Bioinformatics and machine learning-driven discovery of candidate tissue diagnostic markers for endometriosis with experimental verification.

BACKGROUND

Endometriosis is a complex gynecological disorder lacking reliable biomarkers. This study aimed to identify core diagnostic genes through integrated computational approaches. Multiple endometriosis transcriptomic datasets were analyzed.

METHODS

Differential expression analysis and Weighted Gene Co-expression Network Analysis (WGCNA) screened disease-associated genes. Functional enrichment, Protein-Protein Interaction (PPI) network construction, and an ensemble machine learning framework (113 algorithm combinations) were employed to refine hub genes. Diagnostic performance was validated via ROC analysis. Immune infiltration was characterized using CIBERSORT.

RESULTS

Four genes (COL6A3, BGN, LAMA4, THBS2) were identified as robust candidate tissue diagnostic markers, showing consistent upregulation and high discriminatory power (AUC > 0.80). They are implicated in extracellular matrix remodeling. Immune dysregulation was observed, featuring elevated M1 macrophages and plasma cells, alongside reduced resting NK cells, with hub genes correlating with specific immune subsets. To functionally validate these findings, a mice endometriosis model was established and exhibited histopathological features consistent with the disease, including ectopic lesion formation and altered glandular architecture. qPCR and Western blot analyses confirmed significant upregulation of COL6A3, BGN, LAMA4, and THBS2 at both transcriptional and protein levels in ectopic endometrium, further supporting their role in disease pathogenesis.

CONCLUSION

COL6A3, BGN, LAMA4, and THBS2 represent promising candidate tissue diagnostic markers for endometriosis, linked to extracellular matrix and immune microenvironment alterations, providing novel insights for future research and clinical translation.