Improving the Accuracy and Efficiency of Core Fucose Identification Using Machine Learning

In this study, we employed artificial intelligence to address the challenges in identifying core fucose due to migration effects. By knocking out the FUT8 gene in normal mouse brains, we ensured accurate labeling of non-core fucosylated glycans, enabling the identification of mannose glycans with core fucosylation in wild-type mouse brains. We developed two machine learning models—a semi-supervised mapping convergence (MC) model and a self-supervised autoencoder (AE) model—for core fucose recognition. Experimental results demonstrated that both models performed exceptionally well, with the MC model showing potential in identifying non-core fucosylated glycans and the AE model excelling in core fucose detection.