Perugino, Florinda (2025) Computer-driven analysis of taste and aroma compounds, bioactives and toxicants in food: a possible blueprint to piece the analysis of relevant chemicals together. [Tesi di dottorato]

Documento PDF perugino_florinda_38_COMPLETO.pdf Download (14MB)

Abstract

In the last decades, the world has experienced a rapid development of technology resulting in the integration of computational strategies and Artificial Intelligence (AI) also in food chemistry, biology, food science, and related disciplines. Understanding the interaction between macromolecules (i.e. proteins) and small molecules from a mechanistic point of view represents a critical point to elucidate the biological processes underpinning food safety, food perception, and food technology. The integration of AI-based techniques in computational chemistry has already provided advances in pharmacology. Starting from the assumptions that: i) many pharmacologically relevant targets are also involved in the interaction with molecules of food origin; ii) the biological effects of a specific compound depend on its interaction with one or more binding sites of the target protein, the same computational techniques that are widely applied in pharmaceutical research could be successfully applied also to food science. Among all the compounds relevant to food science and food safety, this work investigates three macro categories: taste-active compounds, bioactive compounds, and toxic compounds (with a focus on mycotoxins). This research project aims to represent a proof-of-principle to refine the current understanding of food-borne molecules leveraging on advanced computational methods including artificial intelligence. For this reason, studies on taste compounds have been conducted focusing not only on their capability to elicit taste and aroma, but also on their potential bioactivity and on their possible “side effects”. In this context, the interaction between food-related toxic compounds and taste receptors type 2 (TASRs) was investigated performing a Machine Learning (ML) based virtual screening, providing insights into the potential involvement of extraoral TASRs (especially those expressed in the gut) in the mechanism of toxicity of mycotoxins, such as trichothecenes (i.e. DON and 15-Ac-DON). However, mycotoxins may exert several other adverse effects on human and animal health. Investigating and elucidating effects still largely uncharacterized from a mechanistic point of view may represent a valuable strategy to both strength risk assessment and guarantee food safety. In this perspective, the interaction and possible resulting effects of others concerning mycotoxins within relevant targets has been investigated by applying computational pipelines mostly made up of virtual screening, molecular docking and molecular dynamics simulation. The impact of different single nucleotide variants (SNVs) on the Aromatase enzyme (CYP19A1) was studied revealing that some variants are even more prone to inhibition by zearalenone and α-zearalenol while others are potentially inactive compared to the wild type. Another mycotoxin worthy of major concern is Ochratoxin A (OTA). Its intricate toxicological profile involves multiple mechanisms including oxidative stress, impaired cellular defense and synthesis, and the induction of carcinogenic, nephrotoxic, and neurotoxic effects. Concerning the effects of OTA on protein synthesis, OGFOD1 - a key player in protein translation - was identified as a potential target for both OTA and its thermal degradation product 2’R-OTA. Also, its potential role in Parkinson’s disease has been elucidate from a molecular standpoint proving the ability of OTA and certain congeners to interact with some enzymes involved in the Parkinson’s Adverse Outcome Pathway (i.e. Cathepsin L/D and Complex I). Furthermore, the advancement achieved by AI in protein structure prediction enabled the homology modelling of human Ceramide Synthase 5 (CerS5), a known target of fumonisins. This led to the in-depth investigation of the duality of fumonisins toward CerSs revealing that HFB1 can undergo the reaction through both catalytic mechanisms proposed (i.e. one-step and two-step catalytic mechanism), while FB1 preferentially undergoes the reaction through the two-steps mechanism. Concerning bioactive compounds, this work investigated for the first time to the best of our knowledge the capability of L-peptides and D-peptides to interact with GPR120 (Free Fatty Acid Receptor 4). Starting from this evidence, a CatBoost model to apply as a prefilter for docking analysis was developed with the aim of screening and identifying new GPR120 targets from huge chemical compounds libraries. In conclusion, the main goal achieved in this PhD thesis are i) the identification of a connection between taste receptors and toxic and bioactive compounds; ii) the in depth investigation of the mechanic underpinning the toxicity of different mycotoxins; iii) the progressive implementation of AI-based techniques which proved to be a powerful strategy to empower and accelerate the investigation also reducing time and costs.

Downloads

Actions (login required)

Modifica documento