Shared Immune Genes May Link Depression and Stroke, Study Finds

Major depressive disorder and ischemic stroke are two of the most devastating conditions in modern medicine, ranking among the leading causes of mortality and disability worldwide. Clinicians have long observed that the two travel together: depression is more common in stroke survivors than in the general population, and a history of depression appears to raise the risk of a subsequent stroke. Yet the biological thread connecting a psychiatric disorder of mood with a vascular catastrophe in the brain has remained frustratingly elusive. A new study published in BMC Psychiatry by Xinyun Li, Jingting Zhao, Xiangming Ye, and Zhenfei Xiong, based at Hangzhou Medical College and affiliated hospitals in Zhejiang, China, argues that the missing link may lie in the immune system, and that a small set of shared genes could serve as diagnostic markers for both diseases at once.

The research team approached the problem from two directions. First, they conducted a retrospective cohort study examining the medical histories of patients admitted with ischemic stroke, documenting how many of them carried a prior diagnosis of depression. This clinical grounding matters, because many bioinformatics studies of comorbidity never leave the computer. By confirming in patient records that depression and stroke genuinely cluster together in the same individuals, the researchers established that any shared molecular signature they found would be clinically meaningful rather than a statistical curiosity.

Second, the team turned to public transcriptomic data, mining gene expression datasets deposited in the Gene Expression Omnibus, the National Center for Biotechnology Information’s archive of genomics data. They compared brain or blood expression profiles from patients with ischemic stroke and patients with major depressive disorder against healthy controls, using differential expression analysis to flag genes whose activity rises or falls in each disease. To move beyond single-gene statistics, they applied weighted gene co-expression network analysis, or WGCNA, a technique that groups thousands of genes into modules based on how their expression patterns covary across samples. Modules that behave abnormally in both diseases point toward biological pathways that the two conditions may share, even when no individual gene reaches significance in both datasets on its own.

The intersection of these analyses, refined through protein-protein interaction networks and machine learning feature selection, produced a remarkably compact result: five co-susceptible genes shared between ischemic stroke and major depressive disorder. The genes are CD163, FOLR3, GADD45A, ECHDC2, and RPLP0. Each tells a plausible story. CD163 encodes a receptor on macrophages, the scavenger cells of the immune system, and is a well-known marker of inflammation and of the activation of monocytes that patrol blood vessels. FOLR3, folate receptor gamma-3, participates in folate transport and has been implicated in inflammatory responses. GADD45A, growth arrest and DNA damage inducible alpha, is a stress-response gene that cells switch on when they face damage or inflammatory signaling. ECHDC2 is involved in lipid metabolism, a process intimately tied to atherosclerosis, the underlying cause of most ischemic strokes. RPLP0 is a structural component of the ribosome, the cell’s protein-building machinery, and its dysregulation can reflect broad shifts in cellular activity under stress.

Machine learning played a central role in distilling the candidate list. The researchers used methods including least absolute shrinkage and selection operator regression, known as LASSO, and support vector machine recursive feature elimination, or SVM-RFE, to identify which genes carried the most diagnostic information. They then built a diagnostic nomogram, a graphical scoring tool that combines the expression values of the five genes into a single risk estimate for each disease. Performance was evaluated with receiver operating characteristic curves, quantified by the area under the curve, along with calibration plots, decision curve analysis, and clinical impact curves. In the ischemic stroke validation cohort, the nomogram achieved an area under the curve of 0.96 with an accuracy of 0.946, while in the independent depression validation dataset GSE251778 it reached an area under the curve of 0.803 with an accuracy of 0.719. Those figures suggest a five-gene signature strong enough to distinguish patients from controls in both conditions, with particularly striking performance on the stroke side.

