean paste	GC–O + GC–MS/GC–O + GC–TOF–MS	S. Wang, Liu, et al. (2021), C. Zhao et al. (2021)
γ‐Terpinene	99‐85‐4	Oily	Douchi	HS–GC–IMS + HS–SPME–GC–MS	Y.‐Z. Zhang, Lin, et al. (2022)
γ‐Valerolactone	108‐29‐2	Sweet, milky, popcorn, fruity	Douchi, Soybean paste	GC–O + GC–MS/GC–O + GC–TOF–MS	S. Wang, Liu, et al. (2021), C. Zhao et al. (2021)
Eucalyptol	470‐82‐6	Woody, herbal, spicy, mint, sweet	Soybean paste	GC × GC–TQMS	Lu, et al. (2021)
Phenol	108‐95‐2	Phenolic, animal, rubber‐like, woody	Soybean paste	HS–SPME–GC–MS/GC–O + GC–TOF–MS	J.‐K. Kim, Seo, Chang, and Lee (1993), C. Zhao et al. (2021)
Benzaldehyde	100‐52‐7	Sweaty, fruity, burnt sugar, nutty, soybean like, bitter almonds, woody, cherry, nuts	Douchi, Soybean paste	HS–SPME–GC–MS	A. Li, Liu, et al. (2024), J. Li, Sun, et al. (2024)
Benzoic acid	65‐85‐0	Floral	Soybean paste	GC–MS	S.‐D. Yi, Yang, Jeong, Sung, and Oh (1999)
Ethyl benzoate	93‐89‐0	Caramel, sweaty, cheese, camomile, flower, celery, fruit, medicine	Soybean paste	SAFE + SPME	Jo et al. (2011)
Phenethyl alcohol	60‐12‐8	Honey, rosy, spice, lilac, fresh, floral, sweet	Douchi, Soybean paste	GC–MS	Q.‐C. Chen, et al. (2011), K. Zhang, Tang, Li, et al. (2025)
Phenyl acetaldehyde	122‐78‐1	Flora, rose, honey, sweet, Hawthorne, green, honeyowy, floral, narcissus, wind letter, cherry	Douchi, Soybean paste	GC–IMS/HS–SPME–GC–MS + GC–O–MS	H. Ling et al. (2022), Z. Wu, et al. (2024)
Phenylacetic acid	103‐82‐2	Floral, honey	Douchi	GC–MS	Tu et al. (2025)
Ethyl phenylacetate	101‐97‐3	Fruit, sweet, honey, rose, floral	Douchi, Soybean paste	GC–MS	Tu et al. (2025), K. Zhang, Tang, Li, et al. (2025)
Acetophenone	98‐86‐2	Aromatic, flower, musty, cheese‐like, roses	Douchi, Soybean paste	GC–MS/GC–IMS + GC–MS	A. Li, Feng, et al. (2024), K. Zhang, Tang, Li, et al. (2025)
Benzyl alcohol	100‐51‐6	Sweet, caramel, floral, almond	Douchi, Soybean paste	GC–O + GC–MS/SPME–GC–MS/HS–SPME–GC–MS	Han et al. (2025), J. Li, Sun, et al. (2024), X. Zhang, Wei, et al. (2021)
Propanoic acid	6212‐69‐7	Sour, cheesy	Soybean paste	GC–MS	Shukla et al. (2010)
Diethyl succinate	123‐25‐1	Fruity, green, bean, wine	Soybean paste	SAFE + SPME	Jo et al. (2011)
1,1‐Diethoxybutane	3658‐95‐5	Flower, sweetness	Douchi	HS–GC–IMS + HS–SPME–GC–MS	Y.‐Z. Zhang, Lin, et al. (2022)
Butanoic acid	107‐92‐6	Sweaty, sour, cheese‐like, milk, cream, butter, fruity	Soybean paste	GC–MS	Shukla et al. (2010)
Phenethyl butyrate	103‐52‐6	Fruit, sweet, honey	Douchi	GC–MS	Tu, et al. (2025)
Methyl butanoate	623‐42‐7	Fruity, banana‐like	Soybean paste	SPME–GC–MS	X. Zhang, Wei, et al. (2021)
Ethyl butanoate	105‐54‐4	Fruity, sweet, apple, fresh	Douchi Soybean paste,	SPME–GC–MS/GC–MS	Tu et al. (2025), X. Zhang, Wei, et al. (2021)
Eugenol	97‐53‐0	Clove‐like, woody, spicy	Douchi, Soybean paste	GC–O + GC–MS/GC–O + GC–TOF–MS	S. Wang, Liu, et al. (2021), C. Zhao et al. (2021)
p‐Cresol	106‐44‐5	Burnt	Douchi	GC–O + GC–MS	S. Wang, Liu, et al. (2021)
Dimethyl disulfide	624‐92‐0	Burnt, cooked onion‐like, cabbage	Soybean paste	GC–O/SPME–GC–MS	Inoue et al. (2016), X. Zhang, Wei, et al. (2021)
Dimethyl sulfide	75‐18‐3	Cabbage	Tempeh	GC–O + GC × GC–ToF–MS	Jeleń et al. (2013)
Dimethyl trisulfide	3658‐80‐8	Garlic, sulfur, cabbage, onion	Douchi, Tempeh	GC–O + GC × GC–ToF–MS/GC–O + GC–MS	Jeleń et al. (2013), Han, et al. (2025)
Pantolactone	79‐50‐5	Caramel‐like	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
Linalool	78‐70‐6	Flowery, sweaty	Soybean paste	GC × GC–TQMS	Lu et al. (2021)
Heptanal	111‐71‐7	Fruity, green, rose, perilla	Soybean paste	SPME–GC–MS	Deng et al. (2022), X. Zhang, Wei, et al. (2021)
Heptanoic acid	111‐14‐8	Sweaty	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
Methyl heptanoate	106‐73‐0	Sweet, honey‐like	Soybean paste	SPME–GC–MS	X. Zhang, Wei, et al. (2021)
Ethyl heptanoate	106‐30‐9	Pineapple, fruit	Douchi	HTS + HS–SPME–GC–MS	Luo et al. (2025)
Decanal	112‐31‐2	Fruit, green, apple, orange peel, flowery, wax	Douchi,	GC–O + GC–MS	S. Wang, Liu, et al. (2021)
Decanoic acid	334‐48‐5	Soapy, musty	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
Ethyl caprate	110‐38‐3	Grape	Soybean paste	GC × GC–TQMS	Lu et al. (2021)
4,5‐Dimethyl‐3‐hydroxy‐2, 5‐dihydrofuran‐2‐one	28664‐35‐9	Roasted, seasoning‐like, caramel‐like	Douchi, Soybean paste	GC–MS + GC–O/GC–O + GC–TOF–MS	J. Zhang et al. (2024), C. Zhao et al. (2021)
