Multi-Omics Integration With Machine Learning Identified Early Diabetic Retinopathy, Diabetic Macula Edema and...

Area of Science:

• Ophthalmology
• Metabolomics
• Biochemistry

Background:

• Diabetic retinopathy (DR) and diabetic macular edema (DME) are leading causes of vision loss in diabetes.
• Early detection and prediction of treatment response are crucial for effective management.

Purpose of the Study:

• To identify metabolic and lipidomic features associated with DR stages.
• To develop machine learning models for differentiating DME and predicting anti-vascular endothelial growth factor (anti-VEGF) therapy response.

Main Methods:

• Analysis of aqueous humor samples from type 2 diabetes mellitus patients and healthy controls using ultra-high-performance liquid chromatography-high-resolution-mass spectrometry.
• Application of machine learning to screen metabolic features and build predictive models for DR, DME, and anti-VEGF response.

Main Results:

• Key metabolic markers for DR, early-DR, and DME were identified, including n-acetyl isoleucine, cis-aconitic acid, and L-kynurenine.
• Predictive models demonstrated high accuracy (R² up to 99.9%) in the calibration set and performed well in validation.
• Trigonelline and 4-methylcatechol-2-sulfate predicted a strong response to anti-VEGF therapy.

Conclusions:

• Machine learning effectively identified differential metabolic features in DR patients.
• Metabolic indicators can predict early DR progression and identify potential non-responders to anti-VEGF therapy in DME.