Scholarlink

Background <p>While large-scale pan-cancer studies have extensively characterized host genomic and molecular features, systematic analyses of microbial signatures across diverse cancer types remain scarce. This study aims to address this gap by employing 16&#xa0;S rRNA sequencing to profile intratumoral microbiota across multiple tumor types.</p> Methods <p>We integrated microbiome data from 11 published studies and one large validation cohort, encompassing 2,839 samples across nine cancer types and normal controls. In addition, we included an independent validation set consisting of 541 tumor samples collected at the Cancer Hospital, Chinese Academy of Medical Sciences. Using QIIME2 for preprocessing and DADA2 for ASV detection, we performed microbial diversity analysis, co-occurrence network construction, taxonomic and functional annotation (via PICRUSt2 and KEGG). Machine learning models were applied for diagnostic classification, with feature importance assessed via SHapley additive explanations.</p> Results <p>Significant differences in microbial diversity and composition were observed across cancer types. Colorectal, oral, and bladder cancers exhibited relatively high microbial diversity, whereas gastric cancer showed reduced diversity. Tumor-specific microbial signatures were identified, such as <i>Ralstonia</i> and <i>Treponema</i> enrichment in breast cancer, and <i>Fusobacterium</i> in bladder cancer. Functional pathway analysis revealed distinct metabolic and immune-related microbial profiles among cancers. Microbial co-occurrence networks varied by tumor type, with gastric cancer displaying the most complex interactions. A Random Forest model demonstrated strong classification accuracy, with <i>Pelomonas</i> and <i>Cutibacterium</i> as key features. Grouping tumors by biological or anatomical characteristics further improved diagnostic performance.</p> Conclusion <p>This largest multi-center study to date identifies distinct tumor-associated microbiota signatures across diverse cancers and demonstrates their strong diagnostic potential. Our findings provide novel insights into the ecology of tumor microbiomes and highlight the microbiota as potential therapeutic targets. This opens promising avenues for microbiome-informed diagnostic and therapeutic strategies in oncology, including the potential development of non-invasive liquid biopsy approaches based on microbial signatures.</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 pan-cancer analysis of intratumoral microbial communities for biomarker discovery and diagnostic modeling

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月25日

查看全文
  • Liuliu Quan,
  • Huiteng Rong,
  • Jian Yue,
  • Jing Xu,
  • Yujie Zheng,
  • Peng Yuan

摘要

Background

While large-scale pan-cancer studies have extensively characterized host genomic and molecular features, systematic analyses of microbial signatures across diverse cancer types remain scarce. This study aims to address this gap by employing 16 S rRNA sequencing to profile intratumoral microbiota across multiple tumor types.

Methods

We integrated microbiome data from 11 published studies and one large validation cohort, encompassing 2,839 samples across nine cancer types and normal controls. In addition, we included an independent validation set consisting of 541 tumor samples collected at the Cancer Hospital, Chinese Academy of Medical Sciences. Using QIIME2 for preprocessing and DADA2 for ASV detection, we performed microbial diversity analysis, co-occurrence network construction, taxonomic and functional annotation (via PICRUSt2 and KEGG). Machine learning models were applied for diagnostic classification, with feature importance assessed via SHapley additive explanations.

Results

Significant differences in microbial diversity and composition were observed across cancer types. Colorectal, oral, and bladder cancers exhibited relatively high microbial diversity, whereas gastric cancer showed reduced diversity. Tumor-specific microbial signatures were identified, such as Ralstonia and Treponema enrichment in breast cancer, and Fusobacterium in bladder cancer. Functional pathway analysis revealed distinct metabolic and immune-related microbial profiles among cancers. Microbial co-occurrence networks varied by tumor type, with gastric cancer displaying the most complex interactions. A Random Forest model demonstrated strong classification accuracy, with Pelomonas and Cutibacterium as key features. Grouping tumors by biological or anatomical characteristics further improved diagnostic performance.

Conclusion

This largest multi-center study to date identifies distinct tumor-associated microbiota signatures across diverse cancers and demonstrates their strong diagnostic potential. Our findings provide novel insights into the ecology of tumor microbiomes and highlight the microbiota as potential therapeutic targets. This opens promising avenues for microbiome-informed diagnostic and therapeutic strategies in oncology, including the potential development of non-invasive liquid biopsy approaches based on microbial signatures.