Concordant and Fluid-Specific Glycosylation Alterations in Serum and Saliva Enable Non-Invasive Diagnosis of Colorectal Cancer

Colorectal cancer (CRC) remains a leading cause of cancer-related mortality, highlighting the need for accurate, noninvasive biomarkers. This study investigated protein glycopatterns in saliva and serum from CRC patients compared with healthy controls, and evaluated their diagnostic potential individually and in combination using machine learning. Paired serum and saliva samples from 44 CRC patients and 45 healthy volunteers were analyzed using lectin microarrays, identifying 15 lectins with distinctive binding profiles in serum and 18 in saliva. Nine lectins (e.g., ACA, PNA, HHL) exhibited consistent patterns across both fluids, indicating systemic CRC-associated glycoprotein remodeling, whereas others (e.g., DBA, PTL-II) displayed opposite trends, reflecting fluid-specific responses. Interfluid correlations were further examined through paired analyses, highlighting both shared and distinct glycosylation characteristics. Random forest models achieved diagnostic accuracies of 92.6% using either serum or saliva alone and 96.3% when combining both, demonstrating the enhanced predictive power of integrated analysis. These findings establish glycosylation profiling of saliva and serum as a promising noninvasive approach for CRC detection and support its potential application in population screening pending validation in larger cohorts.