# Machine learning combined with omics-based approaches reveals T-lymphocyte cellular fate imbalance in abdominal aortic aneurysm

Researcher: Ge Zhang
Chinese name: 张格
ORCID identity anchor: https://orcid.org/0000-0002-3116-3246
Canonical researcher: https://drgezhang.com/#person

Journal: BMC Biology
Year: 2025
Type: Article
DOI: 10.1186/s12915-025-02400-x
Canonical page: https://drgezhang.com/papers/doi-10-1186-s12915-025-02400-x.html
Markdown record: https://drgezhang.com/papers/doi-10-1186-s12915-025-02400-x.md

## Full Authors

1. Demin Li
2. Ge Zhang
3. Pengchong Du
4. Chang Cao
5. Xuyu He
6. Yan Lv
7. Peiyu Yuan
8. Yujia Wang
9. Ruhao Wu
10. Yifan Cao
11. Yu Yang
12. Jiamin Gao
13. Bo Lan
14. Guo-Ping Shi
15. Xiaolin Cui
16. Jinying Zhang
17. Junnan Tang

## Official Abstract

### Background

Abdominal aortic aneurysm (AAA) is typically an asymptomatic disease closely associated with immune mechanisms. A deep understanding of cellular responses within AAA tissues, particularly the molecular changes in T-cell populations, is critical for disease diagnosis and treatment. However, the specific mechanisms inducing T-lymphocyte fate imbalance in AAA remain to be elucidated.

### Results

The analysis revealed the core mechanisms driving T-lymphocyte fate imbalance in AAA. We successfully established a comprehensive regulatory map encompassing T-cell infiltration regulatory features, critical transcription factors, and dysregulated immune signaling pathways. Machine learning algorithms identified transcription factors FOSB and JUNB as key biomarkers. Validation across multiple independent datasets and clinical samples confirmed the feasibility and accuracy of FOSB and JUNB as clinical diagnostic biomarkers for AAA.

### Conclusions

Through the analysis of single-cell and bulk data, hallmarks of human AAA cellular landscape and T-cell comprehensive developmental relationships were recapitulated. This study identified important roles of T-cell and the molecular mechanisms for the dynamic T-cell infiltrating process, which could characterize disease status and landscape of human AAA microenvironment. Using the deep learning algorithms, FOSB and JUNB were demonstrated as pivotal biomarkers of AAA, together with screening the potential pharmacologic agents targeting T-cell polarization. Taken together, this expands the current understanding of AAA pathogenesis and may provide a feasible immune-targeted therapeutic strategy.

Text reproduced verbatim from PMC ([source](https://pmc.ncbi.nlm.nih.gov/articles/PMC12465141/)); [DOI](https://doi.org/10.1186/s12915-025-02400-x); [CC-BY-NC-ND-4.0](https://creativecommons.org/licenses/by-nc-nd/4.0/).

<a id="paper-discovery"></a>
## Research & review context

### Research fields

- Aortic aneurysm immunobiology
- Single-cell vascular research

### Review topics

- T-cell heterogeneity in abdominal aortic aneurysm
- Immune-cell-linked molecular markers in human vascular disease

### Scientific role

Human AAA T-cell state description integrated with bulk-expression marker analysis and independent human-sample assessment.

### Evidence position

Human single-cell and bulk transcriptomics with small clinical-sample evaluation and supportive mouse expression; inferred states are not longitudinal fate tracing or proven therapy.

### Related concepts

- T-cell state imbalance
- Cell-state-to-marker integration

### Possible citation contexts

- Background: adaptive immune heterogeneity in abdominal aortic aneurysm.
- Methods: integration of human single-cell immune states with bulk-expression markers, without causal or therapeutic inference from predicted agents.

Possible literature-review contexts, not evidence of existing citations. Read the findings and Evidence Scope before citing the original article.

