BACKGROUND: Endometriosis (EMs) is a chronic inflammatory disease, and the lack of candidate molecular biomarkers remain a major clinical challenge. Circadian rhythm disruption plays a non-negligible role in EMs. However, the diagnostic potential of circadian rhythm-related biomarkers remains insufficiently validated. METHODS: We performed weighted gene co-expression network analysis to screen for endometriosis-related hub modules. Four machine learning algorithms were used for feature selection in the training cohort, and the resulting diagnostic model was evaluated in three independent GEO validation cohorts. To evaluate the immune microenvironment, we applied MCPcounter, CIBERSORT, and single-sample gene set enrichment analysis. Single‑cell RNA‑seq data were processed with Seurat for cell clustering and annotation, cell‑cell communication networks were constructed using CellChat, and pseudotime trajectory analysis was performed with Monocle. RESULTS: SMC2, SUV39H2, and TIMELESS were identified as candidate molecular biomarkers, demonstrating favorable diagnostic performance across three independent validation sets. Computational immune infiltration analysis suggested higher enrichment scores for B cells and CD8+ T cells, along with lower scores for Tregs and NK cells in EMs samples. Single-cell analysis revealed an increased proportion of mesenchymal cells in EMs samples and altered cell-cell communication involving BMP, CHEMERIN, and GRN signaling pathways. Pseudotime analysis further suggested transcriptional relationships among distinct stromal cell populations. CONCLUSION: This study identified SMC2, SUV39H2, and TIMELESS as candidate tissue-based molecular biomarkers for EMs and revealed alterations in stromal cell populations and immune-related features associated with the disease.