Development and validation of a novel sodium-overload related genes signature for prognostic prediction in breast cancer: integrating bioinformatics and experimental approaches.

Introduction: Necrosis induced by sodium overload has recently been identified as a novel form of regulated cell death. However, the specific genes associated with sodium overload in breast cancer (BC) remain uncharacterized. Methods: We identified 753 differentially expressed sodium-overload-related genes (DESORGs) in BC. We performed pathway enrichment analyses, then used univariate Cox regression to select 67 prognostic DESORGs. To build prognostic models, we tested 101 combinations of ten machine learning algorithms. SHAP analysis was used to determine feature importance. Mendelian randomization (MR) was applied to assess causal effects. Experimental validation (in vitro) included overexpression and knockdown studies. GSEA/GSVA and molecular docking were conducted to explore downstream pathways and potential drug candidates. Results: The ridge regression model showed optimal prognostic power. IFNG was identified as the key feature. The computed risk score was an independent prognostic factor, outperforming traditional clinical variables (AUC = 0.845), and a nomogram built with it yielded good calibration (C-index = 0.815). MR suggested a protective causal effect of NR1H3 in BC, and patients with high NR1H3 expression had significantly better overall survival (p = 0.02). Conclusions: These findings highlight NR1H3 as a novel DESORG and a promising therapeutic target in breast cancer.