Foodborne diseases are frequently linked to the consumption of fruits and vegetables, with the virulence genes (VGs) carried by Escherichia coli (E. coli) being one of the primary contributors to these illnesses. However, the trends of VGs in E. coli from fruits and vegetables remain poorly understood. This study analyzed the genome sequences of 1084 E. coli from fruits and vegetables collected worldwide between 2000 and 2023. The results showed a significant upward trend in VGs over time, peaking at 83.11 in 2019. The major changes were observed in toxin-producing and delivery system. Moreover, the number of plasmids and VGs increased significantly over time, suggesting a rising risk of transmission of VGs. Human-induced factors, including the increase of CO₂ concentration and pesticide use, led to the change of VGs of E. coli, further exacerbating food safety risks. Furthermore, the multiple regression model which can predict VGs showed that reduced CO₂ emissions by 20%, VGs could be reduced by 4.85. Among different types of produce, the isolation of E. coli from leafy vegetables exhibited the highest levels of VGs, with spinach (86.86) and lettuce (75.95) showing the highest concentrations. This trend correlates strongly with the frequency of E. coli induced outbreaks, indicating an emerging public health risk.

The analysis of non-volatile compounds in baijiu is still a research gap, especially the exploration of the differences between commercial baijiu with blending base baijiu is almost none. In this study, non-targeted metabolomics was used to analyze the non-volatile compounds of different blending baijiu samples, and a total of 861 non-volatile compounds were identified. Based on differential metabolite analysis, Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis, random forest, and sensory analysis, the differences and correlation between the blending samples and the commercial baijiu were explored, in addition to 7 pathways and 23 distinct markers that were found to be relevant to the formation of baijiu metabolites. The present study is valuable for the research on the blending process and quality control of baijiu.

Riboflavin is the precursor of flavin mononucleotide (FMN) and flavin adenine dinucleotide (FAD), and is widely used in food, feed, and pharmaceutical industries. However, riboflavin production by microbial fermentation in Bacillus subtilis is still limited by the complex multilayered regulation of riboflavin biosynthesis. In this study, we systematically engineered B. subtilis to improve riboflavin production. First, the riboflavin biosynthetic pathway was strengthened by relieving FMN-mediated feedback inhibition, replacing the native promoter of the rib operon, and optimizing ribA copy number, which increased the riboflavin titer to 2144.6 mg/L. Next, the supply of the precursors ribulose-5-phosphate (Ru5P) and guanosine triphosphate (GTP) was enhanced by reinforcing the pentose phosphate and purine pathways, and competing branch pathways were dynamically regulated using an FMN-responsive switch. In particular, dynamic regulation of purA and pgi further increased the titer by 31.4% to 4.6 g/L. Finally, guided by the genome-scale metabolic model etiBsu1209, we combined target prediction with CRISPRi library screening and identified co-repression of ilvD and guaC as an effective strategy, increasing the titer to 6.6 g/L. After further optimization of cultivation conditions, the final strain produced 8.4 g/L riboflavin in shake flasks and 20.2 g/L in a 5 L bioreactor. These results provide an effective strategy for riboflavin overproduction in B. subtilis and for engineering complex metabolic pathways.

Aging is a progressive decline of physiological functions and raises the risk of many kinds of chronic diseases, imposing a heavy burden on the global public health. Although considerable progress has been made in understanding the genetic mechanisms underlying aging, converting this knowledge into effective interventions to delay aging remains a challenge. Recent research shows that changes in gut microbiota composition and function are tightly linked to senescence, making gut microbial regulation a promising strategy for healthy aging. This review thoroughly summarizes the bidirectional crosstalk between aging and gut microbiota and the mitigating effects of conventional and next-generation probiotics on the aging progression, and focuses on the mechanisms by which probiotics act on the hallmarks of aging. Finally, we explore the strategies for the development and functional enhancement of anti-aging probiotics, aiming to provide novel insights for the precise design of probiotics that promote healthy aging.

Advanced glycation end products (AGEs) generated during thermal processing have raised food-safety concerns because of their potential links to oxidative stress, inflammation, and chronic disease risk. Natural inhibitors have attracted substantial attention, but current evidence remains centered largely on polyphenols, whereas the role of polysaccharides from food-medicine homology (FMH) sources in food AGE control has not been reviewed systematically. This review outlines the formation pathways, representative markers, and dietary relevance of AGEs in food systems and integrates current evidence on FMH-derived polysaccharides with respect to antiglycation mechanisms, structure–activity relationships, source-specific findings, and food-system applications. Available studies indicate that polysaccharide preparations may reduce AGE formation through reactive dicarbonyl trapping, competition for glycation sites, metal-ion chelation, antioxidative protection, and regulation of protein conformation. Among FMH sources, Ganoderma lucidum, Dendrobium officinale, Siraitia grosvenorii, and hawthorn provide relatively direct antiglycation evidence, whereas Angelica sinensis, Polygonatum spp., chrysanthemum, and several other catalog-listed materials are supported mainly by preliminary, indirect, or putative evidence. Limited food-system studies suggest that some FMH-derived polysaccharides can reduce AGE accumulation while also affecting texture, stability, or sensory quality. FMH-derived polysaccharides may therefore serve as multifunctional ingredients for low-AGE food design, but practical application requires more standardized evaluation, stronger process-structure-activity evidence, and models that better approximate real dietary exposure.

