Diabetic retinopathy (DR) is a microvascular complication of diabetes, affecting 34.6% of diabetes patients worldwide. Early detection and timely treatment can effectively improve the prognosis of DR. Metabolomic analysis provides a powerful tool for studying pathophysiological processes. Conducting metabolomic analyses on DR-related biofluids helps identify differential metabolic expressions during disease progression, thereby discovering potential biomarkers to support clinical diagnosis and treatment. Here, an innovative workflow for vitreous liquid analysis is established, and a machine learning-based DR analysis platform integrating vitreous liquid metabolic fingerprint (VL-MF) and plasma metabolic fingerprint (P-MF) derived via nanoparticle enhanced laser desorption/ionization mass spectrometry is developed. Direct VL-MF and P-MF are obtained with desirable reproducibility (coefficient of variation, CV <5%) and remarkable speed (3s per sample), and DR patients are distinguished from healthy controls applying dual biofluid-MF with an area under the curve (AUC) of 0.957. Moreover, a biomarker candidate panel from vitreous liquid and plasma with an AUC of 0.945 is constructed and the related metabolic pathways are identified by metabolomics pathway analysis (MetPA). This work offers a powerful multi-biofluid platform that can not only contribute to DR but also provide solid references for other clinical applications.
Read full abstract