Anethole	104‐46‐1	Anise‐like, sweet, spicy, fennel, licorice smell	Soybean paste	SPME–GC–MS	S. Zhao, Sai, et al. (2023)
Hexanol	111‐27‐3	Alcoholic, green, fatty, sweet, herb‐like, fruit, flowery	Douchi, Soybean paste	HS–GC–IMS + HS–SPME–GC–MS/GC–MS	Y. Chen et al. (2021), Nacchio et al. (2022)
Hexanal	66‐25‐1	Grass, fat, fishy, green, tea leaf‐like, aldehyde	Douchi, Tempeh	HS–GC–IMS + HS–SPME–GC–MS/GC–O + GC × GC–ToF–MS	Y. Chen et al. (2021), Jeleń et al. (2013)
Hexanoic acid	142‐62‐1	Cheesy, fatty, sweaty, sour	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
Methyl hexanoate	106‐70‐7	Flowery	Soybean paste	SPME–GC–MS	X. Zhang, Wei, et al. (2021)
Ethyl hexanoate	123‐66‐0	Pine, apple peel, fruit, alcohol, pineapple, banana	Douchi	HTS + HS–SPME–GC–MS	Luo et al. (2025)
m‐Cresol	108‐39‐4	Roasted	Douchi	GC–O + GC–MS	Han et al. (2024)
Camphene	79‐92‐5	Alcoholic	Douchi	HS–GC–IMS + HS–SPME–GC–MS	Y.‐Z. Zhang, Lin, et al. (2022)
2‐Furanmethanol	98‐00‐0	Burnt sugar, coffee, nutty, roasty, flower like, bitter, spicy	Soybean paste	GC–MS	K. Li, Zhang, et al. (2022)
Furfural	98‐01‐1	Almond, roasted bread, sweet, popcorn‐like, burnt, baking‐like, woody	Soybean paste	HS–SPME–GC–MS	J. Li, Sun, et al. (2024)
Maltol	118‐71‐8	Sweet, malty, caramel‐like, butter, sugar, strawberry, fresh	Soybean paste	HS–SPME–GC–MS	Hong et al. (2012)
Nonanal	124‐19‐6	Frowzy, citrus, soapy, fat, green, vinegar, flowery	Douchi	HS–GC–MS	Sui et al. (2024)
Nonanoic acid	112‐05‐0	Fats, wax, cheese, coconut	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
Ethyl nonanoate	123‐29‐5	Winey	Douchi	HS–GC–IMS + HS–SPME–GC–MS	Y.‐Z. Zhang, Lin, et al. (2022)
Ethyl lactate	97‐64‐3	Sweet, fruit, roasted, old rum	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
Ethyl tetradecanoate	124‐06‐1	Ether	Soybean paste	GC × GC–TQMS	Lu et al. (2021)
Pentanol	71‐41‐0	Fruity, balsamic	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
Ethyl valerate	539‐82‐2	Alcoholic, floral, yeast, fruit	Douchi	HS–GC–IMS + HS–SPME–GC–MS	Y.‐Z. Zhang, Lin, et al. (2022)
Citronellol	106‐22‐9	Flora	Douchi	GC–O + GC–MS	S. Wang, Liu, et al. (2021)
Octanol	111‐87‐5	Nutty, creamy, citrus, floral	Soybean paste	GC–MS	Kumazawa et al. (2013)
Octanal	124‐13‐0	Crude oil, pungent, fatty, green, fruits, citrus	Soybean paste	SPME–GC–MS	X. Zhang, Wei, et al. (2021)
Octanoic acid	124‐07‐2	Fatty, cheesy, sweat, pungency	Douchi, Soybean paste	GC–O + GC–MS	Han et al. (2025), Z. Tian, Ameer, et al. (2022)
Methyl octanoate	111‐11‐5	Fruity, sweet, orange	Soybean paste	SPME–GC–MS	X. Zhang, Wei, et al. (2021)
Ethyl caprylate	106‐32‐1	Banana, fruity, sweet, fat, floral, alcohol, wax, milk, cream, fruit, wine	Douchi, Soybean paste	SPME–GC–MS/HS–GC–IMS + HS–SPME–GC–MS	X. Zhang, Wei, et al. (2021), Y.‐Z. Zhang, Lin, et al. (2022)
Linolenic acid	463‐40‐1		Soybean paste	GC × GC–TQMS	Lu et al. (2021)
Ethyl linoleate	544‐35‐4	Fruity, sweet	Douchi, Soybean paste	GC × GC–TQMS	Lu et al. (2021)
Ethanol	64‐17‐5	Alcoholic, solvent‐like, pungent, ether	Soybean paste	GC–MS	Y. G. Kim et al. (2015)
Acetic acid	64‐19‐7	Rancid, sour, vinegar like	Douchi, Soybean paste	GC–MS	A. Li, Yang, et al. (2024), Shukla et al. (2010)
Phenethyl acetate	103‐45‐7	Rose, honey, sweaty, cheese, flowery	Douchi, Soybean paste	GC–MS	Q.‐C. Chen et al. (2011), S.‐H. Yi and Hong (2021)
n‐Butyl acetate	123‐86‐4	Pear, banana	Soybean paste	GC–MS + GC–O	Y. Zhang et al. (2010)
Ethyl acetate	141‐78‐6	Fruity, pineapple, sweet	Douchi, Soybean paste	GC × GC–TQMS/GC–MS	Lu et al. (2021), Tu, et al. (2025)
Isoamyl acetate	123‐92‐2	Fruit, sweet, fragrant	Douchi	GC–IMS + GC–O‐QTOF/MS	Miao et al. (2024)
Isoeugenol	97‐54‐1	Floral, clove‐like, malty, chocolate‐like	Soybean paste	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
Terpinolene	586‐62‐9	Sharp, lemon	Douchi	HS–GC–IMS + HS–SPME–GC–MS	Y.‐Z. Zhang, Lin, et al. (2022)
2‐Methylbutyl isovalerate	2445‐77‐4	Nuts, malt	Douchi	GC–MS	Lan et al. (2023)
Indole	120‐72‐9	Flowery	Soybean paste	SPME–GC–MS	X. Zhang, Wei, et al. (2021)
Guaiacol	90‐05‐1	Woody, sweet, burnt, smoky	Douchi, Soybean paste, Tempeh	GC–O + GC–TOF–MS	C. Zhao et al. (2021)
Myrcene	123‐35‐3	Fruit aroma, floral aroma, green aroma	Soybean paste	GC × GC–TQMS	Lu et al. (2021)
Methyl palmitate	112‐39‐0	Waxy	Soybean paste	GC × GC–TQMS	Lu et al. (2021)
Ethyl palmitate	628‐97‐7	Waxy, butter	Soybean paste	GC × GC–TQMS	Lu et al. (2021)

