Fig. 4
From: Investigating immune cell infiltration and gene expression features in pterygium pathogenesis
Selection of candidate diagnostic biomarkers of pterygium with machine learning approaches. (A-B) LASSO regression analysis was employed to identify diagnostic biomarkers. (C) Diagnostic errors associated with the conjunctiva, pterygium, and total groups were visualized using the random forest model. (D) A column displaying the top 15 DEGs ranked according to their importance scores derived from the random forest analysis. (E) The DEGs with the lowest error rate and highest accuracy after 10-fold cross-validation were selected as the most suitable candidates through the SVM-RFE algorithm. (F) The intersection of the results from the three machine learning algorithms was illustrated using a Venn diagram tool.
