Identification of optimal feature genes in patients with thyroid associated ophthalmopathy and their relationship with immune infiltration: a bioinformatics analysis

**Background:** Thyroid associated ophthalmopathy (TAO) is an organ-specific autoimmune disease with a complex and poorly understood etiology. The study aimed to use bioinformatics to investigate TAO pathogenesis and identify optimal feature genes (OFGs) and immune infiltration patterns.

**Methods:** The GSE58331 microarray dataset was used, containing 27 TAO anterior orbital tissue samples and 22 normal anterior orbital tissue samples. Differential expression analysis identified 366 differentially expressed genes (DEGs) (69 upregulated, 297 downregulated; P<0.05, fold change ≥1.5). Weighted gene coexpression network analysis (WGCNA) with soft threshold β=26 identified seven coexpression modules; the MEgreen and MEblack modules (483 genes) showed the strongest correlation with TAO. The intersection of DEGs and WGCNA module genes yielded 63 key genes. Three machine learning methods (LASSO logistic regression, SVM-RFE, random forest) were applied to these 63 genes to identify OFGs. Immune cell infiltration was assessed using CIBERSORT, and single-sample gene set enrichment analysis (ssGSEA) was performed for pathway analysis.

**Key Results:** Six OFGs were identified: KLB, TBC1D2B, LINC01140, SGCG, TMEM37, and LINC01697, all downregulated in TAO. A nomogram based on these six genes had an AUC of 0.860. Individual ROC AUCs were: KLB 0.857, TBC1D2B 0.865, LINC01140 0.830, SGCG 0.847, TMEM37 0.781, LINC01697 0.859. Immune infiltration analysis showed increased memory B cells, T follicular helper cells, resting NK cells, M0 macrophages, M1 macrophages, resting dendritic cells, activated mast cells, and neutrophils in TAO samples, while M2 macrophages and resting mast cells were decreased. ssGSEA revealed positive correlations between OFGs and lipid metabolism pathways (heterogeneous biological metabolism, fatty acid metabolism, bile acid metabolism, adipogenesis, cholesterol homeostasis, peroxisome).

**Clinical Implications:** The six identified genes (KLB, TBC1D2B, LINC01140, SGCG, TMEM37, LINC01697) are potential diagnostic biomarkers for TAO. Their downregulation may contribute to orbital tissue changes: KLB downregulation could reduce FGF21-induced fat catabolism, TBC1D2B loss may promote fibrosis via E-cadherin degradation, and LINC01140/LINC01697 may regulate macrophage polarization. The immune infiltration findings suggest that targeting T follicular helper cells, macrophages, or mast cells could be therapeutic strategies. The association with lipid metabolism pathways highlights a potential role for metabolic interventions. However, these findings require validation in larger clinical studies and in vivo/in vitro experiments.