Fig 1.
Batch effect correction and data integration.
(A) Principal component analysis (PCA) plot showing the distribution of samples from GSE60436 and GSE102485 before batch effect removal. (B) PCA plot after batch effect removal using the sva package, demonstrating successful data integration.
Fig 2.
Identification and functional enrichment of immune-related DEGs in DR.
(A) Volcano plot of differentially expressed genes (DEGs) between DR and normal retinal samples. Red dots: significantly upregulated genes (log2FC > 1, p < 0.05); blue dots: significantly downregulated genes (log2FC < −1, p < 0.05); gray dots: non-significant genes. (B) Venn diagram showing the intersection between all DEGs and immune-related genes, yielding 123 immune-related DEGs. (C) Heatmap of the top DEGs. Columns represent samples (blue: DR; pink: normal); rows represent genes (red: high expression; blue: low expression). (D-F) Gene Ontology (GO) enrichment analysis of immune-related DEGs, showing top enriched terms in biological processes (D), cellular components (E), and molecular functions (F). Bar length indicates gene count, and color represents adjusted p-value.
Fig 3.
Gene set enrichment analysis (GSEA) reveals immune pathway activation in DR.
(A-H) Enrichment plots showing significantly enriched pathways in DR samples compared to normal controls. Normalized enrichment score (NES) and p-value are shown for each pathway. Pathways include cytokine-cytokine receptor interaction (A), T-cell receptor signaling (B), NK cell-mediated cytotoxicity (C), leukocyte transendothelial migration (D), Toll-like receptor signaling (E), Fc gamma R-mediated phagocytosis (F), IL-17 signaling (G), and neutrophil extracellular trap formation (H).
Fig 4.
Construction and validation of a diagnostic model for DR.
(A) LASSO regression coefficient profiles of immune-related genes. (B) Cross-validation plot for tuning parameter (λ) selection in the LASSO model. (C) Nomogram for predicting DR risk based on five key genes (PLAUR, PLAU, VTN, IL10RA, VGF). (D) Calibration curve assessing the predictive accuracy of the nomogram. The diagonal line represents ideal prediction; the pink line represents the nomogram performance. (E-I) Receiver operating characteristic (ROC) curves and area under the curve (AUC) values for each of the five key genes in discriminating DR from normal samples.
Fig 5.
Correlation, expression, and interaction analysis of key diagnostic genes.
(A-B) Friends analysis showing semantic similarity among the five key genes. (C) Expression levels of the five key genes in DR and normal retinal samples. *p < 0.05, **p < 0.01, ***p < 0.001. (D) Correlation matrix of the five key genes across all samples. Numbers indicate Pearson correlation coefficients; blue represents positive correlation, red represents negative correlation. (E) Protein-protein interaction (PPI) network of genes associated with DR, constructed using the STRING database. Nodes represent proteins; edges represent interactions.
Fig 6.
Immune cell infiltration profiles in DR and correlation with key genes.
(A) Relative proportions of 22 immune cell subtypes in each DR and normal sample, estimated by CIBERSORT. (B) Comparison of immune cell infiltration abundance between DR and normal groups. Red: DR samples; blue: normal samples. *p < 0.05, **p < 0.01. (C) Correlation heatmap between the five key genes and infiltrating immune cell types. Red: positive correlation; blue: negative correlation. Color intensity reflects correlation strength.
Fig 7.
Identification of two molecular subtypes of DR based on immune-related genes.
(A) Consensus clustering matrix for k = 2, showing clear separation of samples into two subtypes (Cluster A and Cluster B). (B) Consensus cumulative distribution function (CDF) plot for k = 2 to k = 9, indicating that k = 2 is the optimal choice. (C) Volcano plot of DEGs between Cluster A and Cluster B. Red: upregulated genes in Cluster B (log2FC > 1, p < 0.05); green: downregulated genes in Cluster B (log2FC < −1, p < 0.05); gray: non-significant genes. (D) Expression levels of the five key genes in Cluster A and Cluster B. Blue: Cluster A; yellow: Cluster B. *p < 0.05, **p < 0.01. (E) GSEA comparing Cluster A and Cluster B, showing enrichment of immune activation pathways in Cluster B. (F) ssGSEA scores for KEGG pathways in Cluster A and Cluster B, revealing distinct metabolic and immune signaling profiles between the two subtypes.
Fig 8.
Distinct immune infiltration patterns in DR molecular subtypes.
(A) Comparison of immune cell infiltration abundance between Cluster A and Cluster B. Blue: Cluster A; yellow: Cluster B. *p < 0.05, **p < 0.01. (B) Correlation matrix of immune cell types in all DR samples. Red: positive correlation; blue: negative correlation. Color intensity reflects correlation strength.
Fig 9.
Validation of key gene expression in an STZ-induced diabetic mouse model.
(A) Body weight changes in normal control (NC) and diabetic (DM) mice over 8 weeks post-STZ injection. (B) Representative Western blot images of PLAU, PLAUR, VGF, and GAPDH in retinal tissues from NC and DM mice. (C) Quantification of protein expression levels normalized to GAPDH. Data are presented as mean ± SD. *p < 0.05, **p < 0.01 compared to NC group.