3.2.1. Acids

Acid compounds, as important flavor substances in fermented bean paste, can be broadly categorized into small‐molecule acids, branched‐chain acids, and long‐chain acids. Small‐molecule acids such as acetic acid, propionic acid, and butyric acid are mainly produced by microbial metabolism (Y. Chen et al. 2021). Acetic acid is a typical metabolite of LAB, with precursors possibly being carbohydrates or fatty acids. Propionic acid may originate from microbial hydrolysis of lipids, while butyric acid can be generated through the metabolism of LAB. Branched‐chain acids such as 2‐methylpropionic acid, 3‐methylbutyric acid, and phenylacetic acid are closely related to amino acid degradation (L. Zhang, Huang, et al. 2021). Their precursors—valine, leucine, and phenylalanine, respectively—are converted via transamination, decarboxylation, and oxidation reactions under microbial action. Research has found that the phosphotransbutyrylase secreted by Bacillus subtilis can convert isovaleryl‐CoA and isobutyryl‐CoA into 3‐methylbutyric acid and 2‐methylpropanoic acid, respectively (X. Chen, Lu, et al. 2022). Long‐chain acids with more than six carbon atoms primarily derive from lipid degradation. The accumulation of these acids occurs through diverse pathways, including alcohol oxidation, ester hydrolysis, or direct microbial synthesis. In flavor formation, different acids play unique roles: acetic acid imparts a sharp, sour aroma, branched‐chain acids contribute cheesy and rich notes, while long‐chain acids are associated with fatty aromas. Additionally, these acids can serve as precursors for esterification reactions, forming a wider variety of flavor compounds.

3.2.2. Alcohols

Alcohols provide sweet and pleasant aromas, and their formation primarily relies on microbial metabolism of sugars, amino acids, and fatty acids (Lan et al. 2023). These alcohols can be classified into those directly derived from raw materials (e.g., α‐phenylethanol and n‐butanol) and those synthesized during fermentation using glucose and amino acids as precursors (e.g., ethanol and higher alcohols). In terms of biosynthetic pathways, ethanol is mainly produced via glycolysis, where yeast and LAB convert sugars into pyruvate, followed by decarboxylation and reduction. Branched‐chain alcohols possess dual biosynthetic pathways. They can be produced from carbohydrate metabolism via the glycolytic pathway and can also be transformed from amino acid degradation through the Ehrlich pathway (An, et al. 2023; Song et al. 2026). For instance, isoleucine is converted into 2‐methylbutanol, and leucine yields 3‐methylbutanol. Higher alcohols are primarily generated through the degradation of amino acids via the Ehrlich pathway. Specifically, aromatic amino acids are metabolized to produce aromatic alcohols such as phenylethanol, while branched‐chain amino acids are converted into branched‐chain alcohols (J. Huang et al. 2021). Alternatively, they can be formed by the reduction and decarboxylation of corresponding α‐keto acids through the Harris pathway (L. Chen et al. 2025). As a typical aromatic alcohol, phenylethanol is derived from phenylalanine via the intermediates phenylacetaldehyde or phenylpyruvic acid under the action of Saccharomyces cerevisiae. In addition, fatty acid metabolism also serves as an important source of alcohol compounds. 1‐Hexanol is generated through the auto‐oxidation process of ω‐6 fatty acids, whereas 1‐octen‐3‐ol, which has a characteristic mushroom aroma, originates from the conversion of linoleic acid 10‐hydroperoxide under the action of hydroperoxide lyase (R.‐Q. He et al. 2020). Although individual alcohol compounds have relatively high flavor thresholds and thus make limited direct contributions to the overall flavor, the synergistic effect of multiple alcohol substances can endow fermented soybean products with unique flavor characteristics. Studies have shown that these substances are predominantly produced in large quantities during the initial stage of fermentation. Apart from being lost through volatilization during the production process, they also participate in the synthesis of important flavor compounds such as esters, ultimately exerting a significant influence on the flavor quality of the products (Shahidi and Hossain 2022).

3.2.3. Esters

Ester compounds are key components that determine the aroma characteristics of products due to their high volatility and high sensitivity to olfactory receptors (Tian et al. 2023). During the fermentation process, yeast, as the dominant microorganism in ester synthesis, catalyzes the formation of esters through a multipathway metabolic network. During this process, fatty acids are first activated by coenzyme A (CoA) to form acyl‐CoA, which then combines with alcohols such as ethanol, isobutanol, and isopentanol under the catalysis of alcohol acyltransferases to produce the corresponding ethyl esters (Que et al. 2023). Among them, Saccharomyces cerevisiae carrying the ATF1 and ATF2 genes (both encoding alcohol acyltransferases) plays a key regulatory role in acetate esters synthesis, and can catalyze the production of a variety of characteristic acetate esters, including fruity‐smelling ethyl acetate and phenylethyl acetate with a rose‐like aroma (Sumby et al. 2010). In addition, ethanol hexanoyltransferases encoded by the EHT1 and EEB1 genes can mediate the binding reaction between ethanol and medium‐ to long‐chain fatty acids, generating flavor compounds such as butyl acetate with a pineapple aroma or ethyl hexanoate with an apple‐like sweet aroma (Sumby et al. 2010). Besides the classic esterification pathway, alternative routes such as alcoholysis, acidolysis, and transesterification also participate in ester synthesis, a process that relies on the catalytic action of key enzymes such as triacylglycerol lipase and carboxylesterase. Notably, lipases secreted by fungi can directly act on soybean lipids, catalyzing the formation of high‐molecular‐weight fatty acid esters such as ethyl palmitate. The composition and content levels of these ester compounds are finely regulated by multiple factors including fermentation temperature, microbial community structure, and oxygen supply (Pu et al. 2023). For example, under constant‐temperature and long‐term fermentation conditions, fungal lipases can efficiently catalyze the conversion of long‐chain fatty acids to form substances such as ethyl laurate, ultimately endowing fermented bean paste with a rich and layered floral‐fruity aroma and a creamy flavor (An et al. 2023).

