Integrative transcriptomic and machine learning analysis identifies key extracellular matrix-related genes in...

Area of Science:

• Ophthalmology
• Genetics
• Biochemistry

Background:

• Diabetic retinopathy (DR) is characterized by extracellular matrix (ECM) remodeling and retinal vascular basement membrane thickening.
• Identifying key ECM-related genes (ECMGs) is crucial for understanding DR pathogenesis.

Purpose of the Study:

• To identify key ECM-related genes (ECMGs) associated with diabetic retinopathy (DR).
• To explore the mechanisms underlying the regulation of DR by identified ECMGs.
• To investigate potential therapeutic targets for DR based on ECMG interactions.

Main Methods:

• Integrated transcriptomic data of DR with ECMGs from MatrixDB to identify differentially expressed ECMGs.
• Employed six machine learning (ML) models for DR classification and SHapley Additive exPlanation (SHAP) for feature interpretation.
• Conducted functional enrichment analysis (GSEA), immune infiltration analysis (CIBERSORT), and constructed regulatory networks (ENCORI, miRWalk, miRNet).

Main Results:

• Identified nine candidate ECMGs, with five key genes (SFRP1, CILP2, FN1, TPSAB1, ECM2) selected based on ML model performance and SHAP values.
• These key ECMGs exhibited distinct expression patterns in healthy, diabetes mellitus (DM), and DR groups and were enriched in neural-related pathways.
• Correlations were found between ECMGs (FN1, TPSAB1, CILP2) and immune cells, and regulatory networks linked key ECMGs to miRNAs, lncRNAs, TFs, drugs, and DM-related diseases.

Conclusions:

• SFRP1, CILP2, FN1, TPSAB1, and ECM2 are identified as key ECMGs in DR pathogenesis.
• These genes play a coordinated role in ECM remodeling, neural signaling, and immune modulation in DR.
• The findings offer novel insights into DR and highlight potential therapeutic targets for the condition.