Front Pharmacol. 2025 ;16 1634985
Introduction: Lactylation, a post-translational modification characterized by the attachment of lactate to protein lysine residues on proteins, plays a pivotal role in cancer progression and immune evasion. However, its implications in immunity regulation and prostate cancer prognosis remains poorly understood. This study aims to systematically examine the impact of lactylation-related genes (LRGs) on prostate cancer.
Methods: Single-cell and bulk RNA sequencing data from patients with prostate cancer were analyzed. Data were sourced from TCGA-PRAD, GSE116918, and GSE54460, with batch effects mitigated using the ComBat method. LRGs were identified from exisiting literature, and unsupervised clustering was applied to assess their prognostic siginificance. The tumor microenvironment and functional enrichment of relevant pathways were also evaluated. A prognostic model was developed using integrative machine learning techniques, with drug sensitivy analysis included. The mRNA expression profiles of the top ten genes were validated in clinical samples.
Results: Single-cell RNA sequencing revealed distinct lactylation signatures across various cell types. Bulk RNA-seq analysis identified 56 prognostic LRGs, classifying patients into two distinct clusters with divergent prognoses. The high-risk cluster exhibited reduced immune cell infiltration and increased resistance to specific targeted therapies. A machine learning-based prognostic signature was developed, demonstrating robust predictive accuracy for treatment responses and disease outcomes.
Conclusion: This study offers a comprehensive analysis of lactylation in prostate cancer, identifying potential prognostic biomarkers. The proposed prognostic signature provides a novel approach to personalized treatment strategies, deepening our understanding of the molecular mechanisms driving prostate cancer and offering a tool for predicting therapeutic responses and clinical outcomes.
Keywords: immune microenvironment; lactylation; machine learning; personalized treatment; prognostic biomarker; prostate cancer