Integrating molecular properties and olfactory receptor binding affinities for identifying environmental odorants.

Odorous volatile compounds released from waste and wastewater treatment, soil remediation, industrial manufacturing, and livestock operations have become a growing environmental and public health concern. However, identifying key odor compounds within complex emissions remains a fundamental challenge, as odor impact is governed not solely by chemical concentration, but by the odor detection thresholds (ODTs) and odor qualities. Although machine-learning approaches have shown promise in predicting odor characteristics of compounds, existing models primarily rely on molecular physicochemical properties while overlooking the biological interactions between odorants and the human olfactory system. Here, a computational framework integrating molecular descriptors, structural fingerprints, and simulated molecule-olfactory receptor (OR) binding affinities was developed to predict ODTs and odor qualities of volatile compounds. Our framework outperforms conventional physicochemical-based models, with particularly strong discriminative capability for garlic, alliaceous, and sulfurous odors-compound classes commonly associated with odor pollution. The molecular polarity, structural complexity, and OR binding affinity are further identified as prior factors for odor perception. Key ORs governing ODT and odor quality predictions are separately pinpointed. By bridging odorant-OR interaction and molecular information, this framework offers a practical tool for screening unknown environmental odorants, supporting targeted odor pollution management, regulatory threshold setting, and development of environmental monitoring strategies.