Discovery and validation of RAB3B as a diagnostic biomarker for prostate cancer with serum PSA below 10 ng/mL based on a multi-omics study

The diagnostic rate of prostate cancer (PCa) remains suboptimal, particularly for patients whose serum prostate-specific antigen (PSA) concentration is less than 10 ng/mL. Single-cell sequencing, machine learning, and Mendelian randomization have emerged in recent years as excellent methods for identifying tumour markers. We identified 3 genes (RAB3B, SPON2, and SLC4A4) through single-cell sequencing and a machine learning screen. Among them, RAB3B (Ras-related protein, Rab-3B) had a higher area under the diagnostic curve for PSA < 10 ng/mL PCa (0.763) than the other two genes did. Mendelian randomization analysis revealed that the occurrence of PCa led to an increase in circulating levels of the RAB3B protein. Western blot and immunohistochemical analyses demonstrated that RAB3B was overexpressed in PCa tumour. Subsequently, elevated levels of RAB3B were detected in the culture medium of PCa cells. Furthermore, urinary RAB3B protein levels were significantly increased in PCa patients. For patients with PSA levels < 10 ng/mL, urinary RAB3B exhibited an area under the receiver operating characteristic curve (AUC) of 0.733 for PCa diagnosis, outperforming both PSA concentration (AUC = 0.682) and age (AUC = 0.651). Notably, the AUC of the multivariate model urine RAB3B + PSA + Age is 0.8. GSEA at al analysis and functional experiments, including gain-of-function and loss-of-function studies revealed that RAB3B promotes the proliferation and migration of PCa cells. In summary, urinary RAB3B, discovered through a multiomics approach, is a promising biomarker that is complementary to serum PSA, and together these factors can improve diagnostic decision-making in the PSA grey zone.