Machine learning identification of key genes in cardioembolic stroke and atherosclerosis: their association with pan-cancer and immune cells.

Background: Cardioembolic stroke (CS) and atherosclerosis (AS) are closely related diseases. Ferroptosis, a novel form of programmed cell death, may play a key role in CS and AS. However, the pathophysiological mechanisms underlying their coexistence remain unclear. This study aims to identify the hub genes and pathways involved in developing both diseases.

Methods: CS (GSE58294) and AS (GSE20129) datasets were obtained from the Gene Expression Omnibus database, and a ferroptosis (FR)-related gene dataset was downloaded from the FR database. A study was conducted to examine differentially expressed genes (DEGs) in healthy individuals and patients diagnosed with CS and AS. Gene ontology and Kyoto encyclopedia of genes and genomes analyses were performed to explore the functions of common FR-related DEGs (FRDEGs). Two machine learning algorithms, Least Absolute Shrinkage and Selection Operator (LASSO) regression and Support Vector Machine Recursive Feature Elimination (SVM-RFE), were used to screen for overlapping FRDEGs in CS and AS. To validate the prediction results, blood samples were collected from healthy controls and patients with CS and AS for quantitative real-time PCR. The correlation between biomarkers and clinical features was also evaluated.

Results: A total of 69 and 39 FRDEGs were identified in CS and AS, respectively. The hub genes, CIRBP, CREB5, MAPK14, PEBP1, and PTGS2, were identified using multiple methods. The area under the curve was > 0.7 for both models constructed using CS and AS datasets. A strong correlation was observed between neutrophil levels and expression of the hub genes. Additionally, several types of cancer indicated elevated expression of these hub genes compared to normal tissues.

Conclusions: In summary, the diagnostic model based on the FR-related gene PTGS2 demonstrated significant and specific diagnostic value for CS and AS, reflecting the status of blood lymphocytes, monocytes, and neutrophils. A pan-cancer study suggested it could serve as a new clinical prognostic marker and therapeutic target across various cancer types. This model may aid in the diagnosis of CS and AS. The findings offer new insights into the pathogenesis of these diseases.