Machine Learning Uncovers Shared Immune Signatures Linking Sepsis and Aortic Aneurysm

Sepsis and thoracic aortic aneurysm sit at opposite ends of the clinical spectrum of vascular disease, yet they may be far more intimately connected at the molecular level than their dramatically different presentations suggest. Sepsis is a life-threatening, dysregulated systemic response to infection that can cascade into multi-organ failure within hours, while a thoracic aortic aneurysm is a slow, often silent weakening and bulging of the body’s largest artery that may go undetected for decades before rupturing catastrophically. A new study published in PLOS One by Chengqun She, Hao Deng, Yaoguang Feng, Chaozhong Long, and Juan Wu now argues that these two conditions share a common transcriptional fingerprint rooted in immune dysregulation and metabolic stress, and that a small set of genes could serve as diagnostic markers for both diseases simultaneously.

The research team approached the question with a strategy that has become increasingly powerful in the era of publicly available genomic data: integrated bioinformatics combined with machine learning. Rather than recruiting new patients, the investigators mined transcriptomic datasets deposited in the Gene Expression Omnibus, a public repository maintained by the National Center for Biotechnology Information that archives gene expression profiles from thousands of studies. By pulling together datasets from patients with sepsis and datasets from patients with thoracic aortic aneurysm, the team could compare, on a genome-wide scale, which genes are switched on or off in each condition and, crucially, which changes appear in both.

The first analytical step was the identification of shared differentially expressed genes, abbreviated DEGs. Differentially expressed genes are those whose activity levels, measured as messenger RNA abundance, differ significantly between diseased and healthy tissue. When the researchers cross-referenced the gene lists derived from sepsis samples with those derived from aneurysm samples, they found 193 genes that were perturbed in both diseases. This number is small relative to the roughly 20,000 protein-coding genes in the human genome, which is precisely what makes it interesting: it represents a narrow molecular intersection where two clinically distinct pathologies converge.

To understand what these 193 shared genes actually do, the team performed functional enrichment analysis and gene set enrichment analysis, two complementary computational techniques that ask whether a group of genes is overrepresented in known biological pathways. The results pointed strongly toward the immune system. The shared genes were significantly enriched in pathways governing T cell activation, the process by which cytotoxic and helper T lymphocytes ramp up their defensive machinery, and in Th17 cell differentiation, the maturation of a pro-inflammatory T cell subset that has been implicated in autoimmunity and chronic inflammation. Alongside these immune pathways, the analysis flagged oxidative stress–related processes, the cellular damage that accrues when reactive oxygen species overwhelm antioxidant defenses.

This combination of findings is biologically coherent. Both sepsis and aneurysm formation involve inflammation of the vasculature and damage to the structural integrity of blood vessel walls. In sepsis, a cytokine storm can injure the endothelial lining of vessels throughout the body; in thoracic aortic aneurysm, chronic inflammation and oxidative stress degrade the elastic fibers and smooth muscle that keep the aortic wall strong. The new study suggests that these are not merely superficial parallels but reflect overlapping gene regulatory programs, with T cell–driven immunity and oxidative metabolic stress acting as common threads.

Identifying shared genes is only the beginning, however. A true diagnostic biomarker must do more than appear altered in disease; it must reliably distinguish patients from healthy individuals, and ideally it should do so across independent cohorts measured in different laboratories. To filter the 193 candidates down to the most promising markers, the researchers deployed sequential machine learning algorithms. Machine learning models can weigh the expression patterns of thousands of variables and identify the subset of features that carries the most discriminatory information, a task that is intractable by simple visual inspection of expression matrices.

Through this sequential screening process, five candidate genes emerged: ARG1, IL18R1, LIX1L, TRIM32, and PRKCH. Each has a plausible connection to the biology at play. ARG1 encodes arginase 1, an enzyme that metabolizes the amino acid arginine and participates in inflammatory and urea-cycle metabolism. IL18R1 is the receptor for interleukin-18, a potent pro-inflammatory cytokine involved in innate immunity. LIX1L, TRIM32, and PRKCH are less familiar names but are involved in cellular stress responses, protein regulation, and intracellular signaling respectively. When the team checked the expression patterns of these five genes across independent datasets, the patterns were generally reproducible, an important validation step in any biomarker study, since findings that hold only in a single cohort are frequently artifacts of batch effects or small sample sizes.

The researchers then asked whether the expression of these five genes tracked with the immune landscape of the diseased tissues. Using immune infiltration analysis, a computational method that estimates the proportions of different immune cell types present in a bulk tissue sample based on gene expression signatures, they found that the candidate genes’ expression levels were associated with the estimated abundance of immune cell populations. This supports the idea that the genes are not passive bystanders but are intertwined with the immune cell activity that characterizes both sepsis and aneurysm tissue.

To move from correlation toward mechanism, the team turned to an experimental model. They treated vascular smooth muscle cells, the contractile cells that form the bulk of the aortic wall’s muscular layer, with lipopolysaccharide, or LPS, a molecule derived from the outer membrane of Gram-negative bacteria that is a standard laboratory trigger of inflammatory responses. In these LPS-treated vascular smooth muscle cells, ARG1 and IL18R1 were upregulated, meaning their expression increased, while LIX1L was downregulated. This experiment is significant because vascular smooth muscle cells are central to aneurysm pathology, and showing that bacterial inflammatory stimulation shifts the expression of the candidate genes in these cells provides a direct experimental link between the infection-driven biology of sepsis and the vessel wall remodeling that underlies aneurysm formation.

The authors conclude that sepsis and thoracic aortic aneurysm share common transcriptional signatures involving immune dysregulation and metabolic stress, and that ARG1 and IL18R1 in particular may represent candidate molecular features associated with inflammatory and vascular responses in both diseases. The study is, by design, computational and exploratory: it relies on public datasets, estimated immune cell proportions, and a single cell-culture experiment, so the five-gene signature will need prospective clinical validation before it could inform diagnosis or risk stratification for either condition. Nevertheless, the work adds to a growing body of evidence that diseases long treated as unrelated can share deep molecular roots, and it offers a concrete, testable set of targets for future studies of how systemic inflammation reshapes the vasculature. If those targets hold up, the same molecular markers might one day help clinicians identify which patients with overwhelming infection are also silently accruing damage to their aortas, turning a chance molecular overlap into genuine clinical foresight.

Subject of Research: Shared diagnostic biomarkers and immune-related mechanisms between sepsis and thoracic aortic aneurysm

Article Title: Identification of shared diagnostic biomarkers and immune-related mechanisms between sepsis and thoracic aortic aneurysm based on integrated bioinformatics and machine learning approaches

Article References: She, C., Deng, H., Feng, Y., Long, C., & Wu, J. (2026). Identification of shared diagnostic biomarkers and immune-related mechanisms between sepsis and thoracic aortic aneurysm based on integrated bioinformatics and machine learning approaches. PLOS One, 21(10), e0359784. https://doi.org/10.1371/journal.pone.0359784

Image Credits: AI Generated

DOI: 10.1371/journal.pone.0359784

Keywords: sepsis, thoracic aortic aneurysm, biomarkers, machine learning, bioinformatics, immune infiltration, transcriptomics, ARG1, IL18R1, vascular smooth muscle cells, oxidative stress, T cell activation