机器学习在CKD基因诊断中的应用

We examined the GSE60861 dataset and found that it consists of two subsets with different data sizes, namely GSE45980 and GSE60860. In GSE45980 (N=43), we observed that the minimal z-score has moderate diagnostic capability for CKD progression (AUC: 0.758). In the future, we can use various machine learning algorithms, including SHAP, to screen for risk genes that make significant contributions, which may further improve the diagnostic performance of the core gene set.