Integrated transcriptomics unravels implications of glycosylation-regulating signature in diagnosis, prognosis and therapeutic benefits of hepatocellular carcinoma

Hepatocellular carcinoma (HCC) patients, featured by markedly heterogeneous tumor microenvironment (TME), meet diverse clinical outcome and neoadjuvant response. Yet the comprehensive influences of aberrant glycosylation on the TME of HCC remain elusive. In this study, by integrated transcriptome profiling, we systemically analyzed the considerable value of glycosylation-regulating signature in diagnosis and prognosis of HCC. A diagnostic model for HCC based on glycosylation-regulating REOs (relative expression orderings) was constructed. A robust glycoscore system was developed to evaluate distinct glycosylation patterns of patients in both the discovery and independent validation cohorts. Mechanisms for prognostic discrepancies between these patterns were dissected in tumor immunoediting, metabolic reprogramming, somatic mutations, and copy number variation (CNV). An individual survival prediction webserver based on a nomogram model (https://survpredict.shinyapps.io/DynNom/) was also established, which facilitates the translational and clinical application of glycoscore. The glycoscore could also effectively predict therapeutic response to sorafenib, Transhepatic Arterial Chemotherapy and Embolization (TACE), and anti-PD-1 therapies in patients with divergent glycosylation patterns, which was validated by a machine learning model. In summary, our study provided a unique insight into the HCC diagnosis and prognostic stratification based on integrated glycosylation-regulating signature. The robust glycosylation scoring system could comprehensively evaluate TME traits, predict prognosis and clinical benefits from neoadjuvant therapies, which may hold promise for promoting personalized clinical decision-making.