Identification of Novel Protein Biomarkers for Myasthenia Gravis by Integrating Human Proteomics with Genetic Instruments

Abstract

Myasthenia gravis (MG) presents significant health and economic challenges. To identify novel biomarkers, we analyzed proteomic data from 52,704 UK Biobank individuals, focusing on 1463 baseline proteins with follow-up >10 years. Baseline and potential MG cases were 1:5 matched to controls by using propensity score matching. We identified 38 consistently up-regulated differentially expressed proteins (DEPs) in both the baseline and potential MG groups compared to controls, categorized into cardiometabolic, inflammation, neurology, and oncology panels. These DEPs showed potential diagnostic value for distinguishing MG from other neuromuscular disorders, with the area under curves ranging from 0.616 to 0.735 across three models (logistic regression, support vector machine, and random forest). To further investigate causality, two-sample Mendelian Randomization (2SMR) and Cox proportional hazard regression were conducted, and we confirmed 18 potential causal proteins associated with MG, including those in the cardiometabolic panel (CEACAM8, OLR1, PGLYRP1, S100A11, and TNC), inflammation panel (CST7, HGF, IL1RN, IL-6, JCHAIN, OSM, PLAUR, and TGFA), neurology panel (CTSS, MMP8, TBC1D17, and VCAN), and oncology panel (S100A12). Currently, no approved drugs for MG specifically target these identified potential causal proteins and ligands. This comprehensive proteomic analysis highlights novel biomarkers associated with MG, suggesting potential targets for identifying risk proteins and future therapeutic interventions.