Screening of Key Biomarkers and Molecular Mechanisms of Major Depressive Disorder Based on Multi-Cohort Transcriptomic Data

Abstract

Purpose: Major depressive disorder (MDD) is a highly heterogeneous psychiatric disorder with a complex etiology. Its molecular pathological mechanisms are not fully understood. To identify stable biomarkers with potential clinical value and explore their associated molecular pathways, we performed a systematic bioinformatic analysis of three publicly available transcriptomic datasets. Methods: Three expression profile datasets related to major depressive disorder (MDD)—GSE54562, GSE54572, and GSE54568—were retrieved from the Gene Expression Omnibus (GEO) database. After normalization, we performed a differential expression analysis on each dataset and visualized the results using box plots, principal component analysis (PCA), and heat maps. The first dataset underwent further analysis using GO, KEGG, and GSEA enrichment. Then, a protein-protein interaction (PPI) network was constructed using the STRING database. The second and third datasets were then used to validate the candidate features independently. Results: In the first dataset, the major depressive disorder (MDD) and control groups exhibited distinct overall expression patterns, and normalization was effective. Differential analysis identified several markedly dysregulated genes. Among them, SPP1, GLDN, and MEIS1 were upregulated in the MDD group, while MORN4, PCBP3, and SORCS1 were downregulated. Gene ontology (GO) enrichment analysis revealed that the differentially expressed genes were primarily associated with the myelin sheath, compact myelin, and gap junctions. KEGG analysis revealed significant enrichment of the phospholipase D (PLD) signaling pathway and several amino acid metabolism pathways. GSEA also indicated aberrant activation of oxidative phosphorylation, processes related to reactive oxygen species, the phagosome, antigen processing and presentation, ferroptosis, and multiple immune-inflammatory pathways. The PPI network revealed interaction modules centered on myelin structural and membrane transport molecules. In the second dataset (GSE54572), the two groups exhibited consistent but partially overlapping separation. All six top probes differed significantly between the groups. 235779_at showed the most pronounced difference. The third dataset (GSE54568) also displayed a clear case-control separation. All six top probes reached statistical significance, and 232769_at showed the most prominent difference. Conclusion: The onset and progression of MDD may be closely associated with myelin structural damage, membrane lipid metabolic disturbance, neuroinflammation, oxidative stress, and energy metabolic dysregulation. SPP1, GLDN, MEIS1, MORN4, PCBP3, and SORCS1 may be key molecular markers with diagnostic potential. Multi-cohort validation supports the robustness and reproducibility of these candidate molecules, providing a new basis for early MDD diagnosis and molecular subtyping.