High-throughput sequencing (HTS) and gas chromatography time-of-flight mass spectrometry (GC-TOF-MS) were used to compare the microbiota structure and metabolic compounds of traditional suancai from Heilongjiang (HLJ), Shanxi (SX) and Qinghai (QH) in China. Besides, the physicochemical properties such as total number of colonies, pH and total acid content were determined, and the related factors of the differences were analyzed. The salinity of the 3 samples was 1.9%, 8.0% and 10.0%, respectively, and the dominant bacterial genera were Loigolactobacillus, Arcobacter, and Marinomonas. Meanwhile, Loigolactobacillus was significantly positively associated with pH and nitrite in HLJ, Arcobacter was inversely related to pH and nitrite, while Marinomonas was negatively correlated with all physicochemical properties in QH which had the highest salinity. In addition, the 5 main differential metabolites in the 3 samples were acetic acid, 4-ethylphenol, 2,2,4-trimethyl-1,3-pentanediol diisobutyrate, 2,4-tert-butylphenol, and 3-butenenitrile. Among them, the ketones and acids were positively correlated with the core bacteria in HLJ with the lowest salinity, and the main genera in SX were positively associated with various alcohols, while there was a positive correlation between Marinomonas and butyronitrile alcohol in QH with the highest salinity. This study provided a guidance for the differences and correlations of microorganisms, flavor compounds, and quality characteristics from a regional perspective by studying the various quality characteristics of the suancai.

Antimicrobial peptides, as small molecular peptides with extensive antibacterial activity, have shown great potential in food preservation and other fields because of their unique antibacterial mechanism. However, traditional screening methods are time-consuming and resource-consuming, and often yield antimicrobial peptides with poor stability and high cytotoxicity, limiting their wide application. In recent years, the rapid development of artificial intelligence technology has brought new opportunities for research on antimicrobial peptides. Artificial intelligence algorithms can be continuously optimized based on prior knowledge and real-time data, which significantly improves the prediction efficiency of antibacterial peptides and reduces research and development costs. Additionally, these algorithms offer the possibility to explore the diversity of antimicrobial peptides and optimize their properties. Currently, several specialized databases have been established, providing rich resources for algorithmic model training. Furthermore, multi-source bioinformatics data such as genomics, transcriptomics and proteomics are also widely used to predict antimicrobial peptides, with a view to identifying peptides with potential antimicrobial activity more accurately. This article reviews the principles and applicability of various current artificial intelligence algorithmic models for predicting antimicrobial peptides, and explores prediction models specifically designed to address the dilemma facing the application of antimicrobial peptides. It aims to guide readers in selecting and designing artificial intelligence algorithms and to promote their innovative applications in the fields of food safety and human health.