Phenolic Profile of Kitaibelia balansae Leaf Extract and Machine-Learning Modelling of Effects on Pathogenic and Lactic Acid Bacteria

Abstract

This study followed a three-stage approach. First, phenolic compounds were extracted from Kitaibelia balansae leaves using ultrasound-assisted extraction (UAE), and the extraction parameters were optimised. The targeted phenolic profile of the optimised extract was analysed, and its effects on two pathogenic bacterial species and two lactic acid bacterial strains were monitored over 24 h. Syringic acid showed the highest measured concentration among the targeted compounds. Relative to the solvent-matched controls, extract-treated cultures showed lower growth measurements for the pathogenic bacteria and higher growth measurements for the selected lactic acid bacteria under the tested conditions. These findings represent microorganism-dependent responses to the whole extract and do not establish that individual phenolic compounds caused the observed effects. In the present evaluation, machine-learning models outperformed MLR, with GPR showing the best performance. Independent biological validation is needed to confirm broader generalizability.