Scholarlink

Background <p>Thyroid eye disease (TED) is an autoimmune disorder characterized by persistent inflammation around the periphery and within the orbit, potentially driven by hypoxic conditions. Effective biomarkers and precise predictive models are still lacking for the early diagnosis of TED.</p> Methods <p>Bulk RNA sequencing was conducted on peripheral blood samples from TED patients, Graves’ hyperthyroidism (GH) patients without ocular involvement, and healthy controls (HC). Differentially expressed genes between TED and HC, hypoxia-related genes and genes identified through weighted gene co-expression network analysis (WGCNA) were intersected to identify candidate biomarkers. Subsequently, nine machine learning algorithms were applied to screen for critical hypoxia-related TED diagnostic genes (HRTDGs). A diagnostic model based on HRTDG score (HRTDGS) was constructed using logistic regression analyses and then evaluated. TED patients were categorized into high and low HRTDGS groups based on the median score. Distinct immunological profiles and underlying pathological functions were investigated between two groups. Single cell RNA sequencing (scRNA-seq) data further explored HRTDGs’ roles at cellular level.</p> Results <p>Hypoxia was identified as a prominent feature of TED. Among all machine learning algorithms, random forest achieved the highest area under curve (AUC) and was used to identify three key HRTDGs: <i>EGFR</i>, <i>PIK3CB</i>, and <i>CREBBP</i>. The HRTDGS model was then established and found to be an independent predictive factor for TED diagnosis (odds ratio (OR): 2.656, 95% confidence interval (CI): 1.735–4.324, <i>p</i> &lt; 0.001). The model demonstrated high diagnostic accuracy in distinguishing TED from both HC (AUC = 0.785 in training set and 0.905 in testing set) and GH (AUC = 0.935). TED patients with higher HRTDGS exhibited elevated levels of thyrotropin receptor antibodies (TRAb) and abnormal free thyroxine (fT4), along with greater infiltration by activated CD4 + T cells and natural killer (NK) cells. ScRNA-seq revealed elevated expression of HRTDGs in fibroblasts, NK and CD4 + T cells, with enriched EGFR signaling pathway between T/NK cells and fibroblasts in TED compared to HC.</p> Conclusions <p>This study presents a novel hypoxia biomarkers-based diagnostic model for TED, facilitating early detection and offering valuable insights into potential therapeutic targets.</p> Graphic abstract <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integrative transcriptomic profiling and machine learning reveal hypoxia-associated molecular signatures for precision diagnosis in thyroid eye disease

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月24日

查看全文
  • Weijin Qian,
  • Tianyi Zhu,
  • Jin Liu,
  • Yining Wei,
  • Li Yang,
  • Lianfei Fang,
  • Jing Sun,
  • Yinwei Li,
  • Sijie Fang,
  • Huifang Zhou

摘要

Background

Thyroid eye disease (TED) is an autoimmune disorder characterized by persistent inflammation around the periphery and within the orbit, potentially driven by hypoxic conditions. Effective biomarkers and precise predictive models are still lacking for the early diagnosis of TED.

Methods

Bulk RNA sequencing was conducted on peripheral blood samples from TED patients, Graves’ hyperthyroidism (GH) patients without ocular involvement, and healthy controls (HC). Differentially expressed genes between TED and HC, hypoxia-related genes and genes identified through weighted gene co-expression network analysis (WGCNA) were intersected to identify candidate biomarkers. Subsequently, nine machine learning algorithms were applied to screen for critical hypoxia-related TED diagnostic genes (HRTDGs). A diagnostic model based on HRTDG score (HRTDGS) was constructed using logistic regression analyses and then evaluated. TED patients were categorized into high and low HRTDGS groups based on the median score. Distinct immunological profiles and underlying pathological functions were investigated between two groups. Single cell RNA sequencing (scRNA-seq) data further explored HRTDGs’ roles at cellular level.

Results

Hypoxia was identified as a prominent feature of TED. Among all machine learning algorithms, random forest achieved the highest area under curve (AUC) and was used to identify three key HRTDGs: EGFR, PIK3CB, and CREBBP. The HRTDGS model was then established and found to be an independent predictive factor for TED diagnosis (odds ratio (OR): 2.656, 95% confidence interval (CI): 1.735–4.324, p < 0.001). The model demonstrated high diagnostic accuracy in distinguishing TED from both HC (AUC = 0.785 in training set and 0.905 in testing set) and GH (AUC = 0.935). TED patients with higher HRTDGS exhibited elevated levels of thyrotropin receptor antibodies (TRAb) and abnormal free thyroxine (fT4), along with greater infiltration by activated CD4 + T cells and natural killer (NK) cells. ScRNA-seq revealed elevated expression of HRTDGs in fibroblasts, NK and CD4 + T cells, with enriched EGFR signaling pathway between T/NK cells and fibroblasts in TED compared to HC.

Conclusions

This study presents a novel hypoxia biomarkers-based diagnostic model for TED, facilitating early detection and offering valuable insights into potential therapeutic targets.

Graphic abstract