Boosting comprehensive two-dimensional gas chromatography with Artificial Intelligence: Computer Vision helps to see what we smell

Artificial intelligence (AI) is an exponentially expanding field poised to transform the landscape of analytical chemistry [1]. As an intriguing test bench for AI tools and concepts food applications offer many challenges due to the compositional complexity of samples (e.g. primary metabolites secondary/specialized metabolites processing derivatives exogenous compounds from microbial communities presence of xenobiotics and contaminants etc) and the properties connected to specific yet unique chemical patterns. Multidimensional analysis systems combining physicochemical discrimination of individual components by chromatography with the structure-elucidation potential of MS are the ultimate solutions for comprehensive investigations. However the analytical data generated by MDA platforms have to be fully exploited by applying non-conventional approaches supported by the AI potential to take a further step ahead and generate new knowledge on food properties going beyond current quality indexes. The contribution aims to illustrate how AI techniques can efficiently explore the complex datasets derived from comprehensive two-dimensional gas chromatography (GC×GC) in the application area of food-omic. By harnessing the power of AI we highlight the potential for streamlined and insightful data processing based on different kinds of features. Through this exploration we showcase the transformative impact of AI in deciphering the complexities of GC×GC data emphasizing efficiency and precision in food markers discovery. Thanks to the multidimensional nature of GC×GC data arrays Computer Vision and Augmented Visualization open new opportunities for decision-making strategies. By these advanced yet peculiar AI tools retention-time shifts that cause image distortion are efficiently compensated thereby facilitating comparative visualization and related strategies. Moreover, by actively exploiting mass spectral signatures in the presence of co-elutions and misalignments unknown-knowns can be specifically targeted and by identifying key food odorants in the presence of interferents aroma signatures can be predicted by an AI-smelling machine [2,3]. Practical examples showcase the challenges and potentials of the application of AI to GC×GC-MS data obtained in studies related to food volatilomics [4] nutrimetabolomics [5] and sensomics [6].