Identification of potential biomarkers of myopia based on machine learning algorithms

Abstract

Abstract Purpose This study aims to identify potential myopia biomarkers using machine learning algorithms, enhancing myopia diagnosis and prognosis prediction. Methods GSE112155 and GSE15163 datasets from the GEO database were analyzed. We used “limma” for differential expression analysis and “GO plot” and “clusterProfiler” for functional and pathway enrichment analyses. The LASSO and SVM-RFE algorithms were employed to screen myopia-related biomarkers, followed by ROC curve analysis for diagnostic performance evaluation. Single-gene GSEA enrichment analysis was executed using GSEA 4.1.0. Results The functional analysis of differentially expressed genes indicated their role in carbohydrate generation and polysaccharide synthesis. We identified 23 differentially expressed genes associated with myopia, four of which were highly effective diagnostic biomarkers. Single gene GSEA results showed these genes control the ubiquitin-mediated protein hydrolysis pathway. Conclusion Our study identifies four key myopia biomarkers, providing a foundation for future clinical and experimental validation studies.

Keywords

• Myopia
• Machine learning
• Biomarkers
• Gene expression
• Diagnosis