Purpose This study aimed to identify differentially expressed genes (DEGs) and pathways in benign prostatic hyperplasia (BPH) by comprehensive bioinformatics analysis. Methods Data of the gene expression microarray (GSE6099) were downloaded from GEO database. DEGs were obtained by GEO2R. Functional and enrichment analyses of selected genes were performed using DAVID database. Protein-protein interaction network was constructed through STRING. Anterior gradient 2 (ARG2) and lumican (LUM) staining in paraffin-embedded specimens from BPH and normal prostate (NP) were detected by immunohistochemistry (IHC). Differences between groups were analyzed by the Student's t test. Results A total of 24 epithelial DEGs and 39 stromal DEGs were determined. The GO analysis results showed that epithelial DEGs between BPH and NP were enriched in biological processes of glucose metabolic process, glucose homeostasis and negative regulation of Rho protein signal transduction. For DEGs in stroma, enriched biological processes included response to ischemia, antigen processing and presentation, cartilage development, T cell costimulation and energy reserve metabolic process. ARG2, as one of the epithelial DEGs, was mainly located in epithelial cells of prostate. In addition, LUM is primarily expressed in the stroma. We further confirmed that compared with NP, the BPH have the lower ARG2 protein level (p = 0.029) and higher LUM protein level (p = 0.003) using IHC. Conclusions Our study indicated that there are possible differentially expressed genes in epithelial and stromal cells, such as ARG2 and LUM, which may provide a novel insight for the pathogenesis of BPH.