Abstract
BACKGROUND: Preservatives, widely used in food and skincare products, may influence prostate cancer (PCa) development. This study explores the effects of common preservatives, especially parabens, on PCa and their potential molecular associations via computational and database-based analyses. METHODS: This study identified potential preservative targets linked to prostate cancer through database screening (Swiss Target Prediction, STITCH, GeneCards) and extracted overlapping genes for Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses. A core gene network was constructed via the STRING database and Cytoscape software, and the top 20 genes by interaction strength were further analyzed using 10 machine learning algorithms to develop an optimal prognostic model. Multivariate Cox regression identified key genes as independent prognostic factors, which were preliminarily evaluated via molecular docking for preservative binding affinity. Tissue expression differences of these genes were also confirmed using the Human Protein Atlas (HPA). RESULTS: This study identified 135 preservative-PCa-related genes; GO enrichment analysis showed these genes were mainly involved in apoptosis regulation, oxidative stress, signal transduction, and biosynthesis processes, while KEGG enrichment analysis linked them to endocrine resistance, chemical carcinogenesis, and lipid metabolism. The results of the machine learning prediction model showed that the Ridge model achieved the best prediction performance among the combinations of 101 prediction models with a C-index score of 0.709 and was validated across four external datasets (Cambridge, Taylor, CancerMap, GEO46602). Multivariate Cox regression identified 8 key genes (AR, BCL2L1, CASP3, CDK1, HDAC6, MMP2, PIK3CA, XIAP) as independent PCa prognostic factors-with AR, CASP3, CDK1, HDAC6, MMP2 as risk factors and BCL2L1, PIK3CA, XIAP as protective factors. Molecular docking showed all 8 genes could bind spontaneously to four parabens (methylparaben, ethylparaben, propylparaben, butylparaben), and HPA data confirmed significant expression differences of these genes between normal prostate and PCa tissues. CONCLUSION: This study uses computational and database-based approaches to systematically explore potential associations between parabens and PCa, identifying 8 key genes that may mediate this association and providing a theoretical foundation for formulating safer preservative usage guidelines and exploring PCa prognostic markers. Importantly, the current findings are derived from in silico prediction and public database analysis, not from experimental verification involving paraben exposure controls. The study identifies potential correlations rather than verifying direct molecular mechanisms of parabens in PCa; thus, it generates valuable scientific hypotheses that require further validation via in vitro cell experiments, animal models, and human exposure cohort studies to confirm the actual molecular mechanisms linking parabens to PCa.