## Evidence Scale

**Machine learning combined with omics-based approaches reveals T-lymphocyte cellular fate imbalance in abdominal aortic aneurysm**

Integrated human single-cell and bulk transcriptomic analyses describe AAA-associated T-cell state heterogeneity, derive TIRS and evaluate candidate biomarkers in public datasets and small independent human samples, with supportive mouse expression evidence.

| Evidence | Value |
| --- | --- |
| Human AAA scRNA donors | 4 aneurysmal abdominal aorta samples |
| Post-QC single cells | 7,343 cells |
| T cells | 3,841 cells |
| T-cell phenotypes | 8 |
| Bulk Dataset 1 | AAA n=80; controls n=10 |
| Bulk Dataset 2 | AAA n=9; controls n=10 |
| Bulk Dataset 3 | dilated PVAT n=30; non-dilated PVAT n=30 |
| In-house aortic wall | AAA n=5; controls n=4 |
| In-house peripheral blood | AAA n=24; controls n=15 |

### Counting note

Datasets contain different biological and analytical units. GSE57691 + GSE98278 contribute to Dataset 1. Cell counts must never be added to participant or sample counts. Public cohort samples must not be summed into a synthetic unique participant total without participant-level deduplication. Mouse measurements are separate preclinical units.

## Research Question

Can integrated human single-cell and bulk transcriptomic analyses reconstruct T-cell state imbalance in abdominal aortic aneurysm and identify reproducible T-cell-linked molecular markers across public datasets and independent clinical samples?

## Author Evidence Summary

The study combined human AAA single-cell profiles, public bulk transcriptomic cohorts and independent clinical samples to describe T-cell heterogeneity and infer transcriptional state relationships. Four aneurysmal aortic samples yielded 7,343 post-QC cells, including 3,841 T cells assigned to eight phenotypes. Monocle inferred a branching trajectory; it did not directly follow individual cells over time. Integrative gene-set analysis derived a 50-gene TIRS, and LASSO plus SVM-RFE prioritized FOSB, JUNB, CST7 and TBC1D4. Marker discrimination varied across public cohorts and was evaluated in small independent aortic-wall samples (5 AAA, 4 controls) and peripheral blood (24 AAA, 15 controls). BAPN + AngII mouse experiments provided supportive FOSB/JUNB expression evidence. These findings support transcriptomic associations and candidate-marker evaluation, while prospective diagnostic utility, human causal mechanisms and treatment benefit remain unestablished.

## Key Findings

<a id="kf1"></a>
### KF1: Human AAA single-cell profiling characterized a heterogeneous cellular landscape.

Context: Four human aneurysmal aortic samples; cells are analytical units and are not independent participants.

| Evidence | Value |
| --- | --- |
| aneurysmal aortic samples | 4 |
| post-QC cells | 7,343 |
| populations | 16 |
| main cell types | 9 |

Source locator: Results: Single-cell transcriptome profiles of human AAA; Fig. 1

<a id="kf2"></a>
### KF2: Monocle inferred a bifurcated T-cell trajectory linking heterogeneous transcriptional states in human AAA.

Context: Pseudotime reconstructs state relationships; the study did not directly observe individual T cells changing fate over time.

| Evidence | Value |
| --- | --- |
| T cells | 3,841 |
| T-cell phenotypes | 8 |
| infiltration states | 3 |
| trajectory | Bifurcated trajectory with 2 terminal fates |

Source locator: Results: Heterogeneity of T lymphocytes in human AAA and high-dimensional single-cell lineage mapping; Figs. 2-3; Supplementary Fig. S3

<a id="kf3"></a>
### KF3: Public bulk cohorts showed AAA-associated immune and transcriptomic patterns.

Context: Disease-associated patterns across different tissue and cohort comparisons; these are associations, not causal evidence.

| Evidence | Value |
| --- | --- |
| Dataset 1 | AAA n=80; controls n=10 |
| Dataset 2 | AAA n=9; controls n=10 |
| Dataset 3 | dilated PVAT n=30; non-dilated PVAT n=30 |

Source locator: Results: bulk immune-landscape analysis; Fig. 4; Supplementary Fig. S4

<a id="kf4"></a>
### KF4: Integrated cell-state, fate-associated and trajectory gene sets yielded a 50-gene bulk-overlapping T-cell infiltration regulatory signature, TIRS.

Context: Transcriptomic signature derivation; genes selected by these analyses are not established causal regulators.

| Evidence | Value |
| --- | --- |
| cell-state DE genes | 968 |
| BEAM fate leader genes | 437 |
| trajectory ordering genes | 375 |
| intersection | 374 |
| final bulk-overlapping TIRS | 50 genes |

Source locator: Results: Biological implications underlying TIRS; Fig. 5; Supplementary Fig. S5

<a id="kf5"></a>
### KF5: LASSO and SVM-RFE prioritized FOSB, JUNB, CST7 and TBC1D4, with variable discrimination across public cohorts.

