Integrative Machine Learning Framework Revealing TRPM4-Associated Signatures and Identifying SPATA6 as a Potential Biomarker in Prostate Cancer.

Background: The non-selective cation channel TRPM4 can induce necrotic cell death through sodium overload, yet its role in prostate cancer (PCa) progression remains poorly characterized. Materials and methods: Using TCGA-PCa transcriptomic data centered on TRPM4, we identified transcriptional signatures linked to sodium overload. Leveraging prognostic features, we developed robust prognostic models via ten machine learning algorithms and their combinations, training on TCGA data and validating on internal validation set, GSE46602 and GSE116918. We assessed the model's associations with clinicopathological features, prognosis, immune infiltration, and drug response. Expression of the 10 key model genes was validated in PCa cell lines versus a normal prostate epithelial cell. For SPATA6-the top-contributing gene-we overexpressed it in PCa cells to assess its functional impact. Results: We identified 91 overlapping genes from TRPM4-associated and PCa-related differentially expressed genes. Functional enrichment implicated these genes in small GTPase activity, Rap1 signaling, and cAMP signaling. A TRPM4-related signature model (TRSM) comprising 10 key genes demonstrated strong prognostic performance across training and validation cohorts. TRSM-based risk stratification revealed significant differences in disease-free survival, clinicopathological features, immune infiltration, and immunotherapy response. Drug sensitivity analysis indicated heightened docetaxel sensitivity in the high-risk group. Conclusion: Our findings underscore the importance of TRPM4-associated molecular features in PCa prognosis. TRSM shows potential as a predictive tool for patient outcomes and a guide for personalized therapy.