The role of senescence-related genes in major depressive disorder: insights from machine learning and single cell analysis

Abstract

Background: Evidence indicates that patients with Major Depressive Disorder (MDD) exhibit a senescence phenotype or an increased susceptibility to premature senescence. However, the relationship between senescence-related genes (SRGs) and MDD remains underexplored.

Methods: We analyzed 144 MDD samples and 72 healthy controls from the GEO database to compare SRGs expression. Using Random Forest (RF) and Support Vector Machine-Recursive Feature Elimination (SVM-RFE), we identified five hub SRGs to construct a logistic regression model. Consensus cluster analysis, based on SRGs expression patterns, identified subclusters of MDD patients. Weighted Gene Co-expression Network Analysis (WGCNA) identified gene modules strongly linked to each cluster. Single-cell RNA sequencing was used to analyze MDD SRGs functions.

Results: The five hub SRGs: ALOX15B, TNFSF13, MARCH 15, UBTD1, and MAPK14 showed differential expression between MDD patients and controls. Diagnostics models based on these hub genes demonstrated high accuracy. The hub SRGs correlated positively with neutrophils and negatively with T lymphocytes. SRGs expression pattern revealed two distinct MDD subclusters. WGCNA identified significant gene modules within these subclusters. Additionally, individual endothelial cells with high senescence scores were found to interact with astrocytes via the Notch signaling pathway, suggesting a specific role in MDD pathogenesis.

Conclusion: This comprehensive study elucidates the significant role of SRGs in MDD, highlighting the importance of the Notch signaling pathway in mediating senescence effects.

Keywords: Bulk RNA analysis; Endothelial cells; Major depressive disorder; Senescence; Single-cell RNA analysis.

Fig. 1

Analysis of Senescence-Related Differentially Expressed Genes (SR-DEGs) between MDD and Control Groups (A) GSVA analysis of senescence-related gene scores between MDD and control. B Heatmap of SR-DEGs between MDD and control. C Volcano plot of SR-DEGs between MDD and control. D Heatmap of correlation analysis of SR-DEGs. E Boxplot of SR-DEGs between MDD and control

Fig. 2

Establishment and capacity assessment of diagnostic model by hub SR-DEGs (A-B) The RF and SVM-RFE methods were applied to further screen hub SR-DEGs. C Forest plots for SR-DEGs in diagnostic model. D Receiver operating characteristic (ROC) curve of predicted risk scores in MDD diagnosis. E The AUC score for the full dataset was calculated and then 1000 bootstrap samples of the AUC score. F Comparison of standardized net benefit demonstrating high risk threshold of different hub SR-DEGs. G Nomogram of five hub SR-DEGs in the diagnosis of an MDD sample

Fig. 3

Immune infiltration analysis (A) The heatmap of the correlation of the 33 SRGs with the immune cells. B The heatmap of the correlation of the 5 hub SRGs with the immune cells. C The heatmap of the correlation of the 5 hub SRGs with the inflammatory factor

Fig. 4

Consensus clustering analysis of MDD patients based on SRGs A Consensus matrix plots of consensus clustering for 144 MDD samples when k = 2. B Heat map of the SR-DEGs between the two clusters. C GSEA pathway differential analysis shows activated and inhibited pathways in the two clusters. D Box plot of SR-DEGs in the two clusters. E Differential expression of the inflammatory factors in the two clusters

Fig. 5

Weighted gene co-expression network analysis (WGCNA) (A) An analysis of the scale-free fit index and the mean connectivity for selected soft threshold powers (β). B Clustering dendrogram of genes based on topological overlapping. C Gene clustering dendrogram of WGCNA; (D) Enrichment analysis using Metascape for genes in the turquoise module. E-F Proteomics analysis of the proteins that were differently abundant between two clusters: E is cluster1 and F is cluster2

Fig. 6

Single-cell analysis of SRGs in MDD and detailed analysis of cell-cell communications (A-B) UMAP plot of different cell types in the MDD and control. C The dot plot of UCell scores of SRGs in the different cell types. D Cell communication analysis among high and low senescence endothelial cells with other cells. E CellCall analysis determined pathway differences between high and low senescence endothelial cells with other cells. F The intercellular communications from high senescence endothelial cells to astrocytes. G The Sankey plot of the intercellular communications from high senescence endothelial cells to astrocytes

Fig. 7

Transcription factor profiles correlation analysis with SRGs in high and low senescence endothelial cells (A-B) Transcription factor profiles in high and low senescence endothelial cells. C Correlation analysis between the 7 transcription factors unique to 33 SR-DEGs