Context: Public-cohort retrospective marker discrimination; dataset performance does not establish prospective diagnostic utility.

| Evidence | Value |
| --- | --- |
| LASSO + SVM-RFE intersection | FOSB, JUNB, CST7, TBC1D4 |
| Dataset 1 AUCs (FOSB/JUNB/CST7/TBC1D4) | 0.911 / 0.917 / 0.926 / 0.955 |
| Dataset 2 AUCs (FOSB/JUNB/CST7/TBC1D4) | 0.982 / 0.911 / 0.893 / 0.857 |
| Dataset 3 AUCs (FOSB/JUNB/CST7/TBC1D4) | 0.956 / 0.887 / 0.506 / 0.514 |

Source locator: Results: Detection and evaluation of key biomarkers; Fig. 6

<a id="kf6"></a>
### KF6: The candidate markers were evaluated in independent in-house human aortic-wall and peripheral-blood samples.

Context: The aortic-wall comparison was small (5 AAA and 4 controls); these results do not establish approved standalone clinical tests.

| Evidence | Value |
| --- | --- |
| aortic wall | AAA n=5; controls n=4 |
| aortic-wall AUCs (FOSB/JUNB/CST7/TBC1D4) | 1.000 / 1.000 / 0.600 / 0.800 |
| peripheral blood | AAA n=24; controls n=15 |
| blood AUCs (FOSB/JUNB/CST7/TBC1D4) | 0.989 / 0.906 / 0.933 / 0.736 |

Source locator: Results: Detection and evaluation of key biomarkers; Fig. 7A-G

<a id="kf7"></a>
### KF7: The BAPN + AngII mouse AAA model showed higher FOSB/JUNB signal in aortic immunohistochemistry and serum ELISA under the reported conditions.

Context: Supportive preclinical expression evidence; it does not prove a human causal mechanism or treatment efficacy.

| Evidence | Value |
| --- | --- |
| mouse AAA model | BAPN + AngII |
| expression assays | Higher FOSB/JUNB signal in aortic IHC and serum ELISA |

Source locator: Fig. 7H-L; Supplementary Fig. S6

## Study Design & Analytical Framework

### Study profile

- Study Design: Integrative multicohort transcriptomic study combining human AAA single-cell RNA sequencing, multiple public bulk transcriptomic cohorts, independent human clinical validation samples and supportive mouse AAA expression experiments
- Evidence Type: Human single-cell and multicohort transcriptomic association with independent small-sample clinical validation and supportive preclinical expression evidence
- Population: Human abdominal aortic aneurysm tissue, peripheral blood and public transcriptomic datasets, with supportive experimental mouse AAA evidence
- Primary Endpoint: T-cell state/fate-associated transcriptional architecture in AAA
- Secondary Endpoint: Identification and cross-dataset evaluation of T-cell infiltration regulatory genes and candidate AAA biomarkers
- External dataset evaluation: Yes
- Data modalities: single-cell RNA sequencing, bulk transcriptomics, qRT-PCR, immunohistochemistry, ELISA, mouse AAA histopathology

## What This Study Adds

- A human AAA single-cell description of T-cell heterogeneity and inferred state relationships.
- A 50-gene TIRS derived from complementary transcriptomic analyses.
- Cross-dataset candidate-marker evaluation with small independent human-sample assessment.
- Supportive mouse FOSB/JUNB expression observations.

## Evidence Scope

### Supports

- Human AAA single-cell T-cell heterogeneity.
- Inferred T-cell transcriptional state and fate structure.
- Multicohort AAA-associated transcriptomic patterns and TIRS derivation.
- FOSB/JUNB/CST7/TBC1D4 marker discrimination in the analyzed datasets.
- Small independent in-house human validation, including aortic wall n=5 AAA and n=4 controls.
- Supportive mouse expression evidence.

### Does Not Establish

- Pseudotime as directly observed longitudinal cell fate.
- 7,343 cells as 7,343 participants.
- FOSB/JUNB as established causal drivers of human AAA.
- Prospective diagnostic performance or approved standalone clinical tests.
- Predicted pharmacologic agents as effective AAA therapy.
- Mouse expression evidence as human treatment evidence.

### Limitations

- The in-house aortic-wall sample size was small: 5 AAA and 4 controls.
- Pseudotime is an inferred transcriptional ordering, not direct longitudinal lineage observation.
- Candidate-marker discrimination does not establish prospective clinical diagnostic performance.
- Mouse expression observations do not establish human causal mechanisms or therapeutic benefit.

