A candidate CD4⁺ T-cell gene expression panel for multiple sclerosis highlights MgADP-associated genes

Multiple sclerosis (MS) is an autoimmune disease characterized by immune cell dysregulation. HLA class II alleles drive the disease risk, thereby highlighting CD4⁺ T cells as central contributors to MS initiation. While genome-wide association studies have identified additional non-HLA genetic risk loci, the mechanisms linking these variants to downstream molecular processes remain largely unexplored. Gene expression profiling provides a more comprehensive view of molecular activity, and by combining this with modern machine learning approaches, novel cell type-specific biomarkers may be uncovered. Given that clinically isolated syndrome (CIS) is the earliest clinical manifestation of MS, we constructed a supervised machine learning model using CD4⁺ T cell expression data from people with CIS and healthy individuals to identify disease-specific gene signatures that differentiated people with MS from healthy individuals. These biomarkers were identified in untreated individuals and may thus offer mechanistic insight into the molecular pathways that may drive disease development. In the training set, we used recursive feature elimination to select the three most predictive genes for the model (ELOVL1, GALK2, and PRKCD). These genes showed high discriminative performance across two independent test sets, comprising data from T cells or peripheral immune cells from people with MS and healthy individuals (AU-ROC = 0.84 and 0.83). Notably, two of these genes encode proteins that are associated with Magnesium-ADP (MgADP) in a protein-chemical interaction network. MgADP is a known activator of ATP-sensitive potassium (K_ATP) channels, whose activation has been shown to have neuroprotective and anti-inflammatory effects in an experimental animal model of MS. Adding two additional MgADP-associated genes (IP6K2 and FUCA1) yielded an exploratory five-gene candidate panel with improved or comparable performance across the test sets (AU-ROC = 0.87 and 0.91). Among these genes, GALK2 provides a biochemical link to MgADP production, thus it may have potential downstream effects on ATP-sensitive potassium channel activity. Together, we report a three-gene candidate biomarker panel for MS with an exploratory five-gene extension. Based on these findings, we hypothesize that dysregulation of certain genes that are associated with MgADP may contribute to MS pathogenesis, which requires future functional validation.