Multidimensional chromatography has been successfully applied for many years as core technology for odorant patterns characterization in the flavor and fragrance field. Very recently a sensomics-based expert system (SEBES) [1] capable to predict key-aroma signatures of food without using human olfaction has been implemented in a comprehensive two-dimensional gas chromatography (GC×GC) platform. The strategy, also referred to as Artificial Intelligence Smelling, conceptually opens many opportunities for odorants pattern recognition, accurate quantification avoiding time-consuming sample preparation/extraction steps, and samples sensory qualification/discrimination based on computer vision strategies. Moreover, due to the information density of each analysis, fingerprinting can be extended to different samples features (e.g., origin traceability, shelf-life evolution, processing impact etc.). The contribution illustrates the potentials of MDGC platforms in the context of Artificial Intelligence Smelling and extended untargeted/targeted fingerprinting through a key-food product for confectionery industry: high-quality hazelnuts (Corylus avellana L.). Sensory quality of raw and roasted hazelnuts depends to the presence of non-volatile precursors patterns and to (key)-aroma compounds in well-balanced proportions. Moreover, odorants deriving by lipid oxidation processes and/or enzyme-catalyzed reactions carried out by bacteria and molds [2,3] might evoke unpleasant notes with detrimental impact on kernels quality. Odorant patterns, strongly correlated to sensorial qualities, can be effectively detected by computer vision with composite-class images generated combining 2D chromatographic signals from samples showing specific features (e.g., good reference kernels, spoiled kernels with mould, rancid, stale, solvent-like odors). On the other hand, the comprehensive exploration of the detectable volatilome (including all volatiles) by untargeted/targeted fingerprinting drives machine learning toward samples identitation highlighting botanical/geographical signatures [4], shelf-life trajectories, and climate impact on hazelnut metabolome. The virtuous synergy between Academic research and industry defines new quality concepts, accelerates innovation, and shapes the future of our food. The key is the implementation of modern omics concepts and strategies that bring food analysis closer to flavor characterization.
Artificial Intelligence smelling: Can multidimensional chromatography play a (key) role?
Chiara Cordero
;Simone Squara;Andrea Caratti;Carlo Bicchi
2022-01-01
Abstract
Multidimensional chromatography has been successfully applied for many years as core technology for odorant patterns characterization in the flavor and fragrance field. Very recently a sensomics-based expert system (SEBES) [1] capable to predict key-aroma signatures of food without using human olfaction has been implemented in a comprehensive two-dimensional gas chromatography (GC×GC) platform. The strategy, also referred to as Artificial Intelligence Smelling, conceptually opens many opportunities for odorants pattern recognition, accurate quantification avoiding time-consuming sample preparation/extraction steps, and samples sensory qualification/discrimination based on computer vision strategies. Moreover, due to the information density of each analysis, fingerprinting can be extended to different samples features (e.g., origin traceability, shelf-life evolution, processing impact etc.). The contribution illustrates the potentials of MDGC platforms in the context of Artificial Intelligence Smelling and extended untargeted/targeted fingerprinting through a key-food product for confectionery industry: high-quality hazelnuts (Corylus avellana L.). Sensory quality of raw and roasted hazelnuts depends to the presence of non-volatile precursors patterns and to (key)-aroma compounds in well-balanced proportions. Moreover, odorants deriving by lipid oxidation processes and/or enzyme-catalyzed reactions carried out by bacteria and molds [2,3] might evoke unpleasant notes with detrimental impact on kernels quality. Odorant patterns, strongly correlated to sensorial qualities, can be effectively detected by computer vision with composite-class images generated combining 2D chromatographic signals from samples showing specific features (e.g., good reference kernels, spoiled kernels with mould, rancid, stale, solvent-like odors). On the other hand, the comprehensive exploration of the detectable volatilome (including all volatiles) by untargeted/targeted fingerprinting drives machine learning toward samples identitation highlighting botanical/geographical signatures [4], shelf-life trajectories, and climate impact on hazelnut metabolome. The virtuous synergy between Academic research and industry defines new quality concepts, accelerates innovation, and shapes the future of our food. The key is the implementation of modern omics concepts and strategies that bring food analysis closer to flavor characterization.File | Dimensione | Formato | |
---|---|---|---|
Chiara Cordero Talk_v2.pdf
Accesso aperto
Descrizione: Presentation Slides
Tipo di file:
PDF EDITORIALE
Dimensione
10.05 MB
Formato
Adobe PDF
|
10.05 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.