Based on multiple machine learning to identify the ENO2 as diagnosis biomarkers of glaucoma

Abstract

Abstract Purpose Glaucoma is a generic term of a highly different disease group of optic neuropathies, which the leading cause of irreversible vision in the world. There are few biomarkers available for clinical prediction and diagnosis, and the diagnosis of patients is mostly delayed. Methods Differential gene expression of transcriptome sequencing data (GSE9944 and GSE2378) for normal samples and glaucoma samples from the GEO database were analyzed. Furthermore, based on different algorithms (Logistic Regression (LR), Random Forest (RF), lasso regression (LASSO)) two diagnostic models are constructed and diagnostic markers are screened. GO and KEGG analyses revealed the possible mechanism of differential genes in the pathogenesis of glaucoma. ROC curve confirmed the effectiveness. Results LR-RF model included 3 key genes (NAMPT, ADH1C, ENO2), and the LASSO model outputted 5 genes (IFI16, RFTN1, NAMPT, ADH1C, and ENO2), both algorithms have excellent diagnostic efficiency. ROC curve confirmed that the three biomarkers ADH1C, ENO2, and NAMPT were effective in the diagnosis of glaucoma. Next, the expression analysis of the three diagnostic biomarkers in glaucoma and control samples confirmed that NAMPT and ADH1C were up-regulated in glaucoma samples, and ENO2 was down-regulated. Correlation analysis showed that ENO2 was significantly negatively correlated with ADH1C (cor = -0.865714202) and NAMPT (cor = -0.730541227). Finally, three compounds for the treatment of glaucoma were obtained in the TCMs database: acetylsalicylic acid, 7-o-methylisomucitol and scutellarin which were applied to molecular docking with the diagnostic biomarker ENO2. Conclusions In conclusion, our research shows that ENO2, NAMPT, and ADH1C can be used as diagnostic markers for glaucoma, and ENO2 can be used as a therapeutic target.

Keywords

• Diagnostic markers
• Logistic regression
• Random forest
• Lasso regression