Background Solid-state vinegar fermentation, predominantly practiced in China, is characterized by high microbial diversity, multi-substrate co-decomposition, abundant metabolite production, pronounced environmental heterogeneity, and complex interaction networks that collectively shape vinegar quality. However, this process remains largely experience-driven, with limited digitalization and an incomplete understanding of microbial interaction mechanisms. Key challenges include unclear interaction mechanisms, insufficient application of modeling tools such as genome-scale metabolic models, and limited capacity for dynamic process control. Addressing these gaps is essential for improving fermentation efficiency, stabilizing flavor quality, and advancing vinegar modernization. Scope and approach This review is centered on microbial interactions and outlines the metabolic division of labor of key microorganisms in solid-state fermentation and their roles in flavor formation. It examines microbial ecological relationships and interaction mechanisms, and analyzes how environmental heterogeneity regulates microbial interactions. The review further introduces synthetic microbial communities as tools for mechanism validation and functional reconstruction. Finally, based on clarified mechanisms, it discusses how key microbial, metabolic, and environmental information can be translated into model inputs to construct digital twin systems for directional control. Key findings and conclusions Environmental heterogeneity plays a role in shaping microbial interaction patterns, thereby influencing metabolic division of labor, fermentation efficiency, and flavor formation. Microbial interactions, rather than individual species, are drivers of community stability, functional metabolite production, and flavor complexity. The integrated application of synthetic microbial communities, metabolic flux models, and digital twin technologies constitutes a predictive framework for dissecting microbial interactions, optimizing key consortia, and achieving targeted regulation of cereal vinegar fermentation.

The efficient biosynthesis of cytidine-5 '-diphosphocholine (CDP-choline), a vital pharmaceutical intermediate, relies on a sequential two-step pathway catalyzed by choline kinase (CKI) and CTP:phosphocholine cytidylyltransferase (CCT). However, this bioprocess is largely bottlenecked by the sluggish kinetics of CCT, an amphipathic membrane-bound enzyme whose rational engineering is severely hindered by the lack of solved crystal structures. To alleviate this kinetic limitation, this study established a physics-informed and evolution-guided integrated strategy. First, a sequence-structure-dynamics three-dimensional funnel screening workflow was constructed. By combining AlphaFold2-based structural topology filtering with DLKcat kinetic pre-screening, we successfully captured a high-potential candidate CCT from Maudiozyma saulgeensis (MsCCT), with superior chassis compatibility, which showed an 8.69-fold increase in total catalytic activity compared to the control CCT from Saccharomyces cerevisiae (ScCCT). Subsequently, we employed a multi-view integrated Protein Language Model (PLM) zero-shot prediction method. Through the cross-validation of ProGen (evolutionary perspective), SaProt (global structural perspective), and ProSST (local micro-environmental perspective), combined with a domain-focusing strategy, the mutant MsCCTM3 was identified, exhibiting a 2.4-fold catalytic enhancement over the wild-type MsCCT. Upon integrating the evolved MsCCTM3 into a metabolically optimized Bacillus subtilis chassis, multi-omics analysis deciphered a critical kinetic mismatch, revealing that the hyper-active enzyme shifted the metabolic bottleneck to upstream substrate transport and precursor supply capacity. Ultimately, the engineered strain achieved a high titer of 5.65 g/L with a space-time yield of 0.202 g/(L & sdot;h) in a 5 L bioreactor, representing a 34.7% improvement over the control strain. This work not only provides an efficient strain for the industrial production of CDP-choline but also offers an integrated data-driven workflow that demonstrates promising potential for engineering complex membrane proteins lacking structural templates.

