Predictive model for proliferative diabetic retinopathy using single-cell transcriptomics

Proliferative diabetic retinopathy (PDR) is a severe and advanced complication of diabetic retinopathy, representing a leading cause of blindness worldwide. A hallmark of PDR is the formation of fibrovascular membranes (FVMs), which pose a significant threat to visual function due to their propensity to contract and exert traction on the retina. A comprehensive understanding of the underlying molecular mechanisms is essential for addressing this condition effectively. Herein, we first integrated scRNA-seq data from nine FVMs of PDR patients and retinas from healthy donors obtained from the Gene Expression Omnibus (GEO) database. Differential gene expression analysis was then performed on astrocytes, microglia, and rod cells from PDR patients and healthy donors. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses revealed distinct enrichment patterns specific to each cell type in PDR. Additionally, a machine learning-based predictive model was developed as a tool for the detection and monitoring of PDR, achieving an area under the curve AUC of 0.81 in the test dataset and 0.83-0.96 in the validation cohorts. Therefore, this study provides new insights into the cellular and molecular mechanisms underlying FVM development, underscore the potential for detection, monitoring and identifying targets of PDR.