Valorisation of premium Italian wines by volatile signature exploration with GC×GC-TOF MS and Computer Vision

Computer vision is a branch of artificial intelligence (AI) that allows systems to extract useful information from digital files to execute actions or make predictions based on that information. Computer vision was applied in the data processing of 54 samples obtained by microvinification of Corvina and Corvinone grape cultivars harvested in five different vineyards, all located in the Veneto region, Italy. Wines’ headspace was sampled with an SPME device after the addition of sodium chloride to enhance the repartition equilibrium at the condensed/gas interface, and further analysed through comprehensive two-dimensional gas chromatography coupled with Time-of-Flight mass spectrometry (GC×GC-ToF MS). The AI approach entails creating a cumulative class-image from the combination of chromatographic data from various samples. It is a data-fusion technique in which the pixels (spectral events) from many 2D images are realigned, registered, and arithmetically summed to create a cumulative pattern that resembles the compositional complexity of all samples within a given group. Composite images were then pairwise-compared to extract crucial biological information from samples, such as cultivar and geographical location of the vineyards. Despite the location of the vineyard not showing meaningful impact on the wine volatilome with unsupervised approaches, the grape cultivar achieves up to 90% correctness in classification accuracy with supervised chemometrics algorithms. Alcohols such as 2-hexen-1-ol, penten-3-ol, and 3-hexen-1-ol were the primarily discriminating analytes (eight) mainly prevalent in Corvina cultivar, followed by esters (seven) and terpenes (five), the last two showing a higher relative distribution in the Corvinone one.