Identification and validation of palmitoylation-related signature genes based on machine learning for prostate cancer.

Abstract

Prostate cancer (PCa) remains a leading cause of cancer-related mortality in men, with challenges in diagnosis and treatment due to tumor heterogeneity. This study identifies palmitoylation-related signature genes as potential diagnostic and therapeutic targets. Integrating GEO datasets, six differentially expressed genes (DEGs) linked to palmitoylation were identified. Machine learning algorithms (LASSO, RF, SVM) selected three core genes: TRPM4, LAMB3, and APOE. A diagnostic model based on these genes achieved an AUC of 0.929, demonstrating robust accuracy in distinguishing PCa from normal tissues. Functional analysis revealed roles in lipid metabolism and immune modulation, with ssGSEA highlighting correlations between key genes and immune cell infiltration. Experimental validation showed that LAMB3 overexpression suppressed PCa cell proliferation, migration, and invasion, while knockdown enhanced these processes. Molecular docking identified diethylstilbestrol as a potential therapeutic agent targeting LAMB3 and APOE. These findings emphasize the clinical relevance of palmitoylation-related genes in PCa diagnosis and therapy, offering novel biomarkers and insights for personalized treatment strategies.