<a id="citation-use-cases"></a>
## When This Study Is Useful to Cite

### CU1: What evidence describes human AAA T-cell heterogeneity at single-cell resolution?

Human AAA single-cell profiling characterized a heterogeneous cellular landscape. Four human aneurysmal aortic samples; cells are analytical units and are not independent participants. Monocle inferred a bifurcated T-cell trajectory linking heterogeneous transcriptional states in human AAA. Pseudotime reconstructs state relationships; the study did not directly observe individual T cells changing fate over time.

Evidence: [KF1](#kf1), [KF2](#kf2)

### CU2: How were T-cell infiltration states and pseudotime reconstructed in AAA?

Monocle inferred a bifurcated T-cell trajectory linking heterogeneous transcriptional states in human AAA. Pseudotime reconstructs state relationships; the study did not directly observe individual T cells changing fate over time.

Evidence: [KF2](#kf2)

### CU3: How was the T-cell infiltration regulatory signature TIRS derived?

Integrated cell-state, fate-associated and trajectory gene sets yielded a 50-gene bulk-overlapping T-cell infiltration regulatory signature, TIRS. Transcriptomic signature derivation; genes selected by these analyses are not established causal regulators.

Evidence: [KF4](#kf4)

### CU4: What multicohort transcriptomic evidence describes AAA-associated immune signals?

Public bulk cohorts showed AAA-associated immune and transcriptomic patterns. Disease-associated patterns across different tissue and cohort comparisons; these are associations, not causal evidence. Integrated cell-state, fate-associated and trajectory gene sets yielded a 50-gene bulk-overlapping T-cell infiltration regulatory signature, TIRS. Transcriptomic signature derivation; genes selected by these analyses are not established causal regulators.

Evidence: [KF3](#kf3), [KF4](#kf4)

### CU5: How were FOSB/JUNB/CST7/TBC1D4 prioritized using machine learning?

LASSO and SVM-RFE prioritized FOSB, JUNB, CST7 and TBC1D4, with variable discrimination across public cohorts. Public-cohort retrospective marker discrimination; dataset performance does not establish prospective diagnostic utility.

Evidence: [KF5](#kf5)

### CU6: What independent human aortic-wall and blood evidence evaluated candidate markers?

The candidate markers were evaluated in independent in-house human aortic-wall and peripheral-blood samples. The aortic-wall comparison was small (5 AAA and 4 controls); these results do not establish approved standalone clinical tests.

Evidence: [KF6](#kf6)

### CU7: What supportive preclinical expression evidence was reported for FOSB/JUNB?

The BAPN + AngII mouse AAA model showed higher FOSB/JUNB signal in aortic immunohistochemistry and serum ELISA under the reported conditions. Supportive preclinical expression evidence; it does not prove a human causal mechanism or treatment efficacy.

Evidence: [KF7](#kf7)

<a id="citation-boundaries"></a>
## What This Study Should Not Be Cited to Claim

- Pseudotime as directly observed longitudinal cell fate.
- 7,343 cells as 7,343 participants.
- FOSB/JUNB as established causal drivers of human AAA.
- Prospective diagnostic performance or approved standalone clinical tests.
- Predicted pharmacologic agents as effective AAA therapy.
- Mouse expression evidence as human treatment evidence.

<a id="evidence-matrix"></a>
## Evidence Matrix

<a id="evidence-aaa-tcell-1"></a>
### KF1

- Context: Four human aneurysmal aortic samples; cells are analytical units and are not independent participants.
- Finding: Human AAA single-cell profiling characterized a heterogeneous cellular landscape. aneurysmal aortic samples: 4; post-QC cells: 7,343; populations: 16; main cell types: 9
- Evidence Level: Human single-cell observation
- Scope: Four human aneurysmal aortic samples; cells are analytical units and are not independent participants.
- Source Locator: Results: Single-cell transcriptome profiles of human AAA; Fig. 1
- Evidence refs: [KF1](#kf1)

<a id="evidence-aaa-tcell-2"></a>
### KF2

- Context: Pseudotime reconstructs state relationships; the study did not directly observe individual T cells changing fate over time.
- Finding: Monocle inferred a bifurcated T-cell trajectory linking heterogeneous transcriptional states in human AAA. T cells: 3,841; T-cell phenotypes: 8; infiltration states: 3; trajectory: Bifurcated trajectory with 2 terminal fates
- Evidence Level: Computational pseudotime inference
- Scope: Pseudotime reconstructs state relationships; the study did not directly observe individual T cells changing fate over time.
- Source Locator: Results: Heterogeneity of T lymphocytes in human AAA and high-dimensional single-cell lineage mapping; Figs. 2-3; Supplementary Fig. S3
- Evidence refs: [KF2](#kf2)

