A ten-gene retinal pigment epithelium (RPE)/choroid complex diagnosis signature for age-related macular degeneration

Abstract

Abstract Purpose: Age-related macular degeneration (AMD) is a multifactorial disease in the elderly with a prominent genetic basis. This study aimed to apply machine learning method to develop a novel diagnostic model for AMD based on gene biomarkers in RPE/choroid complex, which may be potential therapeutic targets. Methods: We collected RPE/choroid tissue gene expression profiles of AMD and normal patients from the Gene Expression Omnibus (GEO) database as training and validation cohorts. After differential expression analysis and the selection of gene biomarkers by random forest algorithms, selected genes were applied to the least absolute shrinkage and selection operator (LASSO) logistic regression to construct a diagnostic model in the training cohort. The diagnostic ability of the model was further tested in the validation cohort. Gene set enrichment analysis (GSEA) and immune cell assessment were also conducted for further analyses. Results：A noval diagnostic model based on ten genes (BMPR2, CNOT3, CRLF1, FXYD6, HRASLS5, KRTDAP, NUDT16L1, PI16, PLAGL1, SART1) was constructed in the training cohort. The AUC in the training cohort reached 0.908 (95% CI: 0.823-0.975), while it remained 0.809 (95% CI: 0.522-0.889) in the validation cohort. According to the GSEA analysis, glutathione metabolism and phototransduction pathway are the two shared enriched pathways in the training and validation cohorts. Functional enrichment analysis and immune cell evaluation demonstrated that AMD was significantly correlated with both adaptive and innate immune cells, and the levels of neutrophil in the high-risk group were significantly higher than that in the low-risk group in both training and validation datasets Conclusion: We identified and validated a novel ten-gene-based diagnostic model with high accuracy for AMD. The current study provided a promising tool to be used as a precise and cost-effective non-invasive test in clinical practice.