Objective To explore the gene network underlying EMs，identify potential diagnostic biomarkers，validate their diagnostic efficacy using clinical serum specimens，and elucidate the association between core genes and immune microenvironment. Methods Transcriptomic datasets GSE7305 （10 healthy controls， 10 EMs cases） and GSE51981 （34 healthy controls， 49 EMs cases） from the Gene Expression Omnibus （GEO） database were integrated. Batch effects were corrected by the ComBat algorithm， and differentially expressed genes were screened. Weighted Gene Co-expression Network Analysis （WGCNA） was used to identify EMs-related molecular modules，and 113 machine learning algorithms were employed to construct a diagnostic model， whose efficacy was validated by bootstrap， calibration curves and clinical decision curves. CIBERSORT was applied to analyze immune cell infiltration; combined with single-sample Gene Set Enrichment Analysis （ssGSEA） and Shapley Additive exPlanations （SHAP） analysis， the functions of core genes and their diagnostic contributions were clarified. Additionally， 30 EMs patients and 20 non-EMs controls who visited Shengjing Hospital of China Medical University from March to May 2025 were selected， and the expression of core genes and their encoded proteins in serum was verified by enzyme-linked immunosorbent assay （ELISA）， polymerase chain reaction （PCR） and Western blot. Results Compared with healthy people，a total of 711 DEGs were identified in EMs patients （308 upregulated；403 downregulated）. WGCNA indicated that the greenyellow module had the strongest positive correlation with EMs phenotype. The top 10 genes of the 58 DEGs in this module，ranked by absolute fold change，were evaluated through 113 machine learning models，with random forest （RF） demonstrating optimal diagnostic efficacy for EMs （C-index=0.991）. Subsequently，core genes C10orf54，CALCOCO1，ADAT1，and KIF21A were screened out. SHAP analysis indicated C10orf54 as the primary contributor to disease prediction，and the nomogram model showed that C10orf54 had the highest weight in the disease progression of EMs. Clinical specimens validated that the expression of C10orf54 mRNA was downregulated in the serum of EMs patients （P<0.05）， and C10orf54-encoded protein VISTA was significantly upregulated in the serum of EMs patients（P<0.0001）. ssGSEA revealed that C10orf54 was involved in the regulation of copper ion stress and cell cycle. Immune cell infiltration analysis indicated that the proportions of activated immune cells such as memory B cells and CD8⁺ T cells in EMs patients were increased； what's more，C10orf54 expression was positively correlated with activated immune cells. Conclusions C10orf54 may serve as a potential diagnostic biomarker for EMs. Its aberrant expression is closely associated with dysregulation of the immune microenvironment，providing a novel molecular target for early diagnosis and immune-targeted therapy of EMs.