Dual-center study identifies blood-based biomarkers for diabetic kidney disease diagnosis and progression. Machine learning models provide robust predictive performance.
Key Points
Logistic regression model achieved AUCs of 0.920 for diagnosing DKD, leading to significant predictive accuracy.
Four biomarkers including 1,5-AG and multiple fatty acids were identified as critical for DKD diagnosis.
External validation demonstrated robust performance in predicting diabetic kidney disease progression based on identified biomarkers.
Correlation analysis revealed complex relationships between biomarkers and renal function, highlighting risk factors for DKD.