Nonmodel microorganisms offer substantial potential as next-generation microbial chassis (NGMCs), yet most lack efficient and broadly transferable genome-editing systems. Here we present ETfinder, a framework that uses the conserved C-terminal tail of host single-stranded DNA-binding proteins (SSB-Ct) as a biochemical constraint to guide the discovery of RecET recombineering systems. Applied to Rhodobacter sphaeroides, ETfinder identified 91 candidates from 18 841 α-proteobacterial genomes, and all five experimentally tested RecT homologs supported measurable double-stranded DNA (dsDNA) recombineering, with the Paracoccaceae SJ630 system reaching 8.9 × 10² colony-forming units (CFU) per μg of dsDNA and 100% editing accuracy. Testing in Halomonas further showed that RecT proteins from evolutionarily distant taxa remain functional within the same halophilic chassis, indicating that SSB-Ct-guided selection enriches for portable recombination modules beyond phylogenetic proximity. To facilitate broad adoption, we compiled 25 529 RecT-SSB pairs into a curated database and implemented ETfinder as a standalone, locally deployable toolkit for mining, ranking, and phylogenetic visualization. This framework prioritizes high-compatibility homologs, reduces experimental screening burden, and expands the accessible genome-editing toolbox for NGMCs. ETfinder is freely available at https://github.com/lvdongyuan/ETfinder.

Solid-state fermentation (SSF) is a key process in biomanufacturing; however, surface-associated microbial growth, complex sample pretreatment, and signal interference caused by solid particles and complex matrices make it difficult to rapidly, continuously, and synchronously monitor microbial states and key metabolic indicators during fermentation, thereby hindering precise process control and intelligent production. To address this bottleneck, this study employed the SSF of Shanxi aged vinegar as a model system and developed a Process Analytical Technology (PAT) platform by integrating dielectrophoresis-assisted microfluidic single-cell Raman spectroscopy with machine learning, enabling the simultaneous analysis of microbial community composition and physicochemical indicators. a single-cell Raman spectral database was established, and 54 features were identified for six microbial biomarkers. Using a Logistic Regression model, the method achieved a classification accuracy of 99.6%, enabling effective tracking of microbial community succession. A Multilayer Perceptron algorithm was further employed to directly quantify key physicochemical indicators from complex process spectra, demonstrating excellent predictive performance for products acetic acid (R² = 0.93) and lactic acid (R² = 0.87), and subtract reducing sugars (R² = 0.87). This dual-dimensional PAT strategy was successfully validated in both a fermentation vats and the industrial closed solid-state fermenter, enabling synchronous feedback of microbial community dynamics and metabolic states within 1 hour. This rapid at-line monitoring approach bridges the gap between empirical observation and digital, rational process control, providing a robust data-driven foundation for the intelligent automation of complex solid-state biomanufacturing.

The discrepancy between in situ microbial abundance and actual metabolic performance represents a critical challenge for interpreting microbial function from meta-omic data. Here, we integrated metagenomic and metatranscriptomic sequencing to investigate this decoupling between microbial abundance and cultivation-based physiological potential in Shanxi aged vinegar (SAV) solid-state fermentation. Lactobacillus acetotolerans dominated the community at both the genomic (40.89%) and transcriptomic (55.36%) levels, whereas Pediococcus acidilactici accounted for only 0.11%—a canonical rare-biosphere member. Source tracking via Sankey analysis showed that genes involved in acetate production were primarily attributed to Acetobacter pasteurianus, whereas genes involved in lactate production were predominantly associated with Lactobacillus spp. However, L. acetotolerans exhibited limited acid tolerance and lactic acid production, whereas the low-abundance P. acidilactici AAF1-5 displayed robust stress tolerance and superior lactic acid production under fermentation-relevant conditions—a striking contrast between microbial abundance and physiological performance. Metabolic interaction network analysis predicted that P. acidilactici may be co-inhibited by L. acetotolerans (Ixy = −2.737, resource competition) and A. pasteurianus (Ixy = −1.887, acid stress). To test whether ecological constraints, rather than intrinsic metabolic capacity, underlie this low abundance, we heterologously expressed the heat shock co-chaperone gene grpE from A. pasteurianus in P. acidilactici AAF1-5 as an experimental tool. The recombinant strain P. acidilactici-grpE exhibited significantly enhanced viability under acetic acid stress and, in simulated SAV fermentation, lactic acid content increased by 23.63% compared with the wild-type control. These results demonstrate that meta-omic abundance does not necessarily predict physiological performance and that low abundance may reflect ecological constraints rather than intrinsic functional deficiency. Our study provides an ecological framework for linking microbial abundance with physiological function beyond sequence-based abundance inference in complex fermentation microbiomes.

