Pallante L., Cannariato M., Vezzulli F., Malavolta M., Lambri M., Deriu M.A., 2023, Machine Learning Aided Molecular Modelling of Taste to Identify Food Fingerprints, Chemical Engineering Transactions, 102, 283-288.
Nature has developed fascinating mechanisms for selecting and monitoring nutrients through refined systems for food intake and uptake. One of the most important is the sense of taste. Taste is an emergent property involving a complex network of multilevel biological interactions beginning with the activation of specific protein receptors as a consequence of interaction with food molecules. In this context, crucial information about the mechanisms underlying the functioning of taste can be obtained by using molecular mechanistic modelling and machine learning tools borrowed from the field of drug design and the study of structural biology and protein biophysics. The ultimate goal is to develop predictive models capable of studying the intricate connection of molecular, sub-cellular and cellular phenomena underlying the complex biological mechanisms that regulate the relationships between food constituents and perceived taste. Artificial intelligence-driven digital tools for taste prediction and the study of molecular features of the interaction between food molecules and taste receptors have been recently developed by our group. Such tools are the operating engines of the decision support tool developed during the VIRTUOUS project (https://virtuoush2020.com). In this work, these tools were used to generate molecular fingerprints of coffee starting from its chemical composition. Through methods that integrate molecular modelling techniques and machine learning, molecules extracted from coffee were characterized in terms of binding affinity, specificity, and selectivity toward bitter receptors. The targeting ability of coffee-extracted molecules for human TAS2Rs was studied with an atomistic resolution to obtain a virtual fingerprint that links the molecular structure of food ingredients with their bitter profile. The study fits within the digital transition vision that leverages modelling and computational approaches to develop decision-supporting tools for developing solutions in the areas of nutrition, health and the modern food industry.
