Serum miRNA-based diagnostic models for endometriosis: from discovery to validation

Area of Science:

• Biomarker discovery
• Molecular diagnostics
• Gynecological oncology

Background:

• Circulating microRNAs (miRNAs) show differential expression in endometriosis (END) patients versus controls (CTR).
• Existing research findings are conflicting, with no established miRNA-based diagnostic test for END.
• This study aims to develop and validate a miRNA signature for END diagnosis.

Purpose of the Study:

• To investigate if a serum miRNA signature can serve as a diagnostic biomarker for endometriosis.
• To develop and validate machine learning models for endometriosis diagnosis using serum miRNAs.
• To assess the diagnostic accuracy of miRNA-based models in distinguishing END patients from controls.

Main Methods:

• Serum samples from 364 patients (END and benign gynecological conditions) were analyzed using RT-qPCR for 23 miRNAs.
• Diagnostic models were developed using Random Forest and Logistic Regression algorithms.
• Internal validation was performed using repeated cross-validation.

Main Results:

• An 11-miRNA model achieved 65.8% accuracy (AUC 70.4%, sensitivity 75.6%, specificity 53.5%).
• A 6-miRNA model showed higher accuracy (75.9%, AUC 80.4%) for deep infiltrating endometriosis.
• A model for ovarian endometrioma achieved 62.4% accuracy (AUC 65.8%).

Conclusions:

• The developed miRNA-based models demonstrate potential for endometriosis diagnosis, particularly for deep infiltrating endometriosis.
• Further validation in larger, prospective cohorts is necessary to enhance accuracy and robustness.
• The model's low false-negative rate suggests utility as a screening tool for identifying patients requiring further evaluation.