3.2.4. Aldehydes

Aldehyde compounds are important contributors to the nutty and fruity aromas in fermented bean paste, mainly formed through pathways such as lipid oxidation, amino acid degradation, and microbial metabolism. During microbial metabolism, aminotransferases convert free amino acids into α‐keto acids, which are further degraded into corresponding aldehyde compounds (Tian et al. 2023). Endogenous enzymes in adipose tissue generate fatty aldehydes such as butanal, hexanal, octanal, nonanal, and decanal through the oxidation of unsaturated fatty acids. For example, lipoxygenase catalyzes the conversion of linoleic acid to hexanal, while oleic acid may be degraded into nonanal and decanal. The auto‐oxidation of ω‐6 fatty acids such as linoleic acid and arachidonic acid produces (E, E)‐2,4‐decadienal and (E,E)‐2,4‐nonadienal (Feng et al. 2024). Aldehydes such as 2‐methylpropanal, 2‐methylbutanal, 3‐methylbutanal, benzaldehyde, and phenylethyl aldehyde are converted from amino acids such as isoleucine, leucine, and phenylalanine through the Strecker degradation pathway and can also be generated through amino acid catabolism via the Ehrlich pathway (Jeleń et al. 2013). In addition, these aldehydes can serve as intermediate products and be further oxidized or reduced into other flavor compounds such as acids and alcohols. Their corresponding ketones produce fruity or caramel aromas through microbial‐induced oxidation of lipids and amino acids and the Maillard reaction.

3.2.5. Ketones

Ketone compounds in fermented bean paste are mainly produced through pathways such as the Maillard reaction, amino acid thermal degradation, and lipid oxidation (W.‐C. Wang et al. 2023). Among these, 2,3‐butanedione is one of the most common ketone compounds, with diverse biosynthetic pathways. During fermentation, Bacillus subtilis metabolizes sugars to produce α‐acetolactate, which is decarboxylated by the extracellular enzyme α‐acetolactate decarboxylase to generate acetylacetone (Y. Liu et al. 2018). Yeast also synthesizes α‐acetolactate first, which then undergoes decomposition to generate 2,3‐butanedione (J. Huang et al. 2021). Additionally, Bacillus licheniformis produces oxidase through its own metabolism to catalyze the conversion of 2,3‐butanediol to 2,3‐butanedione (Y. Liu et al. 2018). Furthermore, free fatty acids released from lipids may also contribute to the formation of 2,3‐butanedione through catalysis by microbial enzymes (X. Wang, Guo, et al. 2021). Besides 2,3‐butanedione, other ketone compounds are formed via distinctive pathways. For example, 2,3‐pentanedione is formed by the interaction between glucose degradation products and L‐alanine. 2‐Heptanone can be obtained through the Maillard reaction or microbial conversion of fatty acids. 1‐Octen‐3‐one is derived from the auto‐oxidation of unsaturated fatty acids, while acetophenone is produced through the biosynthetic reaction of lipids and amino acids catalyzed by fungal enzymes or the Maillard reaction (Han et al. 2024).

3.2.6. Phenols

Phenolic substances are an important class of functional flavor components in fermented bean paste, not only endowing the product with unique flavor characteristics but also exhibiting significant physiological activities such as antioxidant and antibacterial properties. Among them, methoxyphenolic substances such as 2‐methoxyphenol, 4‐vinylguaiacol, and 4‐ethylguaiacol can impart unique smoky and spicy flavors to soybean products. The formation of these compounds begins with the degradation of bound phenolic acids in soybean raw materials (Lan et al. 2023). Under the action of laccase secreted by Aspergillus oryzae and other fungi, lignin is decomposed into phenolic precursors such as ferulic acid, coumaric acid, and 4‐hydroxycinnamic acid, which are then further converted through the decarboxylation and reduction reactions of yeasts (such as C. etchellsii, C. versatilis, Torulopsis; Pu et al. 2023). In addition, previous studies have shown that substances such as 4‐ethyl‐2‐methoxyphenol detected during the fermentation process may be produced through the degradation of hydroxycinnamic acid and hydroxybenzoic acid, and yeast carbohydrate metabolism and decarboxylation reactions are also considered key links in their synthesis (Pu et al. 2023).

3.2.7. Others

Pyrazines, as nitrogen‐containing heterocyclic compounds, mainly impart baking and roasting aromas (Peng et al. 2014). Their formation not only depends on the Maillard reaction but is also closely related to microbial enzymatic reactions. For example, although Bacillus subtilis does not directly synthesize pyrazines, free amino acids, α‐acetolactate, and α‐dicarbonyl compounds and other small molecules produced during the fermentation process can promote their formation (J. Huang et al. 2021). The biochemical synthesis pathway of microorganisms is further demonstrated by the fact that glucose is biotransformed to generate acetoin as a precursor, which is then further synthesized into tetramethylpyrazine (J. Huang et al. 2021). Furans, as oxygen‐containing heterocyclic compounds, have more diverse sources. They can be generated through sugar degradation or the Maillard reaction, and can also be produced by microbial metabolism. 2‐Furanmethanol is derived from the thermal degradation of ribose, while 2‐pentylfuran is formed by the metabolism of staphylococci or the oxidation of linoleic acid hydroperoxides (R.‐Q. He et al. 2020). In addition, pentose degradation can generate 4‐hydroxy‐5‐methyl‐3(2H)‐furanone, and L‐threonine, as a precursor, generates 4,5‐dimethyl‐3‐hydroxy‐2,5‐dihydrofuran‐2‐one through a condensation reaction. Yeast decarboxylation is related to the biosynthesis of 4‐hydroxy‐2,5‐dimethyl‐3(2H)‐furanone and 2‐ethyl‐4‐hydroxy‐5‐methylfuran‐3‐one (X. Wang, Guo, et al. 2021). These compounds are not only produced by carbohydrate degradation but can also be generated in large quantities through yeast metabolism of D‐xylulose‐5‐phosphate in soybeans. Thermal reaction products such as 3‐hydroxy‐2‐methyl‐4‐pyranone (maltol) further enrich the layers of caramel aroma, collectively forming the characteristic “soy sauce‐like” flavor basis of fermented bean paste (X. Wang, Guo, et al. 2021).

Sulfur‐containing compounds mainly originate from the degradation process of sulfur‐containing amino acids (C.‐J. Chen et al. 2023). During this process, microbial lyases directly degrade methionine into methanethiol, which can then form dimethyl trisulfide with a garlic‐like odor through spontaneous oxidation reactions. The oxidation of methanethiol can also generate dimethyl trisulfide. Although sulfur compounds are present at low concentrations in the fermentation system, due to their extremely low sensory thresholds, they can still significantly affect the flavor characteristics of the product. For example, methional not only contributes to the aroma of boiled potatoes but also has a flavor‐enhancing effect (X. Wang, Guo, et al. 2021). These thiol compounds are generated through two main pathways: the Strecker degradation and microbial metabolism, ultimately forming a complex and diverse network of sulfur‐containing volatile compounds (J. Huang et al. 2021).

