Artificial intelligence-assisted QSAR modeling of phenolic monoterpenes targeting Helicobacter pylori.

As a gram-negative bacterium implicated in various gastrointestinal problems, Helicobacter pylori is classified by the World Health Organization as a group 1 carcinogen. Considering multidrug resistance, particularly to clarithromycin and metronidazole, the efficacy of standard eradication regimens has been compromised causing the urgent need for novel therapeutic agents. Phenolic monoterpenes, volatile plant-derived metabolites such as thymol, carvacrol, vanillin and eugenol, exhibit promising antimicrobial activity against H. pylori however, with limitations due to poor solubility and rapid metabolism. Their small molecular size, reactive functional groups, and lipophilicity enable structural modifications that enhance pharmacokinetic and biological profiles. The present study is focused on development of non-linear quantitative structure-activity relationship (QSAR) modeling of four series of phenolic monoterpene derivatives applying artificial neural networks (ANNs) after linear models failed to provide reliable predictions. The four series of monoterpene derivatives based on eugenol, vanillin, carvacrol and thymol structures were subjected to pattern recognition analysis including hierarchical cluster analysis and clustering based on ANNs. Pattern recognition analysis revealed (dis)similarities in antibacterial activity and provided complementary insight into compounds grouping beyond scaffold identity highlighting cross-series activity-based groupings that served as an experimental reference for ANN-QSAR modeling. The developed ANN-based QSAR models offer predictive accuracy, supporting rational design and optimization of novel phenolic monoterpene derivatives with stronger antibacterial potential against H. pylori.