<a id="evidence-aaa-tcell-3"></a>
### KF3

- Context: Disease-associated patterns across different tissue and cohort comparisons; these are associations, not causal evidence.
- Finding: Public bulk cohorts showed AAA-associated immune and transcriptomic patterns. Dataset 1: AAA n=80; controls n=10; Dataset 2: AAA n=9; controls n=10; Dataset 3: dilated PVAT n=30; non-dilated PVAT n=30
- Evidence Level: Human bulk transcriptomic association
- Scope: Disease-associated patterns across different tissue and cohort comparisons; these are associations, not causal evidence.
- Source Locator: Results: bulk immune-landscape analysis; Fig. 4; Supplementary Fig. S4
- Evidence refs: [KF3](#kf3)

<a id="evidence-aaa-tcell-4"></a>
### KF4

- Context: Transcriptomic signature derivation; genes selected by these analyses are not established causal regulators.
- Finding: Integrated cell-state, fate-associated and trajectory gene sets yielded a 50-gene bulk-overlapping T-cell infiltration regulatory signature, TIRS. cell-state DE genes: 968; BEAM fate leader genes: 437; trajectory ordering genes: 375; intersection: 374; final bulk-overlapping TIRS: 50 genes
- Evidence Level: Transcriptomic signature derivation
- Scope: Transcriptomic signature derivation; genes selected by these analyses are not established causal regulators.
- Source Locator: Results: Biological implications underlying TIRS; Fig. 5; Supplementary Fig. S5
- Evidence refs: [KF4](#kf4)

<a id="evidence-aaa-tcell-5"></a>
### KF5

- Context: Public-cohort retrospective marker discrimination; dataset performance does not establish prospective diagnostic utility.
- Finding: LASSO and SVM-RFE prioritized FOSB, JUNB, CST7 and TBC1D4, with variable discrimination across public cohorts. LASSO + SVM-RFE intersection: FOSB, JUNB, CST7, TBC1D4; Dataset 1 AUCs (FOSB/JUNB/CST7/TBC1D4): 0.911 / 0.917 / 0.926 / 0.955; Dataset 2 AUCs (FOSB/JUNB/CST7/TBC1D4): 0.982 / 0.911 / 0.893 / 0.857; Dataset 3 AUCs (FOSB/JUNB/CST7/TBC1D4): 0.956 / 0.887 / 0.506 / 0.514
- Evidence Level: Retrospective marker discrimination
- Scope: Public-cohort retrospective marker discrimination; dataset performance does not establish prospective diagnostic utility.
- Source Locator: Results: Detection and evaluation of key biomarkers; Fig. 6
- Evidence refs: [KF5](#kf5)

<a id="evidence-aaa-tcell-6"></a>
### KF6

- Context: The aortic-wall comparison was small (5 AAA and 4 controls); these results do not establish approved standalone clinical tests.
- Finding: The candidate markers were evaluated in independent in-house human aortic-wall and peripheral-blood samples. aortic wall: AAA n=5; controls n=4; aortic-wall AUCs (FOSB/JUNB/CST7/TBC1D4): 1.000 / 1.000 / 0.600 / 0.800; peripheral blood: AAA n=24; controls n=15; blood AUCs (FOSB/JUNB/CST7/TBC1D4): 0.989 / 0.906 / 0.933 / 0.736
- Evidence Level: Small independent human-sample evaluation
- Scope: The aortic-wall comparison was small (5 AAA and 4 controls); these results do not establish approved standalone clinical tests.
- Source Locator: Results: Detection and evaluation of key biomarkers; Fig. 7A-G
- Evidence refs: [KF6](#kf6)

<a id="evidence-aaa-tcell-7"></a>
### KF7

- Context: Supportive preclinical expression evidence; it does not prove a human causal mechanism or treatment efficacy.
- Finding: The BAPN + AngII mouse AAA model showed higher FOSB/JUNB signal in aortic immunohistochemistry and serum ELISA under the reported conditions. mouse AAA model: BAPN + AngII; expression assays: Higher FOSB/JUNB signal in aortic IHC and serum ELISA
- Evidence Level: Supportive mouse expression evidence
- Scope: Supportive preclinical expression evidence; it does not prove a human causal mechanism or treatment efficacy.
- Source Locator: Fig. 7H-L; Supplementary Fig. S6
- Evidence refs: [KF7](#kf7)

## Q&A

### What single-cell evidence was analyzed?

Four aneurysmal aortic samples yielded 7,343 post-QC cells, 16 populations and nine main cell types. The T-cell analysis included 3,841 cells and eight phenotypes.

