From Phytochemical Complexity to Biological Response: Explainable Modeling of the Antimicrobial Activity of Cannabis sativa Resin

Full-spectrum Cannabis sativa resins are complex phytochemical matrices whose antimicrobial activity may vary according to the microorganism, treatment concentration, and biological phenotype evaluated. This study investigated the antibacterial and antifungal activities of Cannabis sativa resins and their ability to disrupt pre-formed biofilms of five chemically characterized cannabis resins and evaluated the relative contribution of selected experimental variables to the observed biological responses using inferential statistics and interpretable machine learning. Cannabinoid and terpene profiles were characterized by HPLC-PDA and gas chromatography, respectively. Antimicrobial activity was evaluated against Staphylococcus aureus, Escherichia coli, Pseudomonas aeruginosa, Acinetobacter baumannii, Candida albicans and Saccharomyces cerevisiae. The resins produced microorganism-, concentration-, and phenotype-dependent responses, with complete planktonic growth inhibition observed only for S. aureus and more heterogeneous responses against pre-formed biofilms. Generalized linear models identified significant effects of the experimental factors, whereas supervised machine-learning models achieved high predictive performance, particularly for planktonic growth (R2 ≈ 0.99). Permutation Importance and SHAP analyses revealed that microbial identity provided the greatest predictive information, followed by treatment concentration, whereas resin identity exerted a smaller, phenotype-dependent contribution. These findings indicate that, within the experimental domain evaluated, microorganism identity and treatment concentration provided greater predictive information than resin identity. Integrating phytochemical characterization with microbiological phenotyping and interpretable modeling of experimental variables provides a useful framework for investigating the biological behavior of complex natural-product matrices.