3.3. Key Technologies for the Detection of Aroma Compounds in Fermented Bean Paste

In the field of flavor detection, technologies such as electronic nose/electronic tongue, GC–MS, GC–IMS, and gas chromatography–olfactometry (GC–O) each possess unique characteristics, with significant differences in their sensitivity and application scenarios (J. Ling and Heldman 2025). As representatives of bionic sensing technologies, the electronic nose and electronic tongue utilize sensor arrays to achieve broad‐spectrum responses to volatile organic compounds and taste substances, enabling rapid screening and dynamic monitoring of flavors (C. Kim, Lee, et al. 2022). Their sensitivity allows for the effective capture of concentration change thresholds of key flavor substances (such as esters and alcohols) during the fermentation process, particularly excelling in distinguishing the flavor characteristics of fermented bean paste produced by different processes or from different regions (Yu et al. 2023). The electronic nose can detect volatile compounds with a lower limit of detection at the ppb level, while the electronic tongue's sensitivity to taste substances approaches the human taste threshold. However, limited by the fixed types of sensors (such as metal oxide or conductive polymer sensors), they exhibit weak responses to certain trace heterocyclic compounds or nonpolar substances (Mahanti et al. 2024). Additionally, samples need to be preprocessed into homogeneous liquids, restricting their direct application in solid‐state fermented foods. In contrast, thanks to the high separation efficiency of capillary columns and the high specificity of mass spectrometry, GC–MS offers higher detection sensitivity and enables simultaneous qualitative and quantitative analysis of over 300 volatile compounds (Stashenko and Martínez 2014). GC–MS combines the separation capability of gas chromatography with the detection and identification power of mass spectrometry and is commonly used for separating and identifying volatile and semivolatile organic compounds in complex sample mixtures. Its principle is primarily based on the efficient separation by gas chromatography, coupled with a highly sensitive and specific mass spectrometry detector, enabling precise identification of various volatile substances (Xu et al. 2021). This facilitates the qualitative and quantitative analysis of components in fermented foods such as bean paste. S. Zhao, Sai, et al. (2023) employed GC–MS to analyze 71 flavor compounds in Chinese and Korean fermented bean pastes, including categories such as alcohols, aldehydes, ketones, acids, esters, hydrocarbons, phenols, heterocycles, and benzene derivatives. The limitations of this technology lie in its applicability only to thermally stable volatile components and the time‐consuming pretreatment steps, such as headspace solid‐phase microextraction, with a single analysis taking 30–60 min (X. Zhang, Wei, et al. 2021). GC–IMS bridges the technological gap between rapid screening and high sensitivity (S. Wang et al. 2020). GC–IMS combines the efficient separation of gas chromatography with the rapid gas‐phase separation and detection capabilities of ion mobility spectrometry, making it particularly suitable for the rapid screening and qualitative analysis of volatile and semivolatile organic compounds in complex mixtures. The selective response of its ion mobility spectroscopy to polar molecules makes it more sensitive in detecting reactive flavor compounds such as aldehydes and ketones. Moreover, it can complete analysis within seconds without the need for complex sample enrichment steps (K. Zhang, Tang, Liu, et al. 2025). H. Ling et al. (2022) precisely identified 51 volatile organic compounds in fermented bean paste using GC–MS, including key flavor substances such as 2‐ethyl‐3,5‐dimethylpyrazine at low concentrations. However, it has weaker separation capabilities for high‐molecular‐weight compounds (such as some furan derivatives) and may lead to identification biases of unknown substances due to insufficient database coverage. GC–O takes a different approach by combining human olfaction with instrumental analysis to achieve ultra‐sensitive screening of aroma‐active substances (Han et al. 2024). GC–O directs compounds separated by split‐flow gas chromatography simultaneously to a detector and an olfactory port, allowing analysts to smell and record odor characteristics in real time. Particularly through methods such as aroma extract dilution analysis, it can accurately and quantitatively evaluate the contribution of each compound to flavor. For example, Lu et al. (2021) found that certain sulfides in Zhuhoujiang were identified by olfaction before being detected by instruments. However, this technology is limited by individual differences among evaluators and fatigue effects, making standardization challenging.

In summary, the four flavor detection techniques, namely, electronic nose/electronic tongue, GC–MS, GC–IMS, and GC–O, each have their own advantages and disadvantages. A single technique is insufficient to comprehensively analyze complex samples. The combined application of multiple technologies has become an inevitable trend in flavor research of foods such as fermented bean paste. For instance, the combination of GC–MS and GC–O enables rapid identification of key aroma compounds (X. Zhang, Wei, et al. 2021). At the same time, the complementary use of electronic nose/electronic tongue and GC–MS allows simultaneous detection of volatile and nonvolatile components (K. Zhang, Tang, Li, et al. 2025). This approach overcomes the limitations of individual methods and achieves a comprehensive, accurate, and efficient analysis of flavor profiles.

4. Correlation Between Microbial Communities and Aroma Compounds

The rich microbial communities in fermented bean paste exhibit intricate and multidimensional relationships with the aroma compounds they metabolize, spanning different taxonomic levels (Feng et al. 2024). These relationships form the core mechanism determining the product's sensory quality and flavor characteristics. Multivariate statistical methods, including principal component analysis, cluster analysis, and partial least squares regression, have identified significant correlations between 72 key flavor compounds and 59 microbial genera (Figure 5A; Feng et al. 2024; Z. Tian, Ameer, et al. 2022; K. Zhang, Tang, Li, et al. 2025). The formation of these flavor compounds exhibits distinct microbial specificity, and the underlying metabolic pathways, along with the mechanisms catalyzed by key enzymes, are particularly crucial (Figure 5B). The core functional microorganisms often exhibit the characteristic of “one bacterium, multiple effects,” which mainly stems from their complex metabolic networks (K. Zhang, Tang, Li, et al. 2025). For example, Tetragenococcus, a genus within LAB, can produce acetic acid via glycolysis and pyruvate metabolic pathways, providing a sour undertone to the product. Simultaneously, it participates in the synthesis of flavor compounds such as benzaldehyde through the Ehrlich pathway (An et al. 2021; Z. Y. Chen et al. 2015). Similarly, Zygosaccharomyces rouxii, a species within the yeast genus, also demonstrates diverse metabolic functions. It can convert branched‐chain amino acids via the Ehrlich pathway to produce 3‐methyl‐1‐butanol, which imparts a malt‐like aroma, while also synthesizing furanone compounds through the Maillard reaction. Additionally, it can generate 3‐methylbutanal and benzaldehyde via the Ehrlich pathway, contributing nutty, honey‐like, and cherry‐like aromatic layers to the product (Z. Y. Chen et al. 2015; Diez‐Simon et al. 2020; L. Zhang, Huang, et al. 2021). Furthermore, functional microorganisms such as Candida also exhibit multiple roles: they can synthesize phenylethanol and other compounds via the Ehrlich pathway, while also producing various esters like ethyl acetate through esterification reactions, thereby contributing rich fruity and sweet aromas (An et al. 2021; Y. Chen et al. 2021; Chun et al. 2021; E. J. Lee et al. 2018). Furthermore, the same aroma compound can be produced by different microorganisms through distinct pathways. For example, the mushroom‐like 3‐octanone can be generated either by Bacillus via β‐oxidation or by Kodamaea through fatty acid degradation (X. Yue et al. 2021). Similarly, acetophenone is positively correlated with seven microbial genera, including Tetragenococcus, Ignatzschineria, and Clostridium (Pu et al. 2023). Notably, environmental factors can also modulate microbial community structure and influence flavor profiles (Feng et al. 2024). For instance, increasing salinity enriches Tetragenococcus halophilus, promoting the accumulation of 3‐methylbutanol and ethyl 3‐methylbutanoate (Feng et al. 2024).

