Tissue-specific antibacterial and anti-biofilm activities of Penthorum chinense Pursh: integration of metabolomics and Random Forest machine learning

Penthorum chinense Pursh is a traditional medicinal and edible plant rich in flavonoids and complex polyphenols. Evaluating its aqueous extracts against foodborne pathogens via biomass assay, microplate spread tests and scanning electron microscopy revealed distinct tissue-specific efficacies: leaf and stem extracts exhibited strong bactericidal effects (16–128 mg/mL), notably against Bacillus cereus, whereas flower extracts demonstrated potent biofilm inhibition (IC50 < 1 mg/mL). Broad-coverage UPLC–ESI–Q–TOF–MS metabolite profiling annotated 93 compounds, and integration with Random Forest nested cross-validation prioritised candidate constituents associated with the measured activities. While confirming established antimicrobials (gallic acid, pinocembrin, pinostrobin, (-)-epicatechin gallate, afzelechin, corilagin, and hydrolysable derivatives of quercetin, kaempferol, and liquiritin) as major active constituents, our ML model first revealed several unreported bioactive compounds. Specifically, turmeronol A, vanillic acid glucoside, lotaustralin, and 2’,6’-dihydroxydihydrochalcone-4’-O-(3′’-O-galloyl)-β-D-glucopyranoside act as crucial bactericidal agents, while Gemin D, lamiophlomiol C, 4’-hydroxyflavanone, and pinocembrin-7-O-[4,6-HHDP]-glucoside primarily contribute to anti-biofilm efficacy. Finally, their multi-target mechanisms were preliminarily explored.