Differential gene expression profiling and machine learning-based discovery of key genetic markers in VTE and CKD

Venous thromboembolism (VTE) and chronic kidney disease (CKD) are multifactorial disorders characterized by complex genetic and molecular mechanisms. However, their shared genetic signatures and potential interrelations remain poorly understood. This study aimed to identify key genes and molecular pathways linking VTE and CKD through comprehensive transcriptomic and machine learning analyses.

Gene expression profiles from patients with VTE and CKD, along with corresponding controls, were analyzed to identify differentially expressed genes (DEGs). Functional enrichment analyses were performed using Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. The intersection of DEGs between VTE and CKD was used for feature selection via three machine learning algorithms: Least Absolute Shrinkage and Selection Operator (LASSO), Support Vector Machine-Recursive Feature Elimination (SVM-RFE), and Random Forest (RF). A diagnostic nomogram was constructed based on key genes, followed by receiver operating characteristic (ROC) curve analysis, gene set enrichment analysis (GSEA), and immune infiltration assessment. Validation was performed using independent datasets (GSE37171 and GSE48000) and single-cell RNA sequencing data.

A total of 637 DEGs (413 upregulated and 224 downregulated) were identified in VTE patients, and 671 DEGs (99 upregulated and 572 downregulated) were identified in CKD patients. Enrichment analyses revealed that VTE DEGs were primarily involved in cytoplasmic translation, immune activation, and oxidative phosphorylation, while CKD DEGs were enriched in muscle contraction regulation, ATPase activity, and vascular smooth muscle contraction. Twenty-three overlapping DEGs were found between CKD and VTE, including CCNL2, HNRNPA0, PI4KA, FOS, and HBD. Machine learning analyses identified HNRNPA0 and PI4KA as the most robust feature genes, both exhibiting excellent diagnostic performance (AUC = 1.000). A diagnostic nomogram based on these genes showed strong predictive accuracy and calibration. GSEA and immune infiltration analyses revealed their involvement in immune-related and metabolic pathways. Validation in external datasets confirmed significantly lower expression of HNRNPA0 and PI4KA in CKD samples. Single-cell RNA sequencing further delineated their expression across 11 cellular clusters corresponding to eight major cell types.

This study identifies HNRNPA0 and PI4KA as key genes shared between VTE and CKD, providing new insights into their genetic and immunological links. The diagnostic model based on these genes offers a promising tool for CKD prediction and highlights potential targets for future mechanistic and therapeutic investigations.