Identification and validation of cellular senescence biomarkers in recurrent pregnancy loss using machine learning and single-cell RNA sequencing

Recurrent pregnancy loss (RPL) is a major reproductive health problem in women of childbearing age, and its molecular mechanisms remain incompletely understood. This study aimed to identify key cellular senescence–related biomarkers for RPL and to evaluate their diagnostic value. Multiple RPL-related transcriptomic datasets were retrieved from the Gene Expression Omnibus (GEO) database, integrated, and corrected for batch effects. Differential expression analysis was performed, and the results were intersected with senescence-associated genes from the CellAge database to obtain senescence-related differentially expressed genes (DEARGs). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses, protein–protein interaction (PPI) network construction, and gene set enrichment analysis (GSEA) were conducted to explore potential functional pathways. Multiple machine-learning algorithms were then applied to screen for hub genes, yielding five key senescence-related genes: CYB5R3, LEO1, PAK2, RHOA, and GADD45G. A nomogram model was developed based on these genes, and its diagnostic performance was assessed using calibration curves, decision curve analysis, and receiver operating characteristic (ROC) analysis. Single-cell RNA-sequencing analysis indicated that GADD45G was highly expressed in decidual stromal cells, and virtual knockout analysis suggested that it may be involved in extracellular matrix remodeling, adhesion, and related signaling pathways. In addition, drug enrichment analysis predicted potential candidate small-molecule compounds. Finally, quantitative real-time PCR (qRT-PCR) validation in clinical decidual tissue samples showed that CYB5R3, PAK2, RHOA, and GADD45G were significantly upregulated in RPL, whereas LEO1 was significantly downregulated. Collectively, this study identified key senescence-related genes associated with RPL and established a diagnostic model with potential clinical utility, providing new insights for mechanistic research and precision diagnosis and treatment of RPL.