Scholarlink

Background <p>Endometriosis is a chronic gynecological disease associated with pain, infertility, and delayed diagnosis. Non-invasive biomarkers are urgently needed to facilitate earlier detection and reduce the reliance on diagnostic laparoscopy. MicroRNAs (miRNAs) are stable in body fluids and hold promise as diagnostic tools.</p> Methods <p>In this pilot study, urine samples from 34 patients with histologically confirmed endometriosis and 18 control patients (laparoscopically confirmed absence of disease) were analyzed using next-generation miRNA sequencing. Differential expression analysis was performed with DESeq2. Feature selection applied variance filtering, univariate analysis, mutual information, and recursive feature elimination (RFE). The top 20 miRNAs were used to train four classification models: logistic regression, decision tree, random forest, and support vector machine (SVM). Model performance was evaluated by accuracy, precision, recall, F1-score, and area under the ROC curve (AUC).</p> Results <p>Among all detected miRNAs, hsa-miR-10400-5p was significantly downregulated in endometriosis compared to controls (log₂ fold change − 2.70; adjusted <i>p</i> = 0.015). RFE identified 20 miRNAs, including hsa-mir-183, hsa-mir-500a, hsa-miR-3184-5p, hsa-miR-151b, and hsa-mir-196a-1, as the most informative for classification. The random forest model achieved the best performance (AUC = 0.91; accuracy and F1-score = 0.81), outperforming logistic regression (AUC = 0.83) and SVM (AUC = 0.81). Several identified miRNAs have been previously implicated in endometriosis pathogenesis, and we additionally identified hsa-miR-10400-5p as a significantly downregulated and previously unreported biomarker candidate, representing a novel finding with potential diagnostic relevance.</p> Conclusions <p>Urinary miRNA profiling, combined with machine learning, shows promise as a completely non-invasive approach for the diagnosis of endometriosis. The identified miRNA signature, particularly the novel hsa-miR-10400-5p, warrants validation in larger, independent cohorts to confirm its clinical utility and potential to reduce diagnostic delays.</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

Urinary microRNAs for the non-invasive diagnosis of endometriosis identified by next-generation sequencing and machine learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月7日

查看全文
  • Tomas Kupec,
  • Julia Wittenborn,
  • Chao-Chung Kuo,
  • Laila Najjari,
  • Rebecca Senger,
  • Philipp Meyer-Wilmes,
  • Elmar Stickeler,
  • Jochen Maurer

摘要

Background

Endometriosis is a chronic gynecological disease associated with pain, infertility, and delayed diagnosis. Non-invasive biomarkers are urgently needed to facilitate earlier detection and reduce the reliance on diagnostic laparoscopy. MicroRNAs (miRNAs) are stable in body fluids and hold promise as diagnostic tools.

Methods

In this pilot study, urine samples from 34 patients with histologically confirmed endometriosis and 18 control patients (laparoscopically confirmed absence of disease) were analyzed using next-generation miRNA sequencing. Differential expression analysis was performed with DESeq2. Feature selection applied variance filtering, univariate analysis, mutual information, and recursive feature elimination (RFE). The top 20 miRNAs were used to train four classification models: logistic regression, decision tree, random forest, and support vector machine (SVM). Model performance was evaluated by accuracy, precision, recall, F1-score, and area under the ROC curve (AUC).

Results

Among all detected miRNAs, hsa-miR-10400-5p was significantly downregulated in endometriosis compared to controls (log₂ fold change − 2.70; adjusted p = 0.015). RFE identified 20 miRNAs, including hsa-mir-183, hsa-mir-500a, hsa-miR-3184-5p, hsa-miR-151b, and hsa-mir-196a-1, as the most informative for classification. The random forest model achieved the best performance (AUC = 0.91; accuracy and F1-score = 0.81), outperforming logistic regression (AUC = 0.83) and SVM (AUC = 0.81). Several identified miRNAs have been previously implicated in endometriosis pathogenesis, and we additionally identified hsa-miR-10400-5p as a significantly downregulated and previously unreported biomarker candidate, representing a novel finding with potential diagnostic relevance.

Conclusions

Urinary miRNA profiling, combined with machine learning, shows promise as a completely non-invasive approach for the diagnosis of endometriosis. The identified miRNA signature, particularly the novel hsa-miR-10400-5p, warrants validation in larger, independent cohorts to confirm its clinical utility and potential to reduce diagnostic delays.