Staling causes quality deterioration of starchy food (SF) products, often leading to serious food waste due to untimely identification of staling extent. However, the staling assessment of SF products cannot be performed promptly in the supply chain stage because current rapid detection systems are not available for sensing the indices related to staling. In this work, we developed a staling assessment system based on a triboelectric sensor enabled by deep learning. This device captures texture dynamics of SF products in situ by converting the mechanical changes, characterized by increased hardness and decreased elasticity, into distinct electrical signals. To achieve the high sensitivity required for sensing these variations, the sensor features a hierarchical hollow carbon nanofiber (HCNF) electrode that integrates zero-dimensional hollow spheres within one-dimensional nanofibers. This architecture creates nanoscale cavity confinement to suppress charge decay, while providing an ultrahigh surface area (1601.92 m²/g) to enhance signal output. The collected signals were used to train a convolutional neural network (CNN), which achieves high recognition accuracies of 96.5%–98.1% in samples with different aging rates. Notably, the proposed sensor exhibited good compatibility with the packaging materials commonly utilized for SF, suggesting that this system could non-destructively detect the staling levels of SF in the supply chain to effectively mitigate food waste.

Food-derived bioactive peptides (FBPs) are a key focus of functional food research due to their potential for health interventions in areas such as angiotensin-converting enzyme (ACE) inhibition, antioxidant activity, antibacterial activity, anti-hyperglycemic effects, immunomodulation, and anticancer activity. Faced with the vast sequence space composed of 20 natural amino acids (the theoretical combinations of 10-residue peptides alone reach approximately 10¹³), the traditional “enzymatic hydrolysis–identification–screening” paradigm faces bottlenecks in terms of efficiency and cost; AI offers a new solution. This study systematically reviewed 467 publications from 2021 to 2026, identifying 247 highly relevant studies (including 229 research papers and 18 reviews). Thematic analysis based on Term Frequency-Inverse Document Frequency (TF-IDF) weighting and hierarchical clustering reveals that existing work is highly concentrated on “sequence–activity prediction” (86.6%) and “virtual proteolysis prediction” (57.1%), while research centered on “peptide structure design” accounts for only 2.4%, highlighting the current imbalance in task distribution. This paper analyzes three major flaws in the existing evaluation paradigm: sequence similarity leakage caused by random partitioning, the systematic absence of independent external validation, and the limitations of post-hoc attribution tools such as Shapley additive explanations (SHAP) in guiding sequence optimization. In terms of design approaches, structure-guided directed release aligns more closely with the industrial pathway for natural protein hydrolysates. Although generative de novo design offers a broader space for sequence exploration, it remains constrained by food-specific annotated data, multi-constraint modeling, and industrial production costs. Accordingly, future research should focus on: (1) Establishing standardized evaluation guidelines based on sequence clustering. (2) Developing food-protein-specific characterization models for short peptides of 2–15 amino acids. (3) Constructing an end-to-end closed-loop design framework that explicitly incorporates sensory and gastrointestinal stability constraints. This paper argues that the value proposition of AI should gradually shift from “efficient screening assistance” toward “robust, rational design.”

Nisin demonstrates that ribosomally synthesized and post-translationally modified antimicrobial peptides can become practical food preservatives, but its success does not provide a universal route for next-generation peptide antimicrobials. Commenting on the compartmentalized lanthipeptide biosynthesis platform reported by Suo et al., this article argues for biosynthesis-first engineering, defined here as evaluating antimicrobial activity together with pathway compatibility, modification fidelity, host tolerance, recoverability, and food-matrix performance from the outset, rather than treating production and application constraints as downstream considerations. For lanthipeptides, antimicrobial activity must be considered together with precursor expression, post-translational modification, maturation, host compatibility, recovery, food-matrix stability, sensory neutrality, regulatory feasibility, and production cost. A sequence-to-biosynthesis-to-food framework can help distinguish peptides that are merely active in vitro from those that are realistic candidates for food preservation.