Screening of potential probiotic lactic acid bacteria with antimicrobial properties and selection of superior bacteria for application as biocontrol using machine learning models

Abstract

144 lactic acid bacteria (LAB) strains were screened to find those with the best antimicrobial activity and potential probiotic properties. Firstly, bacteria were tested for resistance to low pH and high bile salt, antimicrobial properties against the common agent of bacterial contamination in plant tissue cultures. Then, the selected strains were assayed for probiotic characteristics including autoaggregation, coaggregation, hydrophobicity, biofilm formation, and adhesion ability. Finally, ranking and selection of probiotic potential bacteria for harnessing as antibacterial agents in plant tissue cultures were performed using supervised machine learning models. The highest values of autoaggregation percentages were observed for Lactobacillus paracasei S23 (96.1%), while the highest coaggregation values with Staphylococcus aureus, Bacillus cereus, and Listeria monocytogenes were observed for strains Lactobacillus plantarum S57 (63.4%), L. plantarum S70 (57.4%), and Lactobacillus casei S81 (58.6%), respectively. Principal component analysis (PCA) and machine learning algorithms allowed to select and prioritizing of 15 out of 144 LAB strains, based on their probiotic potential properties or their antimicrobial activity, of which four strains including Lb. paracasei S23, Lb. plantarum S70, Lb. casei S81, and Lb. plantarum S57 were proposed as appropriate antimicrobial candidates for harnessing in plant tissue cultures.