Acetic acid fermentation (AAF) is a vital stage in the production of Shanxi aged vinegar (SAV), during which microbial community succession and metabolic reprogramming jointly determine acidification efficiency and flavor development. However, the system-level mechanisms linking microbial dynamics to functional metabolism during AAF remain insufficiently understood. In this study, metatranscriptomic analysis was employed combined with KEGG pathway enrichment, CAZy profiling, and global metabolic network visualization to elucidate the functional landscape of AAF across different fermentation stages. The results indicated a pronounced stage-dependent metabolic shift during AAF. In the early stage, nucleotide biosynthesis, and sugar metabolism pathways were highly active, accompanied by strong expression of CAZy-related genes, suggesting a growth-oriented metabolic strategy driven by efficient polysaccharide degradation and carbon assimilation. As fermentation progressed, mainly Acetobacter and Komagataeibacter, became functionally dominant, reinforcing nitrogen metabolism and central carbon metabolism. In the late stage, increasing acid stress induced a transition toward maintenance- and stress-adaptation-oriented metabolism, characterized by enhanced oxidative phosphorylation, ATP-dependent proton transport, nitrogen assimilation, and amino acid metabolism. Notably, amino acid metabolism emerged as a key metabolic axis linking stress tolerance and flavor maturation, while a clear functional division of labor between Lactobacillus and Acetobacter was observed, forming a cooperative metabolic network that stabilized fermentation performance. Collectively, this study provides a system-level view of microbial and metabolic coordination during SAV AAF and offers mechanistic insights into the self-organizing nature of solid-state vinegar fermentation.

l-Histidine is an essential amino acid with important applications in pharmaceuticals and nutrition, highlighting the demand for efficient microbial production platforms. This study developed a high-performance Escherichia coli cell factory through systematic metabolic engineering. First, we mined a feedback-resistant hisG∗ smar and the entire mutant his operon from a previously obtained high l-histidine-producing mutant of Serratia marcescens, offering novel enzymatic parts beyond conventional sources. Combined with precursor supply enhancement and redox balancing, the engineered strain yielded 4.46 g/L l-histidine. Second, a machine learning-based platform (TransDW) was utilized to predict and validate a novel efflux transporter, Cgl1374, increasing titer to 4.82 g/L. Third, we implemented a growth phase-dependent system to dynamically regulate pgi expression, redirecting carbon flux and achieving 5.49 g/L in shake flasks. Finally, applying a novel carbon evolution rate (CER)-based control strategy in fed-batch fermentation, the optimized strain achieved 49.8 g/L of l-histidine in a 5-L bioreactor, with a yield of 0.265 g/g glucose, which is the highest yield reported for engineered E. coli. This work establishes a synergistic framework combining non-model gene discovery, computational transport engineering, and real-time physiological feedback control, providing a versatile blueprint for next-generation microbial cell factories.

