Highly effective antimicrobial films for food packaging using Schisandra chinensis extract and uncovering the contributing components via a novel deep learning approach

Abstract

The growing demand to replace synthetic antimicrobials in food packaging has driven the development of effective and safe natural alternatives. In this study, high-performance antimicrobial films were developed by incorporating Schisandra chinensis extract into a PLA-PBAT polymer matrix. Extraction was first optimized using ethanol at concentrations ranging from 25 % to 100 %, with 100 % ethanol yielding the highest antibacterial activity based on inhibition zone assays. The extract exhibited broad-spectrum antimicrobial effects against Staphylococcus aureus, MRSA, Escherichia coli, and Listeria monocytogenes, showing lower minimum inhibitory concentration and minimum bactericidal concentration values compared to conventional preservatives. A deep learning model trained on 40,259 compounds achieved 99.90 % accuracy on the training set. When applied to the 104 compounds identified in Schisandra chinensis, the model, using a probability threshold of 0.8, classified 65 as antimicrobial agents, including known compounds such as quercetin, catechin, and caffeic acid, alongside numerous understudied constituents, indicating a synergistic antimicrobial mechanism. The optimized film (SCE_AE), prepared with anhydrous ethanol as a solubilizing agent, demonstrated strong inhibition against all four pathogens over a 12 h period, and demonstrated reusability without loss of activity. In practical fruit preservation tests, the prepared film significantly extended the shelf life of strawberries and apricots, outperforming both conventional plastic wrap and silver-based antimicrobial films. This work establishes an integrated strategy combining experimental extraction, film preparation, and deep learning-based compound screening to accelerate the development of high-performance and sustainable food packaging materials.