An integrated experimental and deep learning modeling approach to harness plant-derived antimicrobials for biodegradable food packaging films

Abstract

The significant threat of microbial contamination to food safety necessitates the development of innovative antibacterial packaging strategies. To address this challenge, this study developed biodegradable antibacterial films using the solution casting method, incorporating plant extracts into biodegradable polymer matrices such as PLA, PBAT, and PVA. Extracts from 13 plant species were obtained using various solvents and extraction procedures, and their antimicrobial efficiency was assessed through inhibition zone tests. Among them, Syzygium aromaticum, Schisandra chinensis, Punica granatum peel, Scutellaria baicalensis, and Prunus mume exhibited strong antibacterial activity against common foodborne pathogens, including S. aureus, MRSA, L. monocytogenes, and E. coli. A Transformer-based deep learning model identified 175 compounds in Prunus mume and 152 compounds in Scutellaria baicalensis as potential antibacterial agents, using a probability threshold of 0.99. Further analysis revealed that organic acids in Prunus mume and phenolic compounds in Scutellaria baicalensis were the main antimicrobial constituents, indicating distinct antibacterial pathways. The resulting films demonstrated significant food preservation effects on beef samples and desirable biodegradability after six months of soil burial. This study provides an effective strategy that integrates experimental research with artificial intelligence to advance the development of plant-based antibacterial packaging.