The growing demand for N-acetylneuraminic acid (NeuAc) has driven the need for efficient and environmentally sustainable biomanufacturing processes. Microbial fermentation offers a promising route, yet optimizing cell factories with excellent phenotypes remains challenging. Here, we engineered Escherichia coli to enable high-efficiency co-utilization of glucose and glycerol. We refactored two synthetic pathways with the same start and end to enhance N-acetylmannosamine (ManNAc) precursor levels and optimized NeuAc synthase using artificial intelligence (AI) techniques and machine learning (ML) sequence mining. Subsequently, phosphoenolpyruvate (PEP) levels were boosted by capturing carbon flow from competing regeneration pathways, thus balancing the intracellular PEP:ManNAc ratio for improved NeuAc synthesis. Besides glucose, an additional carbon inlet from glycerol was opened, achieving a NeuAc titer of 70.4 g/l in fed-batch fermentation with a productivity of 1.17 g/l/h. This work demonstrates a highly efficient microbial cell factory for the biosynthesis of NeuAc and provides a versatile system engineering strategy applicable to other high-value compounds.

BACKGROUND:Norovirus (NoV) is the leading cause of foodborne disease outbreaks worldwide, typically spreading via contaminated food and water. Rapid, sensitive, and portable detection of NoV is crucial. RESULTS:Here, we presented a magnetic CRISPR/Cas12a-SERS nanobiosensor capable of detecting NoV with high sensitivity, accuracy, speed, and portability. In this nanobiosensor, SERS nanoprobes linked to magnetic nanoprobes via linker single-stranded DNAs (ssDNAs). The presence of NoV nucleic acid triggered Cas12a's trans-cleavage activity, degrading the linker ssDNA. After magnetic separation, the dissociated SERS nanoprobes were efficiently separated from the magnetic nanoprobes. This enhanced the SERS signal in the supernatant, detectable using a portable Raman spectrometer. The detection limit for NoV is 100 copies/mL within 60 min. The nanobiosensor was further assessed in real-world settings, demonstrating excellent sensitivity and selectivity for detecting trace NoV in complex food samples. SIGNIFICANCE:This approach not only broadens CRISPR-based pathogen detection but also provides a reliable tool for monitoring foodborne viruses. Its potential extends beyond NoV, promising enhanced surveillance of various pathogens in food safety, environmental monitoring, and public health sectors.

The growing demand for N-acetylneuraminic acid (NeuAc) has driven the need for efficient and environmentally sustainable biomanufacturing processes. Microbial fermentation offers a promising route, yet optimizing cell factories with excellent phenotypes remains challenging. Here, we engineered Escherichia coli to enable high-efficiency co-utilization of glucose and glycerol. We refactored two synthetic pathways with the same start and end to enhance N-acetylmannosamine (ManNAc) precursor levels and optimized NeuAc synthase using artificial intelligence (AI) techniques and machine learning (ML) sequence mining. Subsequently, phosphoenolpyruvate (PEP) levels were boosted by capturing carbon flow from competing regeneration pathways, thus balancing the intracellular PEP:ManNAc ratio for improved NeuAc synthesis. Besides glucose, an additional carbon inlet from glycerol was opened, achieving a NeuAc titer of 70.4 g/l in fed-batch fermentation with a productivity of 1.17 g/l/h. This work demonstrates a highly efficient microbial cell factory for the biosynthesis of NeuAc and provides a versatile system engineering strategy applicable to other high-value compounds.

Advances in strain breeding for butanol biosynthesis were quite limited because of physiological complexity of solventogenic Clostridia. Using AI, this study developed a high-throughput screening method for Clostridium acetobutylicum to find strains with inhibitor tolerance and high butanol production. A mutant library was generated from C. acetobutylicum ATCC 824 through ARTP mutagenesis and physiological traits were digitized using color indicators. The classification performance of Machine learning algorithms (PCA, PLS, SVM, ANN) were compared for different butanol-producing strains. Among 2000 strains screened, C. acetobutylicum Tust-f3 was identified, which could tolerate 4.5g/L furfural and yield 10.5g/L butanol from undetoxified lignocellulosic hydrolysate. Proteome analysis reveals that 38 proteins may play a crucial role. Subsequently, seven universal detoxification components for furfural were identified via heterologous expression in E. coli Genes CA_RS19590 and CA_RS08810 showed significant growth improvement (14.44 and 14.28-fold, respectively, compared to control). This study highlights the potential of machine learning in strain selection and breeding.

