PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
April 3, 2026Journal of Translational Medicine0 citationsOpen Access

Oxidative stress reprograms benign prostatic hyperplasia microenvironments: insights from integrative multi-omics and machine learning

JCJie ChenJBJingxing BaiBCBo Chen

Key Points

  • The aim is to uncover the role of oxidative stress in benign prostatic hyperplasia and identify associated biomarkers.
  • Integrated multi-omics analysis of transcriptomic data from BPH and control samples.
  • Application of machine learning algorithms to prioritize biomarker candidates.
  • Validation through experimental approaches including cell culture and immunohistochemistry.
  • Single-cell RNA sequencing and spatial transcriptomics for detailed cellular insights.
  • Identified 499 differentially expressed genes (DEGs), with 26 associated with oxidative stress.
  • Machine learning identified ACOX2, CTSB, and SERPINF1 as promising hub genes with high diagnostic potential.
  • Two distinct BPH subtypes were recognized based on immune infiltration.
  • Experimental validation demonstrated ACOX2's role in promoting cell proliferation and reducing apoptosis under oxidative stress.

Abstract

Oxidative stress (OS) is increasingly implicated in benign prostatic hyperplasia (BPH), yet the underlying cellular programs remain unclear. We integrated multi-omics and machine learning to identify OS-associated biomarkers and to characterize OS-linked stromal states, with targeted experimental validation. Two bulk transcriptomic datasets (12 BPH, 16 controls) were integrated to identify DEGs and OS-associated DEGs, followed by WGCNA, enrichment analyses, and four machine-learning algorithms to prioritize hub genes and build a diagnostic nomogram. Consensus clustering defined molecular subtypes. Candidate compounds were screened in silico. Single-cell RNA-seq (124,616 cells) and spatial transcriptomics were analyzed for OS scoring, trajectories, and inferred intercellular communication. Experimentally, WPMY-1 prostatic stromal cells were exposed to LPS to induce OS; intracellular ROS (flow cytometry), ACOX2 expression (RT-qPCR/western blot), proliferation (EdU), and apoptosis (Annexin V) were assessed under ACOX2 overexpression or shRNA knockdown. ACOX2 expression in human prostate tissues was evaluated by immunohistochemistry. We identified 499 DEGs and 26 OS-DEGs. Machine learning converged on three upregulated hub genes: ACOX2, CTSB, and SERPINF1, which with diagnostic AUCs of 0.880 (0.753-1.000), 0.828 (0.663–0.993), and 0.854 (0.711–0.997); however, given the modest sample size, these findings should be interpreted as hypothesis-generating. Two BPH subtypes were identified (immune-infiltrated vs. non-immune). Single-cell analyses showed elevated OS scores across cell types, highest in fibroblasts; ACOX2-high fibroblasts were expanded in BPH and exhibited trajectory-associated increases in ACOX2 with enriched inferred signaling (including TNFSF12-TNFRSF12A). Spatial transcriptomics revealed regional OS heterogeneity and spatial association of hub-gene expression with OS-enriched areas. In vitro, LPS increased ROS and upregulated ACOX2, ACOX2 overexpression increased proliferation and reduced apoptosis, whereas knockdown showed opposite effects and attenuated LPS-associated proliferative phenotypes. IHC showed stronger ACOX2 staining in hyperplastic vs. normal prostate tissues. These results nominate ACOX2, CTSB, and SERPINF1 as candidate OS-associated markers in BPH and support a hypothesis-generating multi-omics signature that warrants validation in larger independent human cohorts. Experimental perturbation and tissue staining provide validation for an ACOX2-associated activation-like phenotype under OS, motivating biomarker-guided stratification and OS-targeted therapeutic exploration.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Chen et al. (2026) studied this question.

synapsesocial.com/papers/69cf5d055a333a821460aabdhttps://doi.org/10.1186/s12967-026-08055-8
Ask AI
Helpful
Bookmark
Share
View Full Paper