Scholarlink

Background <p>Disinfection by-products (DBPs) are ubiquitous contaminants generated during drinking-water disinfection and have been associated with adverse health outcomes. Although epidemiological and toxicological studies have linked long-term DBP exposure to urinary tract malignancies, their potential molecular relevance to prostate cancer (PRAD) remains insufficiently characterized. Because direct DBP exposure information is not available in most public PRAD transcriptomic cohorts, integrative computational analyses may help prioritize candidate molecular links between DBP-related targets and PRAD-associated gene signatures.</p> Methods <p>DCBQ and NDMA were selected as two DBPs from distinct chemical classes. Putative DBP-related targets were collected from public target-prediction and toxicogenomic databases. Public PRAD transcriptomic datasets were analyzed using differential expression analysis and WGCNA to identify PRAD-associated genes and co-expression modules. The statistical enrichment of the overlap between PRAD DEGs and predicted DBP-related targets was assessed using hypergeometric and Fisher’s exact tests. Candidate genes were then obtained by intersecting predicted DBP-related targets, PRAD-related genes, DEGs, and WGCNA module genes, followed by PPI analysis and machine-learning-based feature prioritization using LASSO regression, random forest, and ANN feature weighting. Diagnostic performance was externally evaluated in GSE46602, and clinical/prognostic relevance was explored in TCGA-PRAD. Single-cell RNA-seq, molecular docking, and 100-ns molecular dynamics simulations were used to characterize cell-type distribution and potential compound–protein interactions.</p> Results <p>Overlap enrichment analysis showed that PRAD DEGs were significantly enriched for predicted DBP-related targets, with 85 observed overlapping genes compared with 35.38 expected by chance (<i>P</i> = 4.65 × 10⁻¹⁴; odds ratio = 2.74). The integrative analysis further identified 34 candidate intersecting genes, of which 24 PPI-connected genes were retained for network and machine-learning analyses. LASSO regression, random forest, and ANN-based feature weighting prioritized SPP1, ANGPT1, ANXA1, and RHOB. External validation in GSE46602 supported their diagnostic potential, with AUC values of 0.704, 0.939, 0.953, and 0.719, respectively. TCGA-PRAD analysis further linked ANGPT1 and SPP1 to progression-free interval. Enrichment analysis highlighted epithelial differentiation, extracellular matrix organization, focal adhesion, and tumor microenvironment-related pathways. Single-cell analysis localized SPP1 mainly to myeloid cells and ANGPT1 mainly to fibroblasts. Docking and molecular dynamics simulations suggested a relatively stable predicted interaction between DCBQ and ANGPT1.</p> Conclusions <p>This study provides an integrative computational prioritization of candidate genes and pathways linking predicted DBP-related toxicological targets with PRAD-associated molecular alterations. The findings highlight ANGPT1- and SPP1-related tumor microenvironment remodeling as a possible molecular context for future investigation of drinking-water DBPs and prostate cancer. These results provide hypothesis-generating candidate targets for exposure-informed epidemiological studies and experimental validation under environmentally relevant conditions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integrative network toxicology and transcriptomic analysis identifies candidate molecular links between drinking-water disinfection by-products and prostate cancer

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年8月5日

查看全文
  • Qilong Wu,
  • Xinyao Zhu,
  • Zhiyu Liu,
  • Yuqi Li,
  • Tao Zhou,
  • Qingfu Deng

摘要

Background

Disinfection by-products (DBPs) are ubiquitous contaminants generated during drinking-water disinfection and have been associated with adverse health outcomes. Although epidemiological and toxicological studies have linked long-term DBP exposure to urinary tract malignancies, their potential molecular relevance to prostate cancer (PRAD) remains insufficiently characterized. Because direct DBP exposure information is not available in most public PRAD transcriptomic cohorts, integrative computational analyses may help prioritize candidate molecular links between DBP-related targets and PRAD-associated gene signatures.

Methods

DCBQ and NDMA were selected as two DBPs from distinct chemical classes. Putative DBP-related targets were collected from public target-prediction and toxicogenomic databases. Public PRAD transcriptomic datasets were analyzed using differential expression analysis and WGCNA to identify PRAD-associated genes and co-expression modules. The statistical enrichment of the overlap between PRAD DEGs and predicted DBP-related targets was assessed using hypergeometric and Fisher’s exact tests. Candidate genes were then obtained by intersecting predicted DBP-related targets, PRAD-related genes, DEGs, and WGCNA module genes, followed by PPI analysis and machine-learning-based feature prioritization using LASSO regression, random forest, and ANN feature weighting. Diagnostic performance was externally evaluated in GSE46602, and clinical/prognostic relevance was explored in TCGA-PRAD. Single-cell RNA-seq, molecular docking, and 100-ns molecular dynamics simulations were used to characterize cell-type distribution and potential compound–protein interactions.

Results

Overlap enrichment analysis showed that PRAD DEGs were significantly enriched for predicted DBP-related targets, with 85 observed overlapping genes compared with 35.38 expected by chance (P = 4.65 × 10⁻¹⁴; odds ratio = 2.74). The integrative analysis further identified 34 candidate intersecting genes, of which 24 PPI-connected genes were retained for network and machine-learning analyses. LASSO regression, random forest, and ANN-based feature weighting prioritized SPP1, ANGPT1, ANXA1, and RHOB. External validation in GSE46602 supported their diagnostic potential, with AUC values of 0.704, 0.939, 0.953, and 0.719, respectively. TCGA-PRAD analysis further linked ANGPT1 and SPP1 to progression-free interval. Enrichment analysis highlighted epithelial differentiation, extracellular matrix organization, focal adhesion, and tumor microenvironment-related pathways. Single-cell analysis localized SPP1 mainly to myeloid cells and ANGPT1 mainly to fibroblasts. Docking and molecular dynamics simulations suggested a relatively stable predicted interaction between DCBQ and ANGPT1.

Conclusions

This study provides an integrative computational prioritization of candidate genes and pathways linking predicted DBP-related toxicological targets with PRAD-associated molecular alterations. The findings highlight ANGPT1- and SPP1-related tumor microenvironment remodeling as a possible molecular context for future investigation of drinking-water DBPs and prostate cancer. These results provide hypothesis-generating candidate targets for exposure-informed epidemiological studies and experimental validation under environmentally relevant conditions.