Figure 5.

Three machine learning methods were utilized to identify hub genes. (A) LASSO coefficient diagram. Each curve represents one specific gene. (B) The cross-validation curve shows the best λ value, and the X-axis above corresponds to the number of variables. (C, D) Visualization of the SVM-RFE results demonstrates that the accuracy and error of the model vary with the number of features. The features with the highest accuracy and lowest error were used for subsequent studies. (E) RF analysis was used to identify the genes with the highest accuracy. (F) Boxplot showing the differential expression of 8 genes. (**p < 0.01; ***p < 0.001; ****p < 0.0001, Student’s t test). (G) The ROC curves of ANPEP, ASMT, CA1, CKMT1B, MINPP1, PDE2A, PLA2G4A, and SDS. (H) Venn diagram depicting the candidate hub genes obtained by comparing the results of three machine learning algorithms.