Carbohydrate fraction characterisation of functional yogurts containing pectin and pectic oligosaccharides through convolutional networks

The carbohydrate fraction of functional yogurts supplemented with citrus and artichoke pectin and their pectic oligosaccharides (POS) has been characterised, including carbohydrates from raw material (milk, pectin and POS) and formed during yogurt manufacture (POS and galactooligosaccharides GOS). Viable cell count and pH values accomplished the quality standards required for yogurts. The content of lactose and lactic and acetic acids was in the range 3.5-4.0; 0.85-1.16 and 0.04-0.07 g 100 g(-1), respectively. Other sugars from pectin (arabinose, galacturonic acid, POS), milk (glucose, galactose, myo-inositol and GOS, GOS) were also quantified. GC-EI-MS spectra of yogurt carbohydrates were classified using machine learning, and structure-relative retention time relationships were calculated determining the abundance of specific fragments on larger oligosaccharide structures. All information generated was correlated using a convolutional network that established characteristic patterns in the complete carbohydrate fraction of each yogurt, as well as changes during fermentation in the carbohydrate profile. It was found that di-, tri- and tetra-POS formed by rhamnose and xylose attached to acetylated-galacturonic acid were released during fermentation in yogurts with artichoke POS. Structures elucidated by these algorithms were confirmed by MALDI-TOF-MS. These models may allow structural differences to be determined among novel oligosaccharides present in food matrices.