Geobacillus thermoglucosidasius NCIMB 11955 possesses advantages, such as high-temperature tolerance, rapid growth rate, and low contamination risk. Additionally, it features efficient gene editing tools, making it one of the most promising next-generation cell factories. However, as a non-model microorganism, a lack of metabolic information significantly hampers the construction of high-precision metabolic flux models. Here, we propose a BioIntelliModel (BIM) strategy based on artificial intelligence technology for the automated construction of enzyme-constrained models. 1). BIM utilises the Contrastive Learning Enabled Enzyme Annotation (CLEAN) prediction tool to analyse the entire genome sequence of G. thermoglucosidasius NCIMB 11955, uncovering potential functional proteins in non-model strains. 2). The MetaPatchM module of BIM automates the repair of the metabolic network model. 3). The Tianjin University of Science and Technology-kcat (TUST-kcat) module predicts the kcat values of enzymes within the model. 4). The Enzyme-insert procedure constructs an enzyme-constrained model and performs a global scan to address overconstraint issues. Enzymatic data were automatically integrated into the metabolic flux model, creating an enzyme-constrained model, ec_G-ther11955. To validate model accuracy, we used both the p-thermo and ec_G-ther11955 models to predict riboflavin production strategies. The ec_G-ther11955 model demonstrated significantly higher accuracy. To further verify its efficacy, we employed ec_G-ther11955 to guide the rational design of L-valine-producing strains. Using the Optimisation Procedure for Identifying All Genetic Manipulations Leading to Targeted Overproductions (OptForce), Predictive Knockout Targeting (PKT), and Flux Scanning based on Enforced Objective Flux (FSEOF) algorithms, we identified 24 knockout and overexpression targets, achieving an accuracy rate of 87.5%. Ultimately, this led to an increase of 664.04% in L-valine titre. This study provides a novel strategy for rapidly constructing non-model strain models and demonstrates the tremendous potential of artificial intelligence in metabolic engineering.

L-isoleucine, an essential amino acid, is widely used in the pharmaceutical and food industries. However, the current production efficiency is insufficient to meet the increasing demands. In this study, we aimed to develop an efficient L-isoleucine-producing strain of Escherichia coli. First, accumulation of L-isoleucine was achieved by employing feedback-resistant enzymes. Next, a growth-coupled L-isoleucine synthetic pathway was established by introducing the metA-metB-based α-ketobutyrate-generating bypass, which significantly increased L-isoleucine production to 7.4 g/L. Upon employing an activity-improved cystathionine γ-synthase mutant obtained from adaptive laboratory evolution, L-isoleucine production further increased to 8.5 g/L. Subsequently, the redox flux was improved by bypassing the NADPH-dependent aspartate aminotransferase pathway and employing the NADH-dependent pathway and transhydrogenase. Finally, L-isoleucine efflux was enhanced by modifying the transport system. After fed-batch fermentation for 48 h, the resultant strain, ISO-12, reached an L-isoleucine production titer of 51.5 g/L and yield of 0.29 g/g glucose. The strains developed in this study achieved a higher L-isoleucine production efficiency than those reported previously. These strategies will aid in the development of cell factories that produce L-isoleucine and related products.

