Integrating Machine Learning and Pharmacophore Features for Enhanced Prediction of H1 Receptor Blockers

Histamine Type I Receptor Antagonists (H1 blockers) are widely used to mitigate histamine-induced inflammation, particularly in allergic reactions. Histamine, a biogenic amine found in endothelial cells, vascular smooth muscle, bronchial smooth muscle, and the hypothalamus, plays a key role in these responses. H1 blockers are essential components of cough syrups and flu medications. They are classified into two generations: first-generation H1 blockers, which are sedating and associated with numerous side effects, and second-generation blockers, which are non-sedating, generally less toxic, but may still exhibit cross-reactivity with other receptors.

In this study, a comprehensive database of compounds was utilized, with fexofenadine serving as a benchmark to discover compounds with potentially superior efficacy and reduced side effect profiles. In particular, multidimensional K-means clustering, a machine-learning technique, was applied to identify compounds with chemical structures similar to fexofenadine.

Utilizing computational prediction of pharmacokinetic profile and molecular docking experiments, the action of these drugs on the H1 receptor was assessed. Furthermore, the crossreactivity of antihistamines was investigated by conducting a structure-based pharmacophore feature analysis of the docked poses of highly toxic antihistamines with various receptors.

By identifying and proposing the removal of common toxic features, this study aims to facilitate the development of antihistamines with reduced adverse effects.