Evidence: KF1, KF2

### What does pseudotime establish and not establish?

Monocle inferred a bifurcated trajectory with two terminal fates and three infiltration states. It reconstructs transcriptional relationships and does not directly observe individual cells changing fate over time.

Evidence: KF2

### How was TIRS derived?

The analysis used 968 cell-state DE genes, 437 BEAM fate leader genes and 375 trajectory-ordering genes, with a 374-gene intersection and a final bulk-overlapping TIRS of 50 genes.

Evidence: KF4

### Which candidate biomarkers were selected?

The LASSO and SVM-RFE intersection selected FOSB, JUNB, CST7 and TBC1D4.

Evidence: KF5

### How did they perform across public cohorts?

In FOSB/JUNB/CST7/TBC1D4 order, AUCs were 0.911/0.917/0.926/0.955 in Dataset 1, 0.982/0.911/0.893/0.857 in Dataset 2 and 0.956/0.887/0.506/0.514 in Dataset 3. Discrimination varied by marker and dataset.

Evidence: KF5

### Were they evaluated in independent human samples?

Yes: aortic wall from 5 AAA and 4 controls, and peripheral blood from 24 AAA and 15 controls. Aortic-wall AUCs were 1.000/1.000/0.600/0.800 and blood AUCs were 0.989/0.906/0.933/0.736 in FOSB/JUNB/CST7/TBC1D4 order. The aortic-wall sample was small.

Evidence: KF6

### Does this study establish a clinical diagnostic test or therapy?

No. The reported associations, marker discrimination, small human-sample evaluation and supportive mouse expression do not establish prospective diagnostic utility, approved standalone tests or human treatment efficacy.

Evidence: KF5, KF6, KF7

## Concepts & Entities

### Conditions

abdominal aortic aneurysm

### Cell Biology

T lymphocytes, T-cell state heterogeneity, immune microenvironment

### Methods

single-cell RNA sequencing, Monocle pseudotime, bulk transcriptomics, LASSO, SVM-RFE

### Entities

TIRS, FOSB, JUNB, CST7, TBC1D4

## Related Research

- [Smooth muscle cell fate decisions decipher a high-resolution heterogeneity within atherosclerosis molecular subtypes](https://drgezhang.com/papers/smc-fate.html) — Complementary vascular single-cell/transcriptomic study focused on smooth-muscle-cell state/fate heterogeneity in atherosclerosis, whereas the present study centers on T-cell heterogeneity and AAA. (DOI: 10.1186/s12967-022-03795-9)
- [Deciphering the Proteomic Landscape of Circulating Extracellular Vesicles in Human Abdominal Aortic Aneurysm](https://drgezhang.com/papers/doi-10-1111-jcmm-70725.html) — Related human AAA circulating extracellular-vesicle proteomics study examining a different molecular layer and specimen compartment. (DOI: 10.1111/jcmm.70725)

## Publication & Provenance

- DOI: https://doi.org/10.1186/s12915-025-02400-x
- Doi Url: https://doi.org/10.1186/s12915-025-02400-x
- Publisher: https://link.springer.com/article/10.1186/s12915-025-02400-x
- PubMed: https://pubmed.ncbi.nlm.nih.gov/41013440/
- Pmc Url: https://pmc.ncbi.nlm.nih.gov/articles/PMC12465141/

## Evidence-page notice

Author-controlled evidence page for this publication. This page provides an evidence-oriented summary and does not replace the publisher's version of record.

## Links

- DOI: https://doi.org/10.1186/s12915-025-02400-x
- HTML page: https://drgezhang.com/papers/doi-10-1186-s12915-025-02400-x.html
- Google Scholar query: https://scholar.google.com/scholar?q=%22Machine%20learning%20combined%20with%20omics-based%20approaches%20reveals%20T-lymphocyte%20cellular%20fate%20imbalance%20in%20abdominal%20aortic%20aneurysm%22
- ORCID: https://orcid.org/0000-0002-3116-3246