Scholarlink

Background <p>Renal cell carcinoma (RCC), which accounts for 70–90% of kidney malignancies, remains difficult to diagnose early due to its asymptomatic onset and the lack of reliable biomarkers. This study aimed to develop a robust diagnostic model by integrating plasma and urine metabolomics profiling.</p> Methods <p>A total of 482 plasma and 434 urine samples from RCC patients, benign renal disease cases, other urological cancers, and healthy controls were analyzed using multi-platform mass spectrometry. Participants were assigned to a discovery or validation cohort to identify RCC-specific metabolites and construct diagnostic models.</p> Results <p>Twenty-six plasma and twelve urine metabolites were selected to build individual models. The integrated plasma-urine model achieved superior diagnostic accuracy (AUC = 0.88) compared with plasma (AUC = 0.86) and urine (AUC = 0.78) models in the validation cohort, with notably improved sensitivity for early-stage RCC. In asymptomatic screening populations, it performed excellently (AUC = 0.94) and maintained high specificity, yielding significantly lower scores for other cancer types (<i>p</i> &lt; 0.01). Pathway analysis identified glycine, serine, and threonine metabolism as the key dysregulated pathway shared across plasma and urine, suggesting a potential therapeutic target.</p> Conclusion <p>This study demonstrates that integrating plasma and urine metabolomics with machine learning yields a robust, non-invasive diagnostic model for renal cell carcinoma. The combined plasma-urine panel outperformed single-fluid models, achieving high accuracy and specificity, and maintained stable performance across tumor stages, grades, and clinical subgroups. The identified metabolic signatures, particularly alterations in glycine-serine-threonine metabolism pathway, provide novel insights into RCC metabolic reprogramming. These findings support the model’s potential for early detection and clinical application, while also offering a basis for future therapeutic exploration.</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

Development and validation of a plasma-urine metabolism diagnostic model for renal cell carcinoma using machine learning

  • 期刊论文

  • 发表日期：

    2025年12月9日

查看全文
  • Zhenkun Dong,
  • Kun Zhai,
  • Bingzhi Geng,
  • Qiang Li,
  • Zhaodu Liu,
  • Fei Shi,
  • Yun He,
  • Hui Chen,
  • Yan Cui

摘要

Background

Renal cell carcinoma (RCC), which accounts for 70–90% of kidney malignancies, remains difficult to diagnose early due to its asymptomatic onset and the lack of reliable biomarkers. This study aimed to develop a robust diagnostic model by integrating plasma and urine metabolomics profiling.

Methods

A total of 482 plasma and 434 urine samples from RCC patients, benign renal disease cases, other urological cancers, and healthy controls were analyzed using multi-platform mass spectrometry. Participants were assigned to a discovery or validation cohort to identify RCC-specific metabolites and construct diagnostic models.

Results

Twenty-six plasma and twelve urine metabolites were selected to build individual models. The integrated plasma-urine model achieved superior diagnostic accuracy (AUC = 0.88) compared with plasma (AUC = 0.86) and urine (AUC = 0.78) models in the validation cohort, with notably improved sensitivity for early-stage RCC. In asymptomatic screening populations, it performed excellently (AUC = 0.94) and maintained high specificity, yielding significantly lower scores for other cancer types (p < 0.01). Pathway analysis identified glycine, serine, and threonine metabolism as the key dysregulated pathway shared across plasma and urine, suggesting a potential therapeutic target.

Conclusion

This study demonstrates that integrating plasma and urine metabolomics with machine learning yields a robust, non-invasive diagnostic model for renal cell carcinoma. The combined plasma-urine panel outperformed single-fluid models, achieving high accuracy and specificity, and maintained stable performance across tumor stages, grades, and clinical subgroups. The identified metabolic signatures, particularly alterations in glycine-serine-threonine metabolism pathway, provide novel insights into RCC metabolic reprogramming. These findings support the model’s potential for early detection and clinical application, while also offering a basis for future therapeutic exploration.