Shanxi aged vinegar microbiome encodes a wide variety of bacteriocins. The aim of this study was to mine, screen and characterize novel broad-spectrum bacteriocins from the large-scale microbiome data of Shanxi aged vinegar through machine learning, molecular simulation and activity validation. A total of 158 potential bacteriocins were innovatively mined from 117,552 representative genes based on metatranscriptomic information from the Shanxi aged vinegar microbiome using machine learning techniques and 12 microorganisms were identified to secrete bacteriocins at the genus level. Subsequently, employing AlphaFold2 structure prediction and molecular dynamics simulations, eight bacteriocins with high stability were further screened, and all of them were confirmed to have bacteriostatic activity by the Escherichia coli BL21 expression system. Then, gene_386319 (named LAB-3) and gene_403047 (named LAB-4) with the strongest antibacterial activities were purified by two-step methods and analyzed by mass spectrometry. The two bacteriocins have broad-spectrum antimicrobial activity with minimum inhibitory concentration values of 6.79 μg/mL-15.31 μg/mL against Staphylococcus aureus and Escherichia coli. Furthermore, molecular docking analysis indicated that LAB-3 and LAB-4 could interact with dihydrofolate reductase through hydrogen bonds, salt-bridge forces and hydrophobic forces. These findings suggested that the two bacteriocins could be considered as promising broad-spectrum antimicrobial agents.

The versatile applications of 5-aminolevulinic acid (5-ALA) across the fields of agriculture, livestock, and medicine necessitate a cost-efficient biomanufacturing process. In this study, we achieved the economic viability of biomanufacturing this compound through a systematic engineering framework. First, we obtained a 5-ALA synthase (ALAS) with superior performance by exploring its natural diversity with divergent evolution. Subsequently, using a genome-scale model, we identified and modified four key targets from distinct pathways in Escherichia coli, resulting in a final enhancement of 5-ALA titers up to 21.82 g/l in a 5-l bioreactor. Furthermore, recognizing that an imbalance of redox equivalents hindered further titer improvement, we developed a dynamic control system that effectively balances redox status and carbon flux. Ultimately, we collaboratively optimized the artificial redox homeostasis system at the transcription level with other cofactors at the feeding level, demonstrating the highest recorded performance to date with a titer of 63.39 g/l for the biomanufacturing of 5-ALA.

Shanxi aged vinegar (SAV) is a famous cereal vinegar in China, which is produced through a solid-state fermentation where diverse microbes spontaneously and complex interactions occur. Here, combined with the metatranscriptomics, the microbial co-occurrence network was constructed, indicating that Lactobacillus, Acetobacter and Pediococcus are the most critical genera to maintain the fermentation stability. Based on an extensive collection of 264 relevant literatures, a transport network containing 2271 reactions between microorganisms and compounds was constructed, showing that glucose (84% of all species), fructose (67%) and maltose (67%) are the most frequently utilized substrates while lactic acid (64%), acetic acid (45%) are the most frequently occurring metabolites. Specifically, the metabolic influence of species pairs was calculated using a mathematical calculation model and the metabolic influence network was constructed. The topology properties analysis found that Lactobacillus was the key role with robust metabolic control of vinegar fermentation ecosystem and acetic acid and lactic acid were the main metabolites with feedback regulation in microbial metabolism of SAV. Furthermore, systematic coordination of positive and negative impacts was proved to be inevitable to form flavor compounds and maintain a natural microbial ecosystem. This study provides a new perspective for understanding microbial interactions in fermented food.

Pyrroloquinoline quinone (PQQ) is one of the important coenzymes in living organisms. In acetic acid bacteria (AAB), it plays a crucial role in the alcohol respiratory chain, as a coenzyme of alcohol dehydrogenase (ADH). In this work, the PQQbiosynthetic genes were overexpressed in Acetobacter pasteurianus CGMCC 3089 to improve the fermentation performance. The result shows that the intracellular and extracellular PQQ contents in the recombinant strain A. pasteurianus (pBBR1-p264-pqq) were 152.53% and 141.08% higher than those of the control A. pasteurianus (pBBR1-p264), respectively. The catalytic activity of ADH and aldehyde dehydrogenase increased by 52.92% and 67.04%, respectively. The results indicated that the energy charge and intracellular ATP were also improved in the recombinant strain. The acetic acid fermentation was carried out using a 5 L self-aspirating fermenter, and the acetic acid production rate of the recombinant strain was 23.20% higher compared with the control. Furthermore, the relationship between the PQQ and acetic acid tolerance of cells was analyzed. The biomass of recombinant strain was 180.2%, 44.3%, and 38.6% higher than those of control under 2%, 3%, and 4% acetic acid stress, respectively. After being treated with 6% acetic acid for 40 min, the survival rate of the recombinant strain was increased by 76.20% compared with the control. Those results demonstrated that overexpression of PQQ biosynthetic genes increased the content of PQQ, therefore improving the acetic acid fermentation and the cell tolerance against acetic acid by improving the alcohol respiratory chain and energy metabolism.

