Direct Glycan Analysis of Biological Samples and Intact Glycoproteins by Integrating Machine Learning-Driven Surface-Enhanced Raman Scattering and Boronic Acid Arrays.

Frequent monitoring of glycan patterns is a critical step in studying glycan-mediated cellular processes. However, the current glycan analysis tools are resource-intensive and less suitable for routine use in standard laboratories. We developed a novel glycan detection platform by integrating surface-enhanced Raman spectroscopy (SERS), boronic acid (BA) receptors, and machine learning tools. This sensor monitors the molecular fingerprint spectra of BA binding to cis -diol-containing glycans. Different types of BA receptors could yield different stereoselective reactions toward different glycans and exhibit unique vibrational spectra. By integration of the Raman spectra collected from different BA receptors, the structural information can be enriched, eventually improving the accuracy of glycan classification and quantification. Here, we established a SERS-based sensor incorporating multiple different BA receptors. This sensing platform could directly analyze the biological samples, including whole milk and intact glycoproteins (fetuin and asialofetuin), without tedious glycan release and purification steps. The results demonstrate the platform's ability to classify milk oligosaccharides with remarkable classification accuracy, despite the presence of other non-glycan constituents in the background. This sensor could also directly quantify sialylation levels of a fetuin/asialofetuin mixture without glycan release procedures. Moreover, by selecting appropriate BA receptors, the sensor exhibits an excellent performance of differentiating between α2,3 and α2,6 linkages of sialic acids. This low-cost, rapid, and highly accessible sensor will provide the scientific community with an invaluable tool for routine glycan screening in standard laboratories.