Taste physiological mechanisms and their health implications, analyzed by genetic, epigenetic, and artificial intelligence analyses

Taste perception is a complex trait influenced by genetic, epigenetic, physiological, and pathological factors, with significant implications for nutrition, health, and disease. Our studies integrated molecular biology, psychophysics, and advanced machine learning methods to explore inter-individual variability in taste function. We demonstrated that supervised learning algorithms, such as CatBoost and Random Forest, can automatically identify TAS2R38 genotypes and accurately analyze taste and oral sensitivity in healthy individuals and patients with chemosensory loss, highlighting the differential contribution of gustatory, olfactory, and trigeminal components. In parallel, convolutional neural networks enabled the reliable and automatic detection of fungiform papillae, overcoming the limitations of the manual method. On the molecular level, we demonstrated that methylation of the Gustin gene influences salivary protein levels and fungiform papillae density, while methylation of the TAS2R38 gene is associated with COVID-19 severity, highlighting the role of epigenetic regulation in taste physiology and the immune system. In addition, we investigated taste impairments in Parkinson’s disease (PD) and their association with α-synuclein (SNCA) gene polymorphisms. Taste deficits in PD were modality-specific, particularly affecting saltiness and astringency, and were modulated by specific SNCA genotypes. Supervised learning models classified PD versus healthy controls and stratified disease severity, identifying key sensory and genetic predictors. Taken together, these findings demonstrate how the integration of computational approaches, genetic/epigenetic profiling, and psychophysical data can advance the understanding of human taste function, offering innovative perspectives for personalized nutrition, clinical diagnostics, and therapeutic applications.