Abstract
Precise diagnosis of early prostate cancer (PCa) is critical for preventing tumor progression. However, the diagnostic outcomes of currently used markers are far from satisfactory due to the low sensitivity or specificity. Here, we identified a diagnostic subpopulation in PCa tissue with the integrating analysis of single-cell and bulk RNA-seq. The representative markers of this subpopulation were extracted to perform intersection analysis with early-PCa-related gene module generated from weighted correlation network analysis (WGCNA). A total of 24 overlapping genes were obtained, the diagnostic roles of which were validated by distinguishing normal and tumorous prostate samples from the public dataset. A least absolute shrinkage and selection operator (LASSO) model was constructed based on these genes and the obtained 24-gene panel showed high sensitivity and specificity for PCa diagnosis, with better identifying capability of PCa than the commercially used gene panel of Oncotype DX. The top two risk factors, TRPM4 and PODXL2, were verified to be highly expressed in early PCa tissues by multiplex immunostaining, and PODXL2 was more sensitive and specific compared to TRPM4 and the pathologically used marker AMACR for early PCa diagnosis, suggesting a novel and promising pathology marker.