Abstract
Dysregulated protein glycosylation is a hallmark of cancer, and systematic investigation of glycosylation patterns is crucial for identifying biomarkers. However, current glycoproteomic studies are constrained by the limited quantification power of single MS files and focus on dysregulated glycopeptides while neglecting the underlying glycoproteins. To address this, we proposed a protein-centric strategy to prioritize proteins susceptible to aberrant glycosylation, aiming to uncover previously overlooked cancer-associated proteins. In this study, we analyzed 200 samples via quantitative glycoproteomics on an integrated platform. Notably, the Glyco-Decipher software's new match-between-run scheme was applied in a large-scale serum-based HCC cohort study to enhance single-shot intact glycopeptide profiling, boosting detection of significantly dysregulated site-specific glycans 4.8-fold compared to conventional method. The protein-centric strategy identified 26 glycoproteins, with Fibronectin emerging as a top diagnostic performer. Specifically, the N1007_H5N4S2 on Fibronectin exhibited excellent diagnostic performance for HCC, achieving an AUC value of 0.917. Furthermore, a machine learning model integrating N1007_H5N4S2 on Fibronectin and N107_H9N3 on Alpha-1-antitrypsin yielded AUC values of 0.950/0.973 (HCC), 0.976/0.922 (TNM-I HCC), and 0.948/0.867 (AFP-negative HCC) in two cohorts, respectively. These findings demonstrated the effectiveness of the protein-centric strategy in identifying robust biomarkers, highlighting the potential of site-specific glycans for improving HCC diagnosis.