Unravelling the human taste receptor interactome: machine learning and molecular modelling insights into...

The understanding of the molecular mechanisms that drive taste perception can have broad implications for public health. This study aims to expand the understanding of taste receptor-associated molecular pathways by resolving the taste receptor interactome. To this end, we propose a comprehensive machine learning approach to accurately predict and quantify protein-protein interactions using an ensemble evolutionary algorithm. 1,647,374 positive and 894,213 negative experimentally verified protein-protein interactions were mined and characterized using 61 functional orthology, sequence, co-expression and structural features. The binary classifier significantly improved the accuracy of existing methods, reconstructing the full taste receptor interactome and was combined with a regressor that estimates the binding strength of positive interactions. Molecular dynamics investigation of top-scoring protein-protein interactions verified novel interactions of TAS2R41. The reconstructed TR interactome can catalyze the study of molecular pathophysiological mechanisms related to taste, the development of flavorsome nutrient-dense food products and the identification of personalized nutrition markers.