Identification of lipid metabolism-related gene markers and construction of a diagnostic model for multiple sclerosis: An integrated analysis by bioinformatics and machine learning

Abstract

Background

Multiple sclerosis (MS) is an autoimmune inflammatory disorder that causes neurological disability. Dysregulated lipid metabolism contributes to the pathogenesis of MS. This study aimed to identify lipid metabolism-related gene markers and construct a diagnostic model for MS.

Methods

Gene expression profiles for MS were obtained from the Gene Expression Omnibus database. Differentially expressed lipid metabolism-related genes (LMRGs) were identified and performed functional enrichment analysis. Least absolute shrinkage and selection operator (LASSO), random forest (RF), and protein-protein interaction (PPI) analysis were employed to screen hub genes. The predictive power of hub genes was evaluated using receiver operating characteristic (ROC) curves. We developed an artificial neural network (ANN) model and validated its performance in three test sets. Immune cell infiltration analysis, Gene set enrichment analysis, and ceRNA network construction were performed to explore the role of lipid metabolism in the pathogenesis of MS. Drugs prediction and molecular docking were utilized to identify potential therapeutic drugs.

Results

We identified 40 differentially expressed LMRGs, with significant enrichment in Arachidonic acid metabolism, Steroid hormone biosynthesis, Fatty acid elongation, and Sphingolipid metabolism. AKR1C3, NFKB1, and ABCA1 were identified as gene markers for MS, and their expression was upregulated in the MS group. The areas under the ROC curve (AUCs) for AKR1C3, NFKB1, and ABCA1 in the training set were 0.779, 0.703, and 0.726, respectively. The ANN model exhibited good discriminative ability in both the training and test sets, achieving an AUC of 0.826 on the training set and AUC values of 0.822, 0.890, and 0.833 on the test sets. Gamma.delta.T.cell, Natural.killer.T.cell, Plasmacytoid.dendritic.cell, Regulatory.T.cell, and Type.1.T.helper.cell were highly expressed in the MS group. A ceRNA network showed a complex regulatory interplay involving hub genes. Luteolin, isoflavone, and thalidomide had good binding affinities to the hub genes.

Conclusion

Our study emphasized the crucial role of lipid metabolism in MS, identifing AKR1C3, NFKB1, and ABCA1 as gene markers. The ANN model exhibited good performance on both the training and testing sets. These findings offer valuable insights into the molecular mechanisms underlying MS, and establish a scientific foundation for future research.