Multi-omics data integration from patients with carotid stenosis illuminates key molecular signatures of atherosclerotic instability
Genome Medicine, 2026
Das V., Narayanan S., Zhang X., Bergman O., Djordjevic D., Kronqvist M., Chemaly M., Karadimou G., Sundman S., Prasad I., Buckler A., Knape K., Michaelsen N., Hedin U., Matic L.
| Disease area | Application area | Sample type | Products |
|---|---|---|---|
CVD | Patient Stratification Will Be: Pathophysiology | Plasma |  Olink Target 96 |
Abstract
Background
Understanding the pathophysiology of unstable atherosclerosis is imperative to prevent myocardial infarction and stroke. Here, we used multi-omics integration to identify key molecular targets with diagnostic and therapeutic potential.
Methods
Biobank of Karolinska Endarterectomies encompassing patients with symptomatic (S) and asymptomatic (AS) carotid atherosclerosis was the main resource. Plaques, peripheral blood monocytes and plasma sampled locally from around plaque or periphery of n > 700 individuals, were profiled by transcriptomics, proteomics and metabolomics. A supervised machine learning method DIABLO was used for patient data integration. Multi-omics layers were integrated separately across local and peripheral disease sites, and their intersection, with stratification for symptomatology. Identified analytes were investigated using scRNAseq, clinical and outcome data.
Results
In peripheral circulation, FABP4, IL6, Bilirubin and Sphingomyelin were the most prominent analytes. F11, ANGPTL3, ICOSLG, ITGB1 and Sphingomyelin were enriched in the local disease site, while FABP4, C1R, IL6, Bilirubin and Sphingomyelin appeared at the intersection. Coagulation, necroptosis, inflammation and cholesterol metabolism were confirmed as key pathways determining symptomatology. Clinical analyses showed an impact of lipid-lowering therapy on ICOSLG expression, anti-hypertensives on plasma FABP4 and BLVRB levels, anti-diabetics on plasma Sphingomyelins, while no medications affected ANGPTL3. Association with future adverse events was shown for plasma Bilirubin, Sphingomyelin, ANGPTL3 and ICOSLG plaque levels. Open-source target analyses suggested genetic involvement of F11, C1S, EGFR, IL6, ANGPTL3 in the disease.
Conclusions
Using an innovative, multi-modal data integration machine learning framework, this study provides confirmatory and novel information on mechanisms behind atherosclerotic instability. The findings raise possibilities for translational prioritizations to aid personalized medicine.