FIGURE 5.

(A) The relationship between microbial flora and flavor substances in fermented bean paste (An et al. 2021; Y. Chen et al. 2021; Z. Y. Chen et al. 2015; Chun, et al. 2021; Diez‐Simon et al. 2020; E. J. Lee et al. 2018; K. Zhang, Tang, Li, et al. 2025; L. Zhang, Huang, et al. 2021). (B) The main formation pathways of flavor compounds in fermented bean paste.

Although correlation‐based analytical methods can effectively reveal the potential relationships between microbial communities and flavor compounds, the results mainly reflect statistical correlations rather than strict causal relationships. In complex fermentation systems, the associations between microorganisms and flavor may arise not only from direct metabolic activities but also from indirect regulatory effects. Therefore, accurately distinguishing between “correlative evidence” and “causal evidence” is a critical prerequisite for achieving precise flavor regulation. Based on this, this review further summarizes the core functional microorganisms closely associated with flavor formation and the levels of evidence supporting their roles (Table 3), aiming to provide a more solid theoretical foundation for the targeted regulation of fermented bean paste flavor and the development and utilization of functional microbial resources.

TABLE 3.

The direct flavor‐related core microbiome in fermented bean paste.

Core microbial community	Related flavor compounds	Verification method	References
L. plantarum	3‐Methylbutyric acid, ethyl 3‐methylbutyrate, 2,5‐dimethyl‐3‐ethylpyrazine	Inoculated fermentation + GC–MS	K. Zhang, Li, et al. (2025)
T. halophilic	2‐Methylbutyraldehyde, 3‐methylbutyraldehyde, 2‐methylbutanol, decanal, (trans)‐2‐octenal, acetic acid, 1‐octen‐3‐ol, 2‐butylfuran, 2‐pentylfuran, trans‐2‐(2‐pentenyl) furan, benzaldehyde, phenylacetaldehyde, 2,6‐dimethylpyrazine, 3‐ethyl‐2,5‐dimethylpyrazine	Inoculated fermentation + GC–MS	J. Li, Sun, et al. (2024)
B. velezensis, A. oryzae	2‐Ethylphenol, 2,5‐dimethylpyrazine, linalool, isobutyric acid 1‐methylbutyl ester, 1‐phenyl‐1‐pentanone, n‐hexanal, phenol	Inoculated fermentation + GC–MS	M. Yue et al. (2026)
P. glabrum, A. oryzae	Isoleucine, ornithine, glutamine, citrulline, 2‐oxoglutarate, l‐homoserine, aspartic acid, asparagine, pyruvic acid, fumaric acid, L‐malic acid, succinic acid, D‐glyceric acid	Inoculated fermentation + GC–TOFMS	Sun et al. (2019)
A. oryzae, P. acidilactici, Penicillium, Yaniella, Tetragenococcus, Candida solani	2,3,5,6‐Tetramethylpyrazine, 2,3,5‐trimethylpyrazine, 2,6‐dimethylpyrazine, acetophenone, dimethyl disulfide, 2‐propenyl hexanoate, methyl 2‐methylbutanoate, indole, Tyr, Pro, quinic acid, malic acid	Inoculated fermentation + metagenomics + GC–MS + HPLC	X. Zhang, Shan, et al. (2022)
P. fluorescens	Tyramine, histamine	Metagenomics	Xie et al. (2020)
Lactobacillus, Enterococcus, Leuconostoc, Streptococcus, Saccharomyces, Schizosaccharomyces	Alanine, glycine, serine, and lysine, leucine and isoleucine, glutamic acid	Metaproteomics	P. Zhang et al. (2018)
Tetragenococcus, Penicillium, Aspergillus, Bacillus	Ethanol, pyruvic acid, lactic acid, fatty acid	Metaproteomics + metabonomics	X. Zhang, Shan, et al. (2022)
Lactobacillus, Tetragenococcus, Melissococcus, Lysinibacillus, Staphylococcus	Ethanol, 1‐octen‐3‐ol, ethyl hexadecanoate, ethyl linoleate, ethyl oleate, acetic acid, hexadecanoic acid, benzaldehyde, phenylacetaldehyde, 3‐octanone	Metatranscriptome + GC–MS	An et al. (2021)
E. faecium, T. halophilus, B. licheniformis, S. succinus	Acetic acid, hexanoic acid, phenylmethanol, 2,3,5,6‐tetramethylpyrazine, pentan‐1‐ol, 3‐methylbutyl acetate, 2‐phenylethanol, ethylbenzene, 2‐ethylhexan‐1‐ol, hexan‐1‐ol, oct‐1‐en‐3‐ol, 3‐hydroxybutan‐2‐one, 2,5‐bis[(trimethylsilyl)oxy]benzaldehyde, 2‐pentylfuran, and methoxy‐phenyl‐oxime	Inoculated fermentation + GC–MS	Jeong et al. (2017)
A. oryzae, B. amyloliquefaciens, B. subtilis, T. halophilus, Z. rouxii	Acetic acid, 2‐methylhexanoic, butanoic, hexanoic, 1‐octen‐3‐ol, 2‐pentylfuran, tetramethylpyrazine, ethanol, benzaldehyde, benzeneacetaldehyde	Inoculated fermentation + GC–MS	E. J. Lee et al. (2018)
B. subtilis	2,5‐Dimethylpyrazine, 2,3,5‐trimethylpyrazine, 2,3,5,6‐tetramethylpyrazine, 2,3,5‐methyl‐6‐ethylpyrazine	Inoculated fermentation + GC–MS + real‐time quantitative PCR	Y. Liu et al. (2018)
B. subtilis	1‐Hexanol, acetoin, 2‐hydroxy‐3‐pentanone, toluene, p‐xylene, trimethylpyrazine, tetramethylpyrazine, 2,5‐dimethylpyrazine, 2‐ethyl‐3,5‐dimethylpyrazine, 2,3,5‐trimethylpyrazine, 2‐methylpyrazine, 2,3‐butanedione	Inoculated fermentation + GC–MS + HPLC	X. Chen, Lu, et al. (2022)
R. oligosporus	1‐Monolinolein, 2‐monolinolein, 1‐monoolein, 2‐monoolein, 1‐monolinolenin, 2‐monolinolenin, α‐linolenic acid, linoleic acid, oleic acid, palmitic acid, stearic acid.	Inoculated fermentation + GC–MS	Kusumah et al. (2020)
R. oligosporus, S. cerevisiae	Styrene, caryophyllene, phenol, maltol, 1‐octen‐3‐ol, butanoic acid	Inoculated fermentation + GC–MS + sensory evaluation	Kustyawati (2017)
Aspergillus, Corynebacterium, Lactococcus, Atopostipes, Peptostreptococcus, Truepera, Gulosibacter, Actinomycetaceae, Rhodobacteraceae	2‐Methyl‐butanal, ethyl 2‐methylbutyrate, isoamyl acetate, 2,6‐dimethylpyrazine, 1‐octen‐3‐ol, 2‐pentyl furan, benzeneacetaldehyde, phenylethyl alcohol, phenethyl acetate, phenethyl butyrate	Inoculated fermentation + amplicon sequencing + GC–MS	H. Yang et al. (2019)
A. oryzae, M. caribbica, M. guilliermondii, C. etchellsii, C. versatilis	Citric acid, butane diacid, malic acid, palmitic acid, stearic acid, oleic acid, octadecenoic acid, linoleic acid, α‐linolenic acid, 4‐ethylguaiacol, 4‐ethylphenol	Inoculated fermentation + amplicon sequencing + GC–MS + HPLC	B. He et al. (2019)
M. racemosus, A. oryzae, Candida, Weissella, Meyerozyma, Bacillus, Ochrobactrum, Serratia, Enterobacteriaceae, Pantoea, Nocardiopsis, Pseudomonas, Enterococcus, Penicillium, Gibberella, Yunnania, Microascus, Acremonium, Fusarium, Debaryomyces	Ethyl isovalerate, isoamyl acetate, methyl phenylacetate, ethyl phenylacetate, trans‐11‐octadecenoic acid methyl ester, methyl linoleate, palmitic acid ethyl ester, ethyl linoleate, phenylethyl alcohol	Inoculated fermentation + amplicon sequencing + GC–MS	B. He et al. (2019), Y. Zhang, Zeng, et al. (2021)
A. elegans, T. halophilus, Z. rouxii, C. versatilis	Benzene acetaldehyde, 1‐octen‐3‐ol, acetophenone, 2‐methoxyphenol, 4‐ethylphenol, 2‐pentylfuran, benzaldehyde, ethyl octanoate	Inoculated fermentation + GC–MS + sensory evaluation	G. He et al. (2016)
B. amyloliquefaciens	1‐Octen‐3‐ol, linalool, (E)‐2‐nonenal, 3‐methylbutanal, nonanal, phenylacetaldehyde, benzaldehyde, acetoin, 2,5‐dimethylpyrazine and styrene	Inoculated fermentation + UPLC–Q–TOF–MS	J. Cai et al. (2026)
Leuconostoc, Pseudomonas, Staphylococcus, Erwinia, Rahnella	α‐Pinene, fatty acid methyl esters	Metagenomic + metabolomic insights	C. Liu et al. (2026)