Because the study’s central hypothesis concerns immunity, the team also compared immune cell infiltration between stroke patients, depression patients, and healthy controls. Using single-sample gene set enrichment analysis, they estimated the relative abundance of different immune cell types in each sample and found characteristic patterns of immune alteration in both diseases. Correlation analyses then linked the five co-susceptible genes to key immune cell populations, and crucially, these relationships held up when the analysis was repeated in independent validation datasets. The consistency across discovery and validation cohorts strengthens the argument that the gene-immune associations are not artifacts of a single dataset but reflect a reproducible biological signal connecting neuroinflammation to both disorders.

The researchers did not stop at computational validation. They confirmed the expression levels of the five co-susceptible genes in animal models of disease, using quantitative real-time polymerase chain reaction to measure gene expression in tissue. The experimental work, conducted with support from Zhejiang Provincial research programs and approved by the ethics committee of Zhejiang Provincial People’s Hospital, included behavioral assessments such as the tail suspension test and open field test, and a middle cerebral artery occlusion model, the standard experimental approach for reproducing ischemic stroke in rodents. The team also manipulated gene activity using adeno-associated virus vectors to knock down or overexpress specific targets, and performed downstream functional analyses to probe what the genes do. In a forward-looking step, they screened targeted small-molecule compounds using the Drug-Gene Interaction Database, generating a shortlist of candidate drugs that act on the shared genetic pathways.

The significance of the work lies in its claim to be the first study to combine bioinformatics and machine learning to screen for co-susceptible genes between ischemic stroke and major depressive disorder and to construct a diagnostic model that establishes both genetic and immune links between the two diseases. If the five-gene signature can be replicated in larger, more diverse cohorts and translated into a clinically practical assay, it could eventually help physicians identify patients at risk of either condition, or of the post-stroke depression that afflicts a substantial fraction of stroke survivors. A blood-based panel measuring CD163, FOLR3, GADD45A, ECHDC2, and RPLP0 is not yet on the horizon as an approved diagnostic, but the nomogram framework demonstrates that the information content of these genes is sufficient to separate patients from controls with high accuracy in the datasets tested.

There are, as with any study of this kind, important caveats. The gene expression datasets analyzed were drawn from archived cohorts with their own sampling choices, tissue sources, and technical platforms, and the retrospective cohort design can document association between prior depression and stroke but cannot fully disentangle causation. The mouse validation, while valuable, cannot capture the full complexity of human psychiatric illness. The authors themselves note that the article was shared early as a peer-reviewed, accepted version subject to further edits. Independent replication, ideally with prospective patient samples and standardized measurement protocols, will be essential before the biomarker panel moves toward clinical use. The open-access study, published under a Creative Commons license with its data and supplementary materials freely available, invites exactly that scrutiny.

Even so, the study adds momentum to one of the most compelling narratives in modern neuroscience: that the boundary between psychiatry and vascular medicine is far more porous than traditional specialty divisions suggest. Inflammation, immune cell trafficking, and stress-response genes appear to sit at the crossroads where a mood disorder and a blood vessel disease meet. By naming five specific genes and showing that they carry diagnostic signal for both conditions, the Chinese team has converted a broad hypothesis about the depression-stroke link into a concrete, testable molecular target list. Whether those targets become the basis of new diagnostics, drug screens, or prevention strategies, they offer researchers a shared map of territory that psychiatry and neurology have long explored separately.

Subject of Research: Shared immune-related genetic biomarkers linking major depressive disorder and ischemic stroke

Article Title: Identification of potential immune-related diagnostic biomarkers for major depressive disorder and ischemic stroke using bioinformatics analysis

Article References: Li, X., Zhao, J., Ye, X., & Xiong, Z. (2026). Identification of potential immune-related diagnostic biomarkers for major depressive disorder and ischemic stroke using bioinformatics analysis. BMC Psychiatry. https://doi.org/10.1186/s12888-026-08682-x

Image Credits: AI Generated

DOI: 10.1186/s12888-026-08682-x

Keywords: major depressive disorder, ischemic stroke, biomarkers, bioinformatics, machine learning, immune infiltration, co-susceptible genes, WGCNA, diagnostic nomogram, neuroinflammation, gene expression, transcriptomics