|
Abstract
|
In this study, we examined 43 Lactiplantibacillus plantarum isolates from Iranian traditional dairy products to
elucidate the correlation between plantaricin gene profiles and antimicrobial activity. Twelve target pln genes
were screened using polymerase chain reaction (PCR), and inhibition against six pathogens—Bacillus cereus,
Yersinia enterocolitica, Salmonella enterica, Listeria monocytogenes, Escherichia coli, and Staphylococcus aureus—was
assessed through reciprocal dilution assays. The inhibition zone sizes ranged from 5 mm to 20 mm, with the most
significant inhibition observed in B. cereus and Y. enterocolitica. Statistical analysis revealed that the Random
Forest model achieved an accuracy of 85% and an R² of 0.62 in predicting antimicrobial activity, indicating a
robust predictive capability.
Random Forest modeling was utilized to establish a connection between gene presence-absence data and
antimicrobial performance. Strains exhibiting broad pln repertoires (e.g., isolates 405 and 465) demonstrated
strong, multi-target inhibition, while some with limited gene sets, such as isolate 954 containing only plnG,
displayed potent but highly specific activity. Across various pathogens, plnB consistently emerged as the most
significant predictor, with plnD, plnC, and plnG contributing in a pathogen-specific manner. The models
accounted for 50–65% of the variance in activity (R² = 0.50–0.65), indicating a reliable predictive capability.
These findings suggest that the genetic composition of the plantaricin cluster is a crucial determinant of
antimicrobial behavior in L. plantarum. The integration of targeted genomics with machine learning offers a
practical approach for selecting strains optimized for food preservation and probiotic applications, with the
potential to reduce reliance on chemical preservatives.
|