Advanced AI-driven cheminformatics models for characterizing air pollutants and predicting interactions with biological targets

Health and environmental risks linked to air pollution represent a major concern. According to the World Health Organization, 9 out of 10 people worldwide are exposed to elevated pollutant concentrations. This exposure has been associated with deleterious effects on vital organs, contributing to chronic respiratory diseases, oxidative stress, hypertension, and a decline in cognitive functions. However, to gain a deeper understanding of the role and impact of pollutant molecules on complex disorders, their molecular properties, biological activities, and toxicological effects need to be further explored. In this context, this PhD thesis proposes two computational tools, APDB and APBIO, aimed at comprehensively characterizing pollutants and finding potential associations with human biological targets. APDB is an online database that collects and analyzes air pollutants by investigating their physicochemical, structural, and quantum mechanical properties in detail. By considering a wide range of molecular properties, APDB provides an insightful analysis of air pollutant similarity. It also offers a publicly available web interface that enables browsing data by category, visualizing and downloading molecular structures, computed descriptors, and predicted similarities. Leveraging the collected pollutant molecules and their associated targets, APBIO implements a chemogenomic approach that integrates molecule bioactivity signatures and target sequence descriptors to train machine learning classifiers, subsequently used to capture and infer unobserved interactions. APBIO derives and employs more sophisticated representations of pollutants that incorporate not only chemical information but also knowledge about their biological traits, demonstrating robustness in predicting associations between target proteins and pollutant molecules. These resources provide valuable knowledge on the characteristics and properties of air pollutants, along with an in-depth exploration of their similarities. Furthermore, potential candidates for experimental validation can be identified and suggested to uncover relationships between pollutant exposure and the harmful effects on human health.