Objective:To provide theoretical basis for the relevant mechanisms in diabetic retinopathy by bioinformatics methods.Methods:Proliferative diabetic retinopathy(PDR)datasets GSE60439 and GSE94019 were screened through GEO database.Expres-sion data of GSE60439 dataset was extracted using R language,and the differential genes were determined,with the screening condi-tions of P<0.05 and|log2FC|≥1;WGCNA was used to select the important modules related to PDR in GSE94019 dataset;key genes were obtained by intersecting differential genes and genes in important modules.Key genes were subjected to GO and KEGG enrich-ment analysis;PPI network was constructed to identify high-risk key proteins.Finally,CIBERSORTx was used to analyze the infiltra-tion of immune cells in PDR,and Pearson correlation analysis was used to screen key genes related to immune cells as core genes.Results:Three core genes were finally obtained,namely Col1a1,Col1a2 and Col3a1,their expressions were positively correlated with M2 macrophages,and the core genes were enriched in the AGEs-RAGE signaling pathway.Conclusion:Positive feedback may be formed between M2 macrophages,AGEs-RAGE signaling pathway and collagen genes(Col1a1,Col1a2,Col3a1),even without the stimulation of high glucose,this vicious circle may continue to damage retinal tissue.