Advancing Food Volatilomics with AI: From Data Visualization to Sensory Prediction

Artificial intelligence (AI) is expanding rapidly and is expected to significantly impact analytical chemistry [1]. Food analysis represents a challenging field for AI applications due to the complex composition of samples, comprising primary and secondary metabolites, processing by-products, microbial metabolites, xenobiotics, and contaminants, all influencing functional and sensory properties. Multidimensional analytical systems, combining chromatographic separation with mass spectrometry (MS)-based structural characterization, offer robust platforms for analyzing such matrices. However, fully leveraging the large datasets generated by multidimensional analysis (MDA) requires non-conventional, AI-assisted approaches to advance food characterization beyond conventional quality indicators. This work explores the application of AI methods for processing and interpreting complex datasets produced by comprehensive two-dimensional gas chromatography (GC×GC) in foodomics. AI-driven approaches streamline data handling, enhance chemical marker discovery, and improve overall interpretative efficiency. The multidimensional structure of GC×GC datasets enables the application of Computer Vision and Augmented Visualization tools for advanced decision-making. These strategies compensate for retention-time variations and correct image distortions, supporting more accurate comparisons. Additionally, selective mass spectral mining amid co-elution and alignment inconsistencies allows for targeted identification of ünknown-knownsänd facilitates aroma prediction through the ”AI-smelling machine” concept [2,3]. Case studies illustrate the potential and challenges of applying AI in GC×GC-MS data analysis, particularly in food volatilomics [4] and sensomics [5]. These developments move analytical chemistry closer to establishing direct links between chemical fingerprints and sensory attributes — effectively enabling us to ”see how it smells.”