The extraction, purification, qualification, and quantification of polyphenols (PPs) in vinegar are challenging owing to the complex matrix of vinegar and the specific physicochemical and structural properties of PPs. This study aimed to develop a simple, efficient, low-cost method for enriching and purifying vinegar PPs. The enrichment and purification effects of five solid phase extraction (SPE) columns and five macroporous adsorption resins (MARs) for PPs were compared. The results show that SPE columns were more effective in purifying vinegar PPs than MARs. Among them, the Strata-XA column showed a higher recovery (78.469 ± 0.949%), yield (80.808 ± 2.146%), and purity (86.629 ± 0.978%) than other columns. In total, 48 PPs were identified and quantified using SPE and gas chromatography-mass spectrometry from the SPE column extracts; phenolic acids, such as 4-hydroxyphenyllactic acid, vanillic acid, 4-hydroxycinnamic acid, 4-hydroxybenzoic acid, protocatechuic acid, and 3-(4-Hydroxy-3-methoxyphenyl) propionic acid, occupy a major position in SAV. Furthermore, considering the potential applications of PPs, the concentrates were characterized based on their bioactive properties. They exhibited high total PP, flavonoid, and melanoidin contents and excellent anti-glycosylation and antioxidant activities. These results indicate that the established methodology is a high-efficiency, rapid-extraction, and environment-friendly method for separating and purifying PPs, with broad application prospects in the food, chemical, and cosmetic industries.

Multi-microorganism solid-state fermentation (SSF) is a traditional technique to produce fermented foods. However, the fermentation kinetics is difficult to establish due to the irregular and complex growth and metabolism profiles of the microorganisms. In this work, the SSF of vinegar was described and predicted for the first time by using two-stage kinetics. The in situ and in vitro kinetics of cell growth, product formation, and substrate utilization of the predominant microorganisms Acetobacter and Lactobacillus with R-2 more than 0.98 were analyzed using Logistic, Luedeking-Piret, and Luedeking-Piret-like modes, respectively. Ethanol, lactic acid, and acetic acid were found to be the main factors responsible for the temporal variation of fermentation profiles. Potential interactions between predominant microorganisms were revealed by in vitro SSF. Acetic acid mainly produced by Acetobacter was proven the main factor for the two-stage profiles, and its modulating role for microbial growth and metabolism was more important than that of lactic acid and ethanol. The thresholds of acetic acid to the negative specific growth rate were 3.05 and 1.68 g/100 g of Cupei for A. pasteurianus and L. helveticus, respectively. These results provide the theoretical modules to understand the SSF of cereal vinegars for further monitoring and modulation.

To investigate the formation of typical melanoidin polymers, methylglyoxal (MGO) with NH3 or alanine (Ala) was used to form coloured compounds, with glyoxal or acetone used as controls. The products were characterised using chromatography, mass spectrometry, and spectroscopy. Spectroscopic results showed that the coloured compounds formed were similar to melanoidins in food. GC-MS results showed that the MGO-based reaction generated similar volatile compounds using the Maillard reaction. Mass spectrometry showed that the molecular weights of structural units in the polymers were mainly 162, 169, and 176 Da, and these could be reassembled using the basic units derived from MGO alone or in combination with nitrogen. Hence, polymers recombined using basic structural units should be considered while determining melanoidin biomarkers. The preparation of coloured compounds using MGO with NH3 can be used as a novel method to produce the control compounds for melanoidin after process optimization.