5. Regulating the Flavor of Fermented Bean Paste Through Intelligent Biomanufacturing Technology

Driven jointly by synthetic biology and digital twin technology, the regulation of fermented bean paste fermentation is advancing from traditional empirical models toward a new phase of predictive science based on first principles. This convergence not only promotes the upgrading of intelligent equipment and the transformation of production models but also demonstrates potential for broader application across fermented food sectors, with promising prospects for enhancing quality, reducing energy consumption, and ensuring safety. However, current technologies still face numerous challenges in practical applications, such as insufficient predictive accuracy of models in complex biological reactions, as well as adaptability and stability issues when scaling up from laboratory to industrial production. These obstacles require continuous research and innovation in future applications. Despite these challenges, this direction has laid a critical foundation for the intelligent development of the food industry and will drive ongoing optimization across the supply chain.

5.1. Applications and Potential of Synthetic Biology in Fermented Bean Paste

SynComs originated in the early 21st century and are grounded in the Design–Build–Test–Learn (DBTL) theoretical framework (Y. Zhang et al. 2026). By integrating two complementary research strategies—top‐down and bottom‐up—this framework provides a complete methodological foundation for the systematic design and optimization of SynComs (Nikoloudaki et al. 2024). The top‐down strategy begins with complex natural microbial communities, employing systematic microbiome analysis to identify core microbial taxa and their functional modules (San León and Nogales 2022). Representative strains with strong functional performance and environmental adaptability are then screened and assembled to construct low‐complexity yet functionally stable synthetic communities. This approach preserves key functions of natural systems while significantly reducing community complexity and uncertainty. In contrast, the bottom‐up strategy is driven by clearly defined application‐oriented goals, centering on the screening of target microorganisms and focusing on desired functions (Mao et al. 2021). It emphasizes the elucidation of metabolic division of labor, metabolic complementarity, and interspecies interactions among strains. By incorporating Genome‐Scale Metabolic Models (GEMs) to quantitatively assess interspecies interactions, this strategy enables the rational assembly of SynComs with predictable functions and stable performance.

Currently, the application of SynComs in fermented bean paste systems has been experimentally validated. For example, Jia et al. (2020) screened and identified five core dominant bacterial genera via amplicon sequencing. Combined with in vitro experiments and in situ verification, they analyzed and reconstructed the structure and function of the microbial community and further constructed an artificial synthetic microbial community capable of stable fermentation under 6% salinity conditions. Niu et al. (2024) constructed a synthetic microbial community consisting of Aspergillus oryzae, Staphylococcus carnosus, Tetragenococcus halophilus, and Weissella confuse and applied it to broad bean paste at 10% salinity, successfully promoting the accumulation of volatile metabolites while maintaining acidity stability under reduced salt conditions. These approaches effectively address the traditional reliance on highly dynamic and uncontrollable microbial communities in natural environments, which often cause batch‐to‐batch variations in product quality and flavor. Moreover, guided by the core principles of “engineering” and “controllability,” a comprehensive technical system has been developed encompassing strain screening, SynComs construction, metabolic regulation, and process optimization, enabling precise control over the fermentation process of fermented bean paste (Figure 6). Therefore, SynComs provide an effective means to shift fermented microbiome research from “highly complex natural systems” to “controllable, reconstructible, and interpretable model systems.” This approach helps elucidate the functional roles of microorganisms at multiple levels—including metabolic pathways, cooperative networks, and competitive strategies—and lays a theoretical foundation for engineering‐oriented design.

FIGURE 6.

Integrating synthetic biology and digital twin technologies to optimize fermentation and enhance the flavor of fermented bean paste.

SynCom enables researchers to precisely manipulate community composition, species ratios, and environmental factors under well‐defined experimental conditions, allowing for the dissection of key ecological processes such as resource competition, metabolic dependence, and functional complementarity. By intentionally adding or removing specific microorganisms, the necessity or redundancy of species within a metabolic network can be determined. Adjusting the initial inoculation ratios allows for the quantification of how interspecies interactions contribute to community assembly. Furthermore, by introducing environmental disturbances such as changes in temperature, oxygen availability, or pH fluctuations, it becomes possible to systematically observe the remodeling of microbial interactions in response to changing environmental conditions, thereby simulating the dynamic niche reconstruction characteristic of solid‐state fermentation. This mechanistic research framework transforms the uncontrollable fermentation process into manageable experimental variables, clearly defining the functional roles of different microorganisms within metabolic pathways, cooperative networks, and competitive interactions, thereby providing robust support for in‐depth investigations into the mechanisms governing microbial communities.

5.2. Applications and Potential of Digital Twin Technology in Fermented Bean Paste

With the rapid emergence of intelligent fermentation technologies, digital twin technology—integrating the Internet of Things, big data, artificial intelligence, and simulation—has gradually become a core tool for achieving real‐time prediction, mechanism analysis, and intelligent control of solid‐state fermentation processes, providing strong technical support for the intelligent transformation and upgrading of the solid‐state fermentation industry (Figure 6; Jamshidi et al. 2023). Building a high‐quality digital twin model for solid‐state fermentation cannot rely solely on the collection of individual process parameters or the application of traditional kinetic models. Its core essence lies in the systematic integration of the complex microbial interaction mechanisms during solid‐state fermentation, including competition, mutualism, cross‐feeding, metabolic complementarity, and cooperative behaviors among microorganisms (Y. Zhang, Zhou, et al. 2025). At the same time, advances in the field of SynCom provide a highly controllable experimental platform and scientifically sound design strategies for the model construction and experimental validation of digital twin technology, further refining the technical framework of digital twins. By integrating multidimensional indicators such as environmental spatial heterogeneity, microbial functional traits, environmental stress resistance, microbial interaction networks and kinetic modeling, and relying on the closed‐loop iterative mechanism of “experimental observation—simulation—data assimilation—state update,” digital twin technology realizes dynamic calibration and accurate predictive analysis of the entire fermentation process. It promotes the transformation of the solid‐state fermentation industry from the traditional experience‐driven model to a scientifically rigorous mechanism‐based interpretation and precise regulation model, laying a solid theoretical foundation for the accurate prediction, process optimization and efficient control of fermentation processes.

Currently, digital twin technology has not yet been applied in the field of fermented bean paste fermentation, but it has been experimentally validated in various other fermentation systems. For instance, during fermentation, digital twin models can track in real time the dynamic evolution of microbial communities, such as the shift of Lactobacillus curvatus from an initial 39.7% to a dominant strain after 100 days of fermentation (J. Y. Kim, Park, et al. 2022). This real‐time visualization of microbial growth curves provides data‐driven support for process adjustments (Y. Zhang, Zhu, et al. 2025). In strain development, researchers can validate key indicators such as metabolic activity, antimicrobial capacity, and environmental adaptability in virtual environments based on genetic and metabolic profiling, enabling the screening and targeted cultivation of optimal strains (Y. Zhang, Zhu, et al. 2025). Therefore, in the future, by integrating key information such as microbial interactions, metabolic division of labor, and environmental heterogeneity during the fermentation of fermented bean paste, it can be transformed into quantifiable and modelable data inputs. This will enable precise control and real‐time dynamic feedback in the production process, allowing for early prediction of fermentation progress and quality. Ultimately, this will drive the transformation and upgrading of the traditional experience‐led production model of solid‐state fermentation of fermented bean paste into a modern fermentation model supported by theoretical foundations, characterized by controllable processes, precision, and high efficiency.

6. Conclusions and Outlook

This article provides a comprehensive review of the research progress in the field of fermented bean paste fermentation science. It systematically elaborates on the spatiotemporal heterogeneity characteristics of microbial communities, the formation mechanisms of multisensory flavor compounds, and the features of detection techniques. It reveals the complexity and synergy of the microbial–metabolite network during traditional fermentation processes. As a typical representative of traditional fermented foods in East Asia, the unique flavor of fermented bean paste originates from the synergistic effects of dynamic microbial community succession and metabolic products. Key functional microbial groups, including molds, LAB, yeasts, and bacilli, decompose proteins, carbohydrates, and lipids, generating volatile and nonvolatile compounds such as esters, alcohols, aldehydes, ketones, and phenols, which collectively construct its multilayered flavor profile. In recent years, molecular biology techniques such as high‐throughput sequencing and metagenomics have precisely analyzed the composition and functional evolution of microbial communities. Meanwhile, analytical methods, including GC–MS, GC–IMS, and electronic nose/tongue, have enabled qualitative and quantitative research on flavor compounds, further clarifying the associations between core flavor compounds and specific microbial metabolic pathways. Although existing technologies have advanced relevant research, limitations still exist, such as the constraints of single analytical techniques and insufficient regulatory capacity over bacterial metabolism. Therefore, future research should focus on integrating complementary multiplatform analytical technologies, constructing synthetic microbial communities via synthetic biology and directionally optimizing their metabolic networks, and combining digital twin technology to achieve real‐time monitoring and intelligent regulation of fermentation processes. Meanwhile, predictive models that accurately correlate flavor compound profiles with sensory perception should be established to break through the current bottleneck of insufficient understanding of characteristic functional microorganisms and their metabolic mechanisms. This interdisciplinary research paradigm will strongly promote the transformation of the fermented bean paste industry from traditional experience‐driven methods to data‐driven precision manufacturing, laying a scientific foundation for the precise design of flavor and quality. While ensuring the uniformity of industrial product quality, it will also preserve the essence of intangible cultural heritage techniques, comprehensively enhancing the quality, safety, and social value of traditional fermented foods, leading the industry into a new stage of intelligent development.

Author Contributions

Kangyong Zhang: conceptualization, methodology, writing – original draft, writing – review and editing, investigation. Hongwei Tang: conceptualization, methodology, writing – review and editing. Libo Liu: conceptualization, methodology, writing – review and editing. Bailiang Li: writing – original draft, writing – review and editing, funding acquisition. Fei Liu: writing – original draft, writing – review and editing. Zhongjiang Wang: writing – original draft, writing – review and editing.

Conflicts of Interest

The authors declare no conflicts of interest.