Decoding Regioselectivity of Disaccharides Using Quantum Transport and Artificial Intelligence

Disaccharides, composed of two monosaccharide units, are crucial cornerstones in the regulation of various forms of cellular activities. Identifying disaccharides is, however, extremely challenging because of the stereochemistry of monomeric subunits and the regioisomeric diversity of glycosidic linkages. To tackle this fundamental challenge, we devise an automated and unbiased platform that employs quantum tunneling and explainable artificial intelligence (AI) to detect each disaccharide constitutional isomer and regioisomer with good sensitivity and specificity. Explainable AI calling of the three most widely known disaccharides, sucrose, lactose, and maltose, as well as six regioisomers of α-d-glucopyranosyl-d-fructose, is performed simultaneously, and 99.2% accuracy is achieved. From the global and local analysis of the degree of influence of each input variable in calling disaccharide isomers, we aim to provide a better understanding of the AI decision-making process. AI-integrated quantum tunneling technology for sequencing disaccharides, which has never been previously reported, could be valuable in decoding complex structure–function relationships of polysaccharides.