Machine learning-based prediction of diagnostic markers for Graves' orbitopathy

Area of Science:

• Endocrinology and Immunology
• Molecular Biology
• Genetics

Background:

• Graves' orbitopathy (also known as thyroid-associated orbitopathy or TAO) pathogenesis remains unclear.
• Abnormal DNA methylation patterns have been observed in TAO.
• Identifying TAO biomarkers linked to DNA methylation could reveal new therapeutic targets.

Purpose of the Study:

• To identify and explore DNA methylation-associated biomarkers for TAO.
• To analyze the functional pathways and diagnostic value of identified biomarkers.
• To investigate the correlation between biomarkers and immune cell infiltration in TAO.

Main Methods:

• Downloaded and analyzed TAO expression and methylation data from the Gene Expression Omnibus database.
• Utilized weighted gene co-expression network analysis, intersected differentially methylated genes (DMGs) with differentially expressed genes to identify TAO-associated DMGs (TA-DMGs).
• Screened diagnostic markers using Lasso regression and Support Vector Machine (SVM) analysis, followed by pathway enrichment and immune infiltration analysis, and validated with qRT-PCR.

Main Results:

• Identified 125 TA-DMGs primarily involved in immune-related pathways.
• Screened NKD2 and S100A11 as diagnostic biomarkers for TAO.
• NKD2 showed a strong positive correlation with CD56 dim natural killer cells, and both biomarkers' expressions were validated by qRT-PCR.

Conclusions:

• NKD2 and S100A11 are identified as potential biomarkers for TAO.
• These biomarkers may be regulated by DNA methylation in TAO.
• This finding offers a new reference for the diagnosis and treatment of TAO patients.