Objective The present study aims to investigate the alterations of serum proteomic and metabolomic profiles in Chinese patients with severe and active Graves' Orbitopathy (GO). Materials and Methods Thirty patients with GO and 30 healthy volunteers were enrolled. The serum concentrations of FT3, FT4, T3, T4, and thyroid-stimulating hormone (TSH) were analyzed, after which TMT labeling-based proteomics and untargeted metabolomics were performed. MetaboAnalyst and Ingenuity Pathway Analysis (IPA) was used for integrated network analysis. A nomogram was established based on the model to explore the disease prediction ability of the identified feature metabolites. Results One hundred thirteen proteins (19 up-regulated and 94 down-regulated) and 75 metabolites (20 increased and 55 decreased) were significantly altered in GO compared to the control group. By combining the lasso regression, IPA network, and protein-metabolite-disease sub-networks, we extracted feature proteins (CPS1, GP1BA, and COL6A1) and feature metabolites (glycine, glycerol 3-phosphage, and estrone sulfate). The logistic regression analysis revealed that the full model with the prediction factors and three identified feature metabolites had better prediction performance for GO compared to the baseline model. The ROC curve also indicated better prediction performance (AUC = 0.933 vs. 0.789). Conclusion A new biomarker cluster combined with three blood metabolites with high statistical power can be used to discriminate patients with GO. These findings provide further insights into the pathogenesis, diagnosis, and potential therapeutic targets for this disease.Copyright© Bentham Science Publishers; For any queries, please email at epub@benthamscience.net.