Abstract
BACKGROUND: Prostate cancer (PCa) is a prevalent malignancy in men, with exosomes playing a key role in tumor microenvironment and disease progression, yet their molecular mechanisms remain unclear.This study aims to identify potential biomarkers and therapeutic targets in PCa by integrating exosome-related genes with differentially expressed genes (DEGs). METHODS: Four GEO datasets (GSE32448, GSE46602, GSE69223, GSE6956) were analyzed. Batch effects were corrected using the ComBat method, followed by DEG analysis and feature selection via machine learning (LASSO regression, random forest, SVM). Functional enrichment and molecular docking validated the findings. RESULTS: Post-correction, sample clustering improved significantly. Of 49 overlapping DEGs and exosome-related genes, EEF2, LGALS3, and MYO1D emerged as key biomarkers, with EEF2 showing the highest predictive power (AUC = 0.786). A risk score model achieved an AUC of 0.886. Immune analysis linked these genes to immune cell subsets, and docking studies revealed strong interactions with small molecules like cycloheximide. CONCLUSION: This study elucidates the molecular role of exosome-related genes in PCa, proposing predictive biomarkers and novel